CN113762043A - Abnormal track identification method and device - Google Patents
Abnormal track identification method and device Download PDFInfo
- Publication number
- CN113762043A CN113762043A CN202110482956.XA CN202110482956A CN113762043A CN 113762043 A CN113762043 A CN 113762043A CN 202110482956 A CN202110482956 A CN 202110482956A CN 113762043 A CN113762043 A CN 113762043A
- Authority
- CN
- China
- Prior art keywords
- node
- graph
- track
- nodes
- road section
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/088—Non-supervised learning, e.g. competitive learning
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Traffic Control Systems (AREA)
Abstract
The application provides an abnormal track identification method and device, which relate to the technical field of computers, and the method comprises the following steps: acquiring traveling track data to be identified, wherein the traveling track data is acquired in a target road section; extracting characteristics of the traveling track data to obtain a node connection sequence corresponding to the traveling track data; the node connection sequence comprises track nodes corresponding to the traveling track data and first directed connection edges among the track nodes; based on graph nodes in a road section graph model of the target road section and second directed connecting edges between the graph nodes, performing abnormity identification on a node connecting sequence corresponding to the traveling track data, and determining whether the traveling track data is abnormal; the road section graph model of the target road section is generated based on a historical node connection sequence corresponding to historical travel track data acquired from the target road section. Therefore, the abnormal track can be identified based on the model with lower complexity, and the complexity of identifying the abnormal track is reduced.
Description
Technical Field
The present application relates to the field of computer technologies, and in particular, to a method and an apparatus for identifying an abnormal trajectory.
Background
The abnormal track behavior identification means that: based on the road monitoring video data, the driving behavior of the vehicles is identified to be represented by track routes with time sequences, most vehicles have approximately the same driving track route, and the driving tracks of a small number of vehicles do not follow the driving tracks of most vehicles, for example: the traffic accident, and the like, which further affect the overall driving safety of the road, need to locate and identify these abnormal driving tracks from the monitoring video.
Disclosure of Invention
In view of the above, it is necessary to provide an abnormal trajectory recognition method, an abnormal trajectory recognition apparatus, a computer device, and a storage medium, which can automatically locate and recognize an abnormal travel trajectory.
An abnormal trajectory identification method, the method comprising:
the track data acquisition module is used for acquiring the traveling track data to be identified, wherein the traveling track data is acquired at a target road section;
the characteristic extraction module is used for extracting the characteristics of the traveling track data to obtain a node connection sequence corresponding to the traveling track data; the node connection sequence comprises track nodes corresponding to the traveling track data and first directed connection edges among the track nodes;
the abnormity identification module is used for carrying out abnormity identification on a node connection sequence corresponding to the traveling track data based on graph nodes in the road section graph model of the target road section and second directed connection edges between the graph nodes, and determining whether the traveling track data is abnormal or not; the road section graph model of the target road section is generated based on a historical node connection sequence corresponding to historical travel track data acquired from the target road section.
An abnormal trajectory recognition apparatus, the apparatus comprising:
acquiring traveling track data to be identified, wherein the traveling track data is acquired in a target road section;
extracting the characteristics of the traveling track data to obtain a node connection sequence corresponding to the traveling track data; the node connection sequence comprises track nodes corresponding to the traveling track data and first directed connection edges among the track nodes;
performing anomaly identification on a node connection sequence corresponding to the travel track data based on graph nodes in the road section graph model of the target road section and second directed connection edges between the graph nodes, and determining whether the travel track data is abnormal; the road section graph model of the target road section is generated based on a historical node connection sequence corresponding to historical travel track data acquired from the target road section.
A computer device comprising a memory and a processor, the memory storing a computer program, the processor implementing the following steps when executing the computer program:
acquiring traveling track data to be identified, wherein the traveling track data is acquired in a target road section;
extracting the characteristics of the traveling track data to obtain a node connection sequence corresponding to the traveling track data; the node connection sequence comprises track nodes corresponding to the traveling track data and first directed connection edges among the track nodes;
performing anomaly identification on a node connection sequence corresponding to the travel track data based on graph nodes in the road section graph model of the target road section and second directed connection edges between the graph nodes, and determining whether the travel track data is abnormal; the road section graph model of the target road section is generated based on a historical node connection sequence corresponding to historical travel track data acquired from the target road section.
A computer-readable storage medium, on which a computer program is stored which, when executed by a processor, carries out the steps of:
acquiring traveling track data to be identified, wherein the traveling track data is acquired in a target road section;
extracting the characteristics of the traveling track data to obtain a node connection sequence corresponding to the traveling track data; the node connection sequence comprises track nodes corresponding to the traveling track data and first directed connection edges among the track nodes;
performing anomaly identification on a node connection sequence corresponding to the travel track data based on graph nodes in the road section graph model of the target road section and second directed connection edges between the graph nodes, and determining whether the travel track data is abnormal; the road section graph model of the target road section is generated based on a historical node connection sequence corresponding to historical travel track data acquired from the target road section.
The abnormal track identification method, the abnormal track identification device, the computer equipment and the storage medium acquire the traveling track data to be identified, which is acquired at a target road section, and perform feature extraction on the traveling track data to obtain a corresponding node connection sequence, which comprises track nodes of a traveling track and a first directed connection edge between the track nodes; and performing abnormity identification on the node connection sequence based on a second directed connection edge between the graph node and the graph node in the road section graph model of the target road section, and determining whether the travel track data is abnormal. The graph model is generated based on historical node connection sequences corresponding to historical travel track data acquired from a target road section. The method converts the identification of the traveling track data collected on the target road section into the matching problem of the node connection sequence in the corresponding graph model, can realize the automatic abnormal identification of the traveling track data based on the model with simpler complexity, and reduces the complexity of the abnormal track identification.
Drawings
FIG. 1 is a diagram of an exemplary application environment for the method of abnormal trajectory recognition;
FIG. 2 is a flow diagram illustrating an abnormal trajectory recognition method according to an embodiment;
FIG. 3 is a schematic diagram of a node connection sequence obtained by extracting features from a trajectory curve in one embodiment;
FIG. 4 is a flowchart illustrating updating a road segment graph model according to a node connection sequence to obtain a new road segment graph model in one embodiment;
fig. 5 is a schematic flow chart illustrating a graph node set in which a road segment graph model is updated based on node feature distances and a second directed connecting edge between graph nodes in the updated graph node set to obtain a new road segment graph model in one embodiment;
FIG. 6 is a flow chart illustrating a method for identifying abnormal trajectories in another embodiment;
FIG. 7 is a schematic diagram illustrating a process for determining an initial road segment graph model of a target road segment in one embodiment;
fig. 8 is a flowchart illustrating a process of performing anomaly identification on a node connection sequence corresponding to travel track data based on graph nodes in a road segment graph model of a target road segment and second directed connection edges between the graph nodes in one embodiment, and determining whether the travel track data is anomalous;
FIG. 9 is a diagram illustrating initial trajectory nodes extracted using lane lines, in accordance with an embodiment;
FIG. 10(1) is a diagram illustrating a graphical model in one embodiment;
FIG. 10(2) is a schematic diagram of a graphical model in another embodiment;
FIG. 10(3) is a schematic diagram of a graphical model in another embodiment;
FIG. 10(4) is a schematic diagram of a graphical model in another embodiment;
FIG. 11 is a block diagram showing the structure of an abnormal trajectory recognition apparatus according to an embodiment;
FIG. 12 is a diagram illustrating an internal structure of a computer device according to an embodiment.
Detailed Description
In order to make the objects, technical solutions and advantages of the present application more apparent, the present application is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the present application and are not intended to limit the present application.
In some embodiments, the abnormal trajectory identification method provided by the present application may be applied to an application environment as shown in fig. 1. Among them, the terminals 102 and 103 communicate with the server 104 via a network. The server 104 acquires traveling track data to be identified, which is acquired at a target road section, from the terminal 102, and performs feature extraction on the traveling track data to obtain a corresponding node connection sequence, which comprises track nodes of a traveling track and a first directed connection edge between the track nodes; and performing abnormity identification on the node connection sequence based on a second directed connection edge between the graph node and the graph node in the road section graph model of the target road section, and determining whether the travel track data is abnormal. The graph model is generated based on historical node connection sequences corresponding to historical travel track data acquired from a target road section. Further, the server 104 may transmit the result of the abnormal trajectory recognition to the terminal 103. The terminal 102 may be, but not limited to, various devices having a camera function, such as a monitoring device disposed on a road, the terminal 103 may be, but not limited to, various personal computers, notebook computers, smart phones, tablet computers, and portable wearable devices, the server 104 may be an independent physical server, a server cluster or a distributed system formed by a plurality of physical servers, and a cloud server providing basic cloud computing services such as cloud service, a cloud database, cloud computing, a cloud function, cloud storage, network service, cloud communication, middleware service, domain name service, security service, CDN, and a big data and artificial intelligence platform.
Cloud computing (cloud computing) is a computing model that distributes computing tasks over a pool of resources formed by a large number of computers, enabling various application systems to obtain computing power, storage space, and information services as needed. The network that provides the resources is referred to as the "cloud". Resources in the "cloud" appear to the user as being infinitely expandable and available at any time, available on demand, expandable at any time, and paid for on-demand. Cloud Computing is a product of development and fusion of traditional computers and Network Technologies, such as Grid Computing (Grid Computing), distributed Computing (distributed Computing), Parallel Computing (Parallel Computing), Utility Computing (Utility Computing), Network Storage (Network Storage Technologies), Virtualization (Virtualization), Load balancing (Load Balance), and the like.
As a basic capability provider of cloud computing, a cloud computing resource pool (called as an ifas (Infrastructure as a Service) platform for short is established, and multiple types of virtual resources are deployed in the resource pool and are selectively used by external clients.
With the development of diversification of internet, real-time data stream and connecting equipment and the promotion of demands of search service, social network, mobile commerce, open collaboration and the like, cloud computing is rapidly developed.
In other embodiments, the abnormal trajectory identification method provided by the present application may also be applied to a calculation unit on a drive test probe, in this embodiment, a road segment graph model generated according to historical travel trajectory data is stored in the calculation unit of the drive test probe, and then the travel trajectory data generated in a target road segment by an object acquired by the drive test probe in real time is identified by graph nodes in the road segment graph model and second directed connecting edges between the graph nodes.
The terms appearing in the examples of the present application are defined:
unsupervised: during the model learning process, truth marking is not needed, and the algorithm can automatically explore the internal relation among data during the operation process.
The following drawings: the graph is a data structure consisting of nodes and edges, wherein each node represents metadata, and the edges represent whether the metadata have connection relations or not.
Directed graph: the connection relationship of the edges has directionality, i.e., node a and B connections and node B and a connections are different.
An authorized graph: a graph with weight values on the edges.
And (3) online learning: the model is based on the input of a real-time data stream, namely the model can only obtain data of the current time and the past time at any time t and cannot obtain data of the future time.
And (3) abnormal track identification: in the road monitoring video data, the driving behavior of the vehicles is represented by track routes with time series, most vehicles have approximately the same driving track route, and the driving tracks of a small number of vehicles do not follow the driving tracks of most vehicles, for example: the traffic accident, and the like, which further affect the overall driving safety of the road, need to locate and identify these abnormal driving tracks from the monitoring video.
In one embodiment, as shown in fig. 2, an abnormal trajectory identification method is provided, which is described by taking the method as an example applied to the server in fig. 1, and includes steps S210 to S230.
Step S210, acquiring the travel track data to be recognized. The travel track data are acquired in the target road section.
The travel track data represent tracks generated by the same object traveling in the target road section; the object is the object whether the travel track required to be monitored is abnormal or not. The travel track data to be recognized represents a track which is generated by an object in a target road section and needs to be subjected to abnormal recognition, for example, track data generated when a vehicle passes through the track is acquired in the target road section. In one embodiment, the object may be a motor vehicle in the target road segment, may also be a non-motor vehicle in the target road segment, or may also be a pedestrian in the target road segment, or the like.
Further, in one embodiment, the track data to be identified, which is acquired by the same object on the target road section, needs to be obtained by combining a target detection and tracking method; the target detection and tracking may be implemented in any manner, for example, the target detection and tracking of the object is implemented through a target detection and tracking model determined through training, so as to obtain the travel track data of the object in the target road segment.
The target road section is a road section which needs to be monitored and abnormal track identification; for example, the target road segment is a road segment within a certain intersection; or the target road section can also be a target road section corresponding to the determined range in the image through image recognition. The track refers to a route which is a track of the movement of the object when the object moves according to a certain rule, and in the embodiment, the travel track data represents a travel track generated by the object in the target road section.
In one embodiment, the travel track data to be recognized may be directly obtained from a monitoring camera disposed in the road, or may be obtained from a storage device connected to the road monitoring camera for storing the travel track data obtained from the road monitoring camera. In other embodiments, the travel track data to be recognized may be acquired in other manners.
Further, in one embodiment, the acquired travel track data of the object in the target road segment is represented in a continuous image sequence, a video form or a track curve.
Step S220, performing feature extraction on the travel track data to obtain a node connection sequence corresponding to the travel track data.
The node connection sequence comprises track nodes corresponding to the traveling track data and first directed connection edges between the track nodes.
In machine learning, pattern recognition and image processing, feature extraction starts with an initial set of measurement data and establishes derivative values (features) that are intended to provide information and non-redundancy, facilitating subsequent learning and generalization steps, and in some cases leading to better interpretability. In this embodiment, feature extraction is performed on the acquired travel track data to obtain a node connection sequence corresponding to the travel track data, specifically including extracting track nodes in the travel track data and connecting directed connecting edges between every two track nodes; in order to distinguish from the subsequent technical features, the directional connecting edge extracted from the travel track data is recorded as a first directional connecting edge in this embodiment. As shown in fig. 3, in a specific embodiment, taking a curve representing travel track data as an example, a node connection sequence obtained by extracting features from the curve includes a track node (a point shown in the graph) and a directed connection edge (a connection line segment between the point shown in the graph and the point) connecting the track node, where the directed connection edge represents a connection direction between the node and the node, and may be a directed line segment between two nodes.
In one embodiment, feature extraction of the travel track data may be performed in any manner. For example, in a specific embodiment, a sliding window may be set, the sliding window slides in the travel track data according to the time sequence, and for the curvatures corresponding to the plurality of point positions on the calculated track in the obtained window, the track point corresponding to the position with the largest curvature in the window is used as the track node in the window, and so on. The position of a plurality of points selected in the track can be randomly selected, and the curvature corresponding to the position of the point can be represented as: and taking the position of the central point as a central point, calculating one half of a track curve between the central point and the previous point and one half of a track curve between the central point and the next point, and calculating the curvatures of the front part and the rear part of the track of the central point.
In one embodiment, the track nodes in the node connection sequence obtained by feature extraction carry node feature information. Further, in one embodiment, the node characteristic information includes at least one of a position of the node, a curvature of the position of the node in the corresponding travel track data, and a track length represented by the position of the node. In a specific embodiment, the position of the node can be represented in the form of coordinates or in the form of latitude and longitude. The length of the trace represented by the position of the node may be: and the sum of one half of the corresponding track length between the position of the node and the previous adjacent track node and one half of the corresponding track length between the position of the node and the next adjacent track node. In one embodiment, the curvature of the position of the node in the corresponding travel track data represents the curvature of the position of the node in the track length represented by the position of the node; the curvature corresponding to the location of the node may be determined in any manner. With continued reference to fig. 3, the nodes A, B and C are sequentially connected, wherein the length of the trajectory represented by the position of the node B is calculated by: x1, which is one-half the length of the track between node B and node A, and X2, which is one-half the length of the track between node B and node C, (X1+ X2) indicates the length of the track represented by the location of node B. Further, the curvature of the position of the node B in the trajectory (X1+ X2) represents the curvature of the node B position in the corresponding travel trajectory data.
Step S230, based on the graph nodes in the road segment graph model of the target road segment and the second directed connecting edges between the graph nodes, performing anomaly identification on the node connection sequence corresponding to the travel track data, and determining whether the travel track data is abnormal.
The road section graph model of the target road section is generated based on a historical node connection sequence corresponding to historical travel track data acquired from the target road section.
The Graphic Models (Graphic Models) refer to graphs composed of points and lines to describe the system. In the present embodiment, the link map model of the target link represents a graph in which points and lines in the target link connect a road structure for describing the target link. The road section graph model of the target road section comprises graph nodes and second directed connecting edges between the graph nodes; a road segment map model of the target road segment is generated from historical travel track data collected within the target road segment. Furthermore, historical traveling track data are collected on the target road section, feature extraction is carried out on the historical traveling track data to obtain a corresponding historical node connection sequence, and a road section graph model corresponding to the target road section is generated based on the historical node connection sequence. The process of generating the road segment graph model based on the node connection sequence will be described in detail in the following embodiments, and will not be described herein again.
In this embodiment, the link map model of the target link includes: and directed connecting edges between the graph nodes, namely second directed connecting edges. Furthermore, according to the graph nodes in the road section graph model and the second directed connecting edge, the node connection sequence corresponding to other travel track data generated in the target road section can be subjected to abnormity identification, and whether the travel track data is abnormal or not is judged. In one embodiment, the nodes in the node connection sequence and the first directed connection edges are subjected to abnormal recognition according to the graph nodes and the second directed connection edges in the road segment graph model, and an abnormal recognition result of the traveling track data is obtained. The specific process of how to perform the abnormal recognition on the node connection sequence according to the graph node and the second directed connection edge will be described in detail in the following embodiments, which is not described herein again.
The abnormal track identification method comprises the steps of acquiring traveling track data to be identified, which are acquired at a target road section, and extracting characteristics of the traveling track data to obtain a corresponding node connection sequence, wherein the node connection sequence comprises track nodes of a traveling track and a first directed connection edge between the track nodes; and performing abnormity identification on the node connection sequence based on a second directed connection edge between the graph node and the graph node in the road section graph model of the target road section, and determining whether the travel track data is abnormal. The graph model is generated based on historical node connection sequences corresponding to historical travel track data acquired from a target road section. The method converts the identification of the traveling track data collected on the target road section into the matching problem of the node connection sequence in the corresponding graph model, can realize the automatic abnormal identification of the traveling track data based on the model with simpler complexity, and reduces the complexity of the abnormal track identification.
Further, in an embodiment, after performing feature extraction on the travel track data to obtain a node connection sequence corresponding to the travel track data, the method further includes: and updating the road section graph model according to the node connection sequence to obtain a new road section graph model.
In this embodiment, not only is the node connection sequence corresponding to the travel track data to be recognized identified abnormally identified by using the road segment graph model, but also the road segment graph model is updated based on the nodes of the node track sequence in the travel track data to be recognized and the first directed connecting edge. Further, the step of updating the road segment graph model based on the nodes in the node connection sequence and the first directed connecting edge may be performed before the node connection sequence corresponding to the travel track data to be recognized is abnormally identified by using the road segment graph model, or may be performed after the node connection sequence corresponding to the travel track data to be recognized is abnormally identified by using the road segment graph model.
In one embodiment, the road segment graph model is updated based on the node connection sequence corresponding to the travel track data to be recognized, and the method comprises two modes of online updating and offline updating. Wherein updating the road segment graph model representation on-line is: when an object moves in a target road section, gradually generating travel track data to be identified, extracting characteristics of the generated partial travel track data to obtain corresponding track nodes and first directed connecting edges between the track nodes, and updating the extracted track nodes and the first directed connecting edges into a road section graph model. Updating the road segment graph model representation offline: when the object runs through the target road section, generating complete traveling track data to be recognized, then performing feature extraction on the complete traveling track data to obtain track nodes and first directed connecting edges corresponding to the complete traveling track data, and further updating the road section graph model according to all the track nodes and the first directed connecting edges.
Furthermore, because the travel track data needs to be identified completely due to the abnormal identification of the travel track data, in the embodiment of updating the road segment graph model on line, the road segment graph model is updated before the abnormal track identification; in the embodiment of updating the road segment map model offline, the updating of the road segment map model may be performed before the abnormal trajectory identification or after the abnormal trajectory identification.
In one embodiment, as shown in fig. 4, the road segment graph model is updated according to the node connection sequence, and a new road segment graph model is obtained, including steps S410 to S430. Wherein:
step S410, a graph node set of the road section graph model of the target road section and a second directed connecting edge between graph nodes in the graph node set are obtained.
The graph node set represents a set formed by graph nodes in the road segment graph model, and connecting edges among the graph nodes in the graph node set are marked as second directed connecting edges.
Step S420, aiming at each track node in the node connection sequence, determining the characteristic distance of the track node and the graph node in the graph node set of the road section graph model.
The node feature distance represents a distance between node features of the nodes. In one embodiment, determining, for each track node in the sequence of node connections, a node feature distance between the track node and a graph node in a set of graph nodes of the road segment graph model comprises: for each track node in the node connection sequence, determining a node feature distance between the track node and a graph node in the road segment graph model based on the node feature of the track node and the node feature of the graph node in the road segment graph model. Further, in one embodiment, the node characteristics of the track nodes include at least one of node positions, curvatures of the node positions in the corresponding travel track data, and track lengths represented by the node positions. Because the graph nodes in the road segment graph model are also generated according to the historical node connection sequence, the graph nodes carry graph node characteristics, and the node characteristics also comprise at least one of the positions of the nodes, the curvatures of the node positions in the corresponding travel track data and the track lengths represented by the node positions.
In one embodiment, the node feature includes two or more of the node position, the curvature of the node position in the corresponding travel track data, and the track length represented by the node position, the node feature distance of the two nodes is calculated, a corresponding weight may be set for each node feature, and the weighted node feature distance of each node feature is calculated as the node feature distance between the two nodes. In another embodiment, the node features include more than two, and calculating the node feature distance between two nodes can calculate the feature distance of each node feature respectively, and averaging to obtain the node distance between two nodes; if calculating the node characteristic distance between the track node A and the graph node X comprises: calculating the characteristic distance of the node position between the track node A and the graph node X, calculating the characteristic distance of the node representing the curvature between the track node A and the graph node X, calculating the distance characteristic of the node representing the length between the track node A and the graph node X, and then calculating the average value of the distance characteristics of the position, the curvature and the length as the node characteristic distance between the track node A and the graph node X. In one embodiment, calculating the feature distance of the two node features may be accomplished in any one of a number of ways.
And step S430, updating a graph node set of the road segment graph model based on the node characteristic distance, and obtaining a new road segment graph model by using a second directed connecting edge between graph nodes in the updated graph node set.
Further, in an embodiment, as shown in fig. 5, the step of updating the graph node set of the road segment graph model based on the node characteristic distance and the second directed connecting edge between graph nodes in the updated graph node set to obtain a new road segment graph model includes steps S431 to S433. Wherein:
and step S431, carrying out node aggregation processing on the target track nodes and the target graph nodes meeting the node aggregation condition based on the node characteristic distance, and updating the target graph nodes.
Wherein the node aggregation condition represents a condition for determining an aggregation node; in this embodiment, the node characteristic distance between the track node and the graph node is smaller than a preset distance threshold, and it is determined that the node aggregation condition is satisfied. The preset distance threshold value can be set to any value according to actual conditions. Further, in one embodiment, the preset distance threshold comprises: a preset position distance threshold, a preset curvature distance threshold, and a preset length distance threshold.
When the node aggregation condition is met between the track node in the node connection sequence and the graph node in the road segment graph model, the track node meeting the node aggregation condition is recorded as a target track node, the graph node meeting the node aggregation condition is recorded as a target graph node, and then node aggregation processing is performed on the target track node and the target graph node.
After the node distances between each track node in the node connection sequence and the graph nodes in the road section graph model are respectively determined, the node characteristic distances of any track node A and any graph node can be sequenced according to the distance, the graph node with the minimum characteristic distance value is used for judging whether the node aggregation condition is met or not, and if the minimum characteristic distance value is smaller than a preset distance threshold value, the node aggregation condition is judged to be met between the track node A and the graph node corresponding to the characteristic distance value.
In one embodiment, the node aggregation processing is performed on the target track nodes and the target graph nodes, and the updating of the target graph nodes includes: and updating the node characteristics of the target track nodes into the node characteristics of the target graph nodes. In one embodiment, updating the node characteristics of the target track node to the node characteristics of the target graph node comprises: and acquiring a first weight corresponding to the newly added target node characteristics and a second weight corresponding to the graph nodes in the road section graph model, and fusing the node characteristics of the target track nodes and the node characteristics of the target graph nodes by using the node characteristics of the target track nodes and the weight corresponding to the node characteristics of the target graph nodes based on the first weight and the second weight. In a specific embodiment, the weight of the newly added target track node is set to 0.1, the weight of the target graph node is set to 0.9, and when the node feature connected with the target track is updated to the node feature of the target graph node, the updated node feature is equal to (0.1 +0.9 of the node feature of the target track node). In other embodiments, updating the node characteristics of the target track node to the node characteristics of the target graph node may also be implemented in other ways.
Step S432, aiming at the non-target track nodes which do not meet the node aggregation condition with any graph node, creating graph nodes corresponding to the non-target track nodes in the road section graph model.
If the track node in the node connection sequence and any graph node do not meet the node aggregation condition, determining the track node as a non-target track node; in one embodiment, if the minimum value of the node characteristic distances between the track node and the graph node is still greater than or equal to the preset distance threshold, the track node is determined to be a non-target track node. In this embodiment, for a non-target track node, a graph node corresponding to the non-target track node is created in a set of graph nodes in the road segment graph model. In one embodiment, node characteristics of non-target trajectory nodes are added to the set of graph nodes.
And step S433, updating the graph node set of the road section graph model and a second directed connecting edge between each graph node in the updated graph node set based on at least one of the updated target graph node and the graph node corresponding to the non-target track node, and obtaining a new road section graph model.
In this embodiment, the node connection sequence is divided into target track nodes and non-target track nodes according to the node characteristic distance: if the target trace nodes are the target trace nodes, the target trace nodes are gathered to the corresponding target graph nodes; and if the node is the non-target track node, creating a new graph node corresponding to the non-target track node in the graph node set. And further, after the graph node set is updated, the second directed connecting edges between the graph nodes can be updated as well. In one embodiment, updating the second directed connecting edge based on at least one of the updated target graph node and the non-target track node corresponding graph node comprises: if the two adjacent track nodes are target track nodes, the weight of a second directed connecting edge between the corresponding target graph nodes in the updated graph node set of the two adjacent target track nodes is + 1; if at least one non-target track node is included in the two adjacent track nodes, a second directed connecting edge is established between the graph nodes corresponding to the two adjacent track nodes in the updated graph node set on the basis of the first directed connecting edge between the two adjacent track nodes, and the weight is set to be 1. And finally obtaining an updated graph node set and a second directed connecting edge, namely an updated road section graph model.
Wherein, in a specific embodiment, assuming that the graph node model includes graph nodes A, B and C, the second directed connecting edge includes A → B, B → C. When the node connection sequence is acquired: a1 → B1 → …, assuming that A1 in the node connection sequence is that the node aggregation condition is satisfied between the target track node and the target graph node A, and the node aggregation condition is satisfied between the target track node B1 and the target graph node B, performing aggregation processing on the target track node A1 and the target graph node A, and performing aggregation processing on the target track node B1 and the target graph node B to obtain an updated graph node B; further, based on the updated graph node a and the updated graph node B, the updated second directed connecting edge is obtained by weighting +1 of the second directed connecting edge a → B. In another embodiment, obtaining the node connection sequence: and Q → B1 → …, wherein Q in the node connection sequence is assumed to be a non-target track node, a node aggregation condition is satisfied between the target track node B1 and the target graph node B, a graph node corresponding to Q is created in the graph node set, the target track node B1 and the target graph node B are subjected to aggregation processing to obtain an updated graph node B, and a second directed connecting edge Q → B is created to obtain an updated second directed connecting edge.
In one embodiment, an upper limit is set for the weight of the second directed connecting edge in the road segment graph model, and when the highest value of the weight in the second directed connecting edge reaches the upper limit, the weights of all the second directed connecting edges in the road segment graph model are uniformly adjusted, so that the weight of the second directed connecting edge in the road segment graph model is always kept below the upper limit. In one embodiment, uniformly adjusting the weights of all the second directed connecting edges includes: and reading the minimum weight in the weights corresponding to the second directed connecting edges, and subtracting the minimum weight from the weights corresponding to all the second directed connecting edges to obtain the adjusted weights corresponding to the second directed connecting edges.
In this embodiment, based on the node characteristic distance between each track node in the node connection sequence and the graph node in the road segment graph model, the node connection sequence is updated to the graph node set and the second directed connecting edge in the road segment graph model, so as to obtain a new road segment graph model. That is to say, as new traveling track data is continuously generated in the target road segment, the road segment graph model also continuously learns the track nodes in the new traveling track data and the first directed connecting edges between the track nodes, so that the road segment graph model can learn the real-time road topological structure information of the target road segment, that is, the traveling track data newly generated in the on-line learning target road segment, and the road segment graph model can still keep consistent with the actual road segment road topological structure information in the situations of setting temporary roadblocks, traffic control and the like.
In one embodiment, as shown in fig. 6, the method further includes steps S610 to S630. Wherein:
step S610, obtaining historical travel track data collected on the target road segment.
The historical travel track data represents historical track data collected in the target road section, namely track data generated by the object traveling in the target road section. In one embodiment, the historical travel track data may be obtained in any number of ways.
And S620, obtaining a corresponding historical node connection sequence based on the historical travel track data.
In one embodiment, the historical travel track data is subjected to feature extraction to obtain a corresponding historical node connection sequence. The same method can be adopted for extracting the characteristics of the historical travel track data and the travel track data, and the details are not repeated herein.
Step S630, updating an initial graph node set in the initial road segment graph model of the target road segment and second directed connecting edges between the initial graph nodes in the initial graph node set based on the historical node connection sequence to obtain the road segment graph model of the target road segment.
The initial road section graph model represents a graph model created in the initial stage of building the road section graph model; in one embodiment, the initial road segment graph model includes only an empty set of graph nodes. In another embodiment, the initial road segment graph model includes only the set of graph nodes determined from the initial trajectory nodes determined from the road identification data for the target road segment. In other embodiments, the initial road segment graph model may also include a second directed connecting edge between the initial graph node and the graph node, which is manually input.
In one embodiment, as shown in fig. 7, the initial road segment graph model of the target road segment may be obtained by steps including step S710 to step S730. Wherein:
in step S710, link identification data of the target link is acquired.
In one embodiment, the road segment identification data for the target road segment includes data for lane lines, lane guide arrows, etc. in the target road segment; according to the road section identification data of the target road section, the respective conditions of the lanes in the target road section and the driving direction data of each lane can be determined. In one embodiment, the road segment identification data of the target road segment may be obtained from a preset database, may also be determined according to a priori knowledge of the target road segment, or may also be obtained by identifying an image captured by the target road segment by a relevant model, and the like.
Step S720, the road section track nodes corresponding to the road section identification data are used as initial graph nodes to obtain an initial graph node set, and second directed connecting edges between the initial graph nodes in the initial graph node set are obtained based on the first directed connecting edges between the road section track nodes.
In one embodiment, after obtaining the road segment identification data, a corresponding road segment trajectory node may be extracted based on the road segment identification data; the road section identification data includes a lane line and a lane guide arrow in the target road section, and in this embodiment, an initial track node in the target road section is extracted based on the lane line and the lane guide arrow. The initial track node extraction based on the road section identification data can be realized in any mode; for example, points can be randomly selected between every two lane lines as initial track nodes according to the positions of the lane lines; or selecting a point between every two lane lines at intervals of a preset distance as an initial track node; or may also be the acquisition of an initial trajectory node manually entered from the road segment identification data, and so on. In one embodiment, the initial track node containing the corresponding node characteristics includes location information of the track node.
In one embodiment, the connection relationship between the initial track nodes may be determined according to the guiding arrow in the road segment identification data; in practical situations, part of the lane lines will be marked with a guiding arrow for indicating the correct driving direction of the lane, and if the guiding arrow is extracted from the road section identification data, the connection relationship between the extracted initial track nodes in the lane can be determined according to the guiding arrow. In one embodiment, if the guiding arrow indicates that the vehicle in the lane should drive eastward, the connection relationship between two adjacent initial trajectory nodes extracted from the lane is determined as follows: from nodes in which the west is near to nodes in which the east is near.
In another embodiment, the directional connecting edges between the initial track nodes may be manually input, and after the initial track nodes are extracted based on the road segment identification data such as the lane lines, if the normal connecting direction between the initial track nodes cannot be identified, direction confirmation information may be generated and fed back to the manual operation for confirmation. In other embodiments, the directional connecting edges between the initial trajectory nodes may be obtained in other manners.
And obtaining a second directed connecting edge between corresponding graph nodes based on the directed connecting edges between the initial track nodes.
Step S730, an initial road segment graph model of the target road segment is obtained based on the initial graph node set and the second directed connecting edges between the initial graph nodes in the initial graph node set.
And obtaining the initial road section graph model of the corresponding target road section after obtaining the second directed connecting edge between the initial graph node set and each initial graph node.
In the embodiment, when constructing the initial road segment map model, the initial track nodes extracted by acquiring the road segment identification data in the target road segment are utilized, and the directed connecting edges among the initial track nodes are used as initial data to construct an initial graph node set in the initial road segment graph model and second directed connecting edges among all initial graph nodes in the initial graph node set, when the track is detected abnormally by the road segment graph model in the follow-up process, more standard initial data is provided, for less irregular sample points (possibly error points caused by target tracking errors, data points in a road section which is not concerned by abnormal track identification and the like) in the traveling track data to be identified, abnormal track detection can be realized, the accuracy of the road section map model in abnormal identification cannot be greatly influenced, and therefore the accuracy of the road section map model in abnormal track identification is improved.
In another embodiment, the initial road segment graph model of the target road segment may also construct an initially empty graph node set, the graph node set is empty, and there is no second directed connecting edge. In this embodiment, in the initialization stage, the graph node set and the second directed connecting edge in the road segment graph model are initially empty, and the initial road segment graph model is updated according to the historical node connection sequence corresponding to the historical travel track data in the target road segment, so as to obtain the road segment graph model for anomaly identification, including the graph nodes and the directed connecting edges between the graph nodes.
Further, in an embodiment, after updating the initial graph node set in the initial road segment graph model of the target road segment and the second directed connecting edge between each initial graph node in the initial graph node set based on the historical node connection sequence to obtain the road segment graph model of the target road segment, the method further includes: carrying out outlier sample-sampling filtering on the graph nodes in the road segment graph model and second directed connecting edges between the graph nodes to obtain second directed connecting edges between the graph nodes and the graph nodes in the filtered road segment graph model; in this embodiment, based on the graph nodes in the filtered road segment graph model and the second directed connecting edges between the graph nodes, the abnormal recognition is performed on the node connection sequence corresponding to the travel track data, and whether the travel track data is abnormal is determined.
In this embodiment, after the graph node set in the initial graph model and the second directed connecting edge between the initial graph nodes in the initial graph node set are updated according to the historical node connection sequence, the updated graph nodes and the second directed connecting edge are subjected to outlier sample filtering.
In one embodiment, an outlier sample can comprise: 1. noise due to target detection tracking errors, such as false detections and false tracking results. 2. The surveillance camera covers some areas outside the main lane (not the area of interest to the surveillance camera). Further, in one embodiment, points outside the preset range may be determined as outlier sample points by defining a preset range within the shooting range of the monitoring camera. If an outlier sample is detected in a graph node of the graph model, filtering the outlier sample; further, if an outlier sample is detected, the second directed connecting edge associated with the outlier sample is also deleted.
In one embodiment, outlier sample filtering of the road map model may be performed prior to extracting the sequence of node connections from the trajectory data. The time duration of the track can be observed by setting a threshold of the track time duration, when the track tracking time duration reaches a certain threshold, for example, 2 to 3 seconds, the track can be considered as a reasonable track, and if the track tracking time duration reaches the threshold, the track is considered as an outlier sample. Alternatively, by defining an area that is not focused by the monitoring probe, for example, the probe only focuses on the main road, and the vehicle on the auxiliary road does not need to focus on the probe, the segment in the trajectory data that is in the auxiliary road species is identified as an outlier sample, and the trajectory and sample points outside these places can be removed according to the focused area during filtering. It will be appreciated that in other embodiments, outlier sample filtering of existing trace data may be performed in other ways.
In another embodiment, outlier sample filtering is performed on the existing trace data, or may be performed after extracting the node-connecting sequence of the trace data. After the node connection sequence corresponding to the existing track data is extracted, the method can be implemented by a method similar to that when the track data is directly filtered, for example, a node connection sequence with a duration length smaller than a certain threshold is determined as an outlier sample, or a node connection sequence segment of a road segment in an area not concerned, such as a secondary road, is identified as an outlier sample, and the like.
In this embodiment, after the second directed connecting edge between the initial graph node and the initial graph node in the initial road segment graph model is updated, whether the obtained road segment graph model contains the outlier sample is detected, and the outlier sample is filtered out, so as to reduce the influence of the outlier sample on the graph model, ensure that the road segment graph model learns the accurate road structure of the target road segment, and improve the accuracy of the road segment graph model.
In a specific embodiment, generating a road segment graph model based on historical travel track data includes the following steps 1-8: step 1, obtaining a corresponding historical node connection sequence according to historical travel track data. And constructing an initial road section graph model, wherein the initial road section graph model comprises an initial graph node set and second directed connecting edges among all initial graph nodes in the initial graph node set.
The specific implementation manner of constructing the initial road segment graph model has been described in detail in the above embodiments, and is not described herein again.
And 2, selecting one unselected track node from the historical node connection sequence as the current track node.
In one embodiment, after a track node is selected from a node connection sequence as a current track node, a selected identifier is set for the current track node; and subsequently, whether the track node is selected or not can be determined according to whether the track node in the node connection sequence is provided with the selected identifier or not. In other embodiments, whether a track node is selected as the current track node may be distinguished in other ways.
And 3, updating the current track node to the current graph node set based on the node characteristic distance between the current track node and each graph node in the current graph node set, and determining a target node corresponding to the current track node in the graph node set. And when updating for the first time, the current graph node set is the initial graph node set.
The specific implementation of updating the current graph node set based on the node characteristic distance and the current trajectory node is described in detail in the embodiments of step S431 to step S433, and is not described herein again.
And 4, reading a current node connection sequence to which the current track node belongs.
In one embodiment, if the historical travel track data includes a plurality of nodes, and similarly, includes a plurality of historical node connection sequences, the node connection sequence to which the current track node belongs needs to be determined and recorded as the current node connection sequence. It can be understood that, if only one piece of trajectory data is included when the road segment graph model is generated, the node connection sequence corresponding to the piece of trajectory data is directly determined as the current node connection sequence.
And 5, searching adjacent track nodes adjacent to the current track node in the current node connection sequence.
In one embodiment, searching for an adjacent track node adjacent to the current track node in the current node connection sequence includes searching for a previous node or a next node adjacent to the current track node in the current node connection sequence. Setting to search only the previous node or the next node adjacent to the current track node, so that the same first directed connecting edge can be prevented from being searched when different track nodes are used as the current track node; for example, when the adjacent track node A points to the track node B, the track node B points to the track node C, and when the current track node is the node A, the next adjacent node is searched for as the node B, and at this time, the first directed connecting edge of the node A pointing to the node B is searched for; when the current track node is the node B, if the previous node is searched, the node a will be searched, and at this time, the first directed connecting edge of the node a pointing to the node B will be repeatedly searched, so that the adjacent next node is also searched when the node B searches, and it can be ensured that the first directed connecting edge of the node a pointing to the node B will not be repeatedly searched.
The previous node or the next node adjacent to the current track node can be determined according to the time corresponding to the track node, the historical node connection sequence corresponds to the historical advancing track data, the generation time points corresponding to different position points in the historical advancing track data are different and have a sequence, and the sequence of each node in the node connection sequence can be determined according to the time corresponding to the node position. Or, the sequence between the nodes may be determined according to the connection direction between the current track node and the adjacent track node, so as to determine the previous node or the next node adjacent to the current track node.
In one embodiment, the searching for the adjacent track node of the current track node in the current node connection sequence may be implemented by using a depth-first search method. Depth First Search (DFS) belongs to one of graph algorithms; the process is briefly that each possible branch path is too deep to be deep, and each node can only be visited once. In one embodiment, the depth-first search may set a depth of a desired search, for example, set the depth to 1, and then search only one neighboring track node for the current track node is stopped; in other embodiments, the depth of the depth-first search may be set to other values according to actual situations.
And 6, determining the corresponding adjacent graph nodes of the adjacent track nodes in the current graph node set.
After reading the adjacent track nodes of the current track node, determining the corresponding adjacent graph nodes of the adjacent track nodes in the current graph node set. In one embodiment, the neighboring track node of the current track node may have been updated into the current graph node set before this step, at which point the corresponding neighboring graph node of the neighboring track node in the current graph node set may be directly determined. In another embodiment, the neighboring track node of the current track node may not be updated to the current graph node set before this step, and at this time, the corresponding neighboring graph node of the neighboring track node in the current graph node set needs to be determined according to the node characteristic distance between the neighboring track node and each graph node in the current graph node set; and determining the adjacent graph nodes corresponding to the adjacent track nodes in the current graph node set in the same way as determining the graph nodes corresponding to the current track nodes in the current graph node set.
And 7, updating the weight of a second directed connecting edge between each graph node in the current graph node set based on the first directed connecting edge between the current track node and the adjacent track node, and returning to the step 2.
The steps already determine a current graph node corresponding to the current track node in the current graph node set and an adjacent graph node of an adjacent track node of the current track node in the current graph node set, and correspondingly update a second directed connecting edge between the current graph node and the adjacent graph node in the current graph node set based on a first directed connecting edge between the current track node and the adjacent track node.
Further, if a second directed connecting edge between the current graph node and the adjacent graph node already exists, the weight of the second directed connecting edge is added to be + 1; and if the second directed connecting edge between the current graph node and the adjacent graph node does not exist, creating a second directed connecting edge corresponding to the first directed connecting edge between the current track node and the adjacent track node in the current road section graph model, and setting the weight as 1.
And 8, when the nodes in all the historical node connection sequences are selected, obtaining a road section graph model according to the current graph node set and a second directed connection edge between all graph nodes in the graph node set.
In the above embodiment, the complete step of generating the road segment graph model according to the historical node connection sequence corresponding to the historical travel track data is described, and by the above method, the travel track data of the object is converted into the form of the road segment graph model, and is represented by the second directed connection edge between the graph node and the graph node, so that the model difficulty can be simplified. This process may also be referred to as an initialization process of the road segment map model. Further, when the road section graph model is generated, based on the traveling track data generated by the object in the target road section, the road section graph model can learn the road topology structure in the target road section through the process, and then the road section graph model is used for carrying out abnormal recognition on other traveling track data in the target road section, so that the analysis and judgment of the track are established on the road structure of the current visual field, and the situation that the data and the training model need to be collected again due to the fact that the abnormal track recognition model is caused by the change of the road structure can be avoided.
Further, in a specific embodiment, the generating a road segment graph model corresponding to the target road segment based on the historical node connection sequence corresponding to the historical travel track data includes: and (4) constructing an initial graph node set and a second directed connecting edge between each initial graph node in the initial graph node set. Secondly, selecting a current node connection sequence, and selecting a first track node which is not selected in the current track node connection sequence as a current track node based on the sequence corresponding to each track node in the node connection sequence. And thirdly, updating the current track node to the current graph node set based on the node characteristic distance, and determining the current graph node corresponding to the current track node in the graph node set. (reading the current node connection sequence identifier.) and (c) searching the previous track node adjacent to the current track node in the current node connection sequence. In this step, because the first track node is selected as the current track node (e.g., node B) in sequence, the previous adjacent track node (e.g., the previous adjacent track node a of node B) to the current track node is searched, and the previous adjacent track node is updated to the node set before the previous adjacent track node to the current track node, so that the adjacent graph node corresponding to the previous adjacent track node to the current track node in the current graph node set can be directly determined (i.e., the adjacent graph node corresponding to the node a in the current graph node set is determined when the node a is used as the current track node). And seventhly, updating the node connection relation of the node A pointing to the node B into the node connection relation set. And returning to the sequence corresponding to each track node in the node connection sequence, and selecting the first track node which is not selected in the current node connection sequence as the current track node. Ninthly, when all the track nodes in the node connection sequence are selected, obtaining a road section graph model according to the current graph node set and a second directed connection edge between all the graph nodes in the current graph node set.
In this embodiment, when the road segment graph model is generated, the sequence of updating the selected track node to the graph node set is determined to be that the first node in the historical node connection sequence is selected as the current track node, and when the adjacent track node of the current track node is searched, only the previous adjacent track node of the current track node is searched. Further, a first node may also be selected as the current trace node in the current node connection sequence, and when searching for an adjacent trace node, a next node of the current trace node (the first node) and a next node (a next node of the next node) are searched, … is performed until a last node in the current node connection sequence is searched, that is, all nodes in the target node connection sequence are searched at one time, that is, a first directed connection edge connection relationship of each trace node from the first node to the last node in the current node connection sequence is updated to a second directed connection edge between each graph node in the current graph node set at one time.
In another embodiment, the current track node may also be selected in the current node connection sequence according to the order from the end node, and when searching for the adjacent track node of the current track node, only the next adjacent track node of the current track node may be searched, or the previous adjacent track node to the current track node may be searched. Further, the last node of the current node connection sequence may also be selected as the current trace node, and when searching for an adjacent trace node, the previous node of the current trace node (last node) and the previous node are searched, …, until the first node in the current node connection sequence is searched, that is, all nodes in the current node connection sequence are searched at one time, that is, the first directed connection edge connection relationship of each trace node from the last node to the first node in the current node connection sequence is updated to the second directed connection edge between each graph node in the current graph node set at one time.
In the embodiment that an end node (a first node or a last node) is used as a current track node, and another end node (a last node or a first node) of a current node connection sequence is searched for at one time, the sequence of the steps from (c) to (b) in the above embodiment can be adjusted to (c): selecting a current node connection sequence, selecting a first node (or a last node) as a current track node, and updating the current track node to a current graph node set based on a node characteristic distance; searching the next adjacent track node (or the previous adjacent track node) of the current track node, adding the node set, … until the end node (or the head node) is searched, and updating the end node to the current graph node set; and starting from the first node (or the last node), updating second directed connecting edges among all graph node sets in the current graph node set sequentially based on the first directed connecting edges of the next adjacent track node (or the last adjacent track node). Or, when the road segment graph model is updated, all the track nodes in the current node connection sequence are updated to the current graph node set, and then the corresponding second directed connection edges are sequentially updated for the first directed connection edges between the current track nodes and the adjacent track nodes. In other embodiments, the above steps may be performed in other orders, as long as it is ensured that all the trace nodes in all the historical node connection sequences and the first directed connection edges between the trace nodes update the graph node set in the segment graph model and the second directed connection edges between the graph nodes in the graph node set.
In the embodiment of updating the road segment graph model based on the node connection sequence corresponding to the travel track data to be identified, the method also extracts the characteristics of the travel track data to be identified to obtain the node connection sequence, wherein the node connection sequence comprises a first directed connection edge between a track node and the track node, and then updates a graph node set in the road segment graph model and a second directed connection edge between graph nodes in the graph node set by using the first directed connection edge between the track node and the track node. Updating the travel track data to be identified to the road section graph model comprises two modes of online updating and offline updating.
And updating the traveling track data to be identified to the road section graph model representation on line: when an object moves in a target road section, gradually generating travel track data to be identified, and at the moment, extracting features from the generated travel track data to be identified to obtain nodes and directed connecting edges of the connecting nodes; and performing a step of updating the road segment map model into the road segment map model, wherein the step comprises the following steps: (1) and taking the extracted latest track node as a current track node, and updating the current track node to a graph node set in the current graph model. (2) And searching the last adjacent track node of the latest track node. (3) Updating a second directed connecting edge between the graph nodes in the current graph model based on the first directed connecting edge of the last adjacent track node and the latest track node; and (4) returning to the step (1) until the object exits the target road segment. The process is similar to the above steps (c) - (c).
Further, after the node connection sequence corresponding to the complete travel track data to be identified is updated to the road section graph model on line, the updated road section graph model is used for carrying out abnormity identification on the travel track data to be identified.
And the off-line updating of the travel track data to be recognized to the road section graph model representation is as follows: when the object runs through the target road section and already runs out of the target road section, generating complete traveling track data to be recognized, extracting features from the complete traveling track data at the moment, obtaining a corresponding node connection sequence which comprises track nodes and a first directed connection edge between the track nodes, and executing the step of updating a road section graph model based on the track nodes and the first directed connection edge between the track nodes; this part of the steps may include: selecting a first node (or a last node) in a node connection sequence corresponding to the travel track data to be identified as a current track node, and updating the current track node to a node set based on a node characteristic distance (between the track node and a graph node); searching the next adjacent track node (or the previous adjacent track node) of the current track node, updating to the current graph node set based on the node characteristic distance, … until the end node (or the first node) is searched, and updating to the current graph node set; and starting from the first node (or the last node), and updating second directed connecting edges between the graph nodes in the current graph node set on the basis of the first directed connecting edges between the first directed connecting edges and the next adjacent track node (or the last adjacent track node) in sequence. Or, when updating the graph node set and the second directed connecting edges between the graph nodes in the graph node set, all the trace nodes in the current node connection sequence are updated to the graph node set, and then the second directed connecting edges between the graph nodes in the graph node set are sequentially updated based on the first directed connecting edges between the current trace nodes and the adjacent trace nodes. The process is similar to the above steps (c) - (c).
In the above embodiments, different implementation manners are provided for generating the road segment graph model and updating the node connection sequence corresponding to the travel track data to be identified into the road segment graph model, and the above embodiments all implement a process of updating a second directed connection edge between a graph node set and a graph node in the road segment graph model based on the track nodes in the node connection sequence and the first directed connection edges between the track nodes, and only part of the sequences are different.
In one embodiment, as shown in fig. 8, based on graph nodes in the link graph model of the target link and second directed connecting edges between the graph nodes, performing anomaly identification on a node connection sequence corresponding to the travel track data, and determining whether the travel track data is abnormal, including steps S231 to S233. Wherein:
step S231, matching the track nodes in the node connection sequence and the first directed connection edges between the track nodes with the graph nodes in the road segment graph model of the target road segment and the second directed connection edges between the graph nodes, and determining, for each first directed connection edge, a weight corresponding to the second directed connection edge matched with the first directed connection edge.
In one embodiment, matching the track nodes in the node connection sequence and the first directed connecting edges between the track nodes with the graph nodes in the road segment graph model of the target road segment and the second directed connecting edges between the graph nodes respectively comprises: and matching the track node with the graph node, and matching the first directed connecting edge with the second directed connecting edge.
Further, matching the track nodes with the graph nodes comprises: and calculating the node characteristic distance between the track node and the graph node based on the node characteristics of the track node and the node characteristics of the graph node, and matching the track node and the graph node according to the node characteristic distance between the track node and the graph node. And more closely, respectively calculating the node characteristic distance between the track node and each graph node aiming at any track node, taking the minimum value of the node characteristic distance to judge, and if the minimum value of the node characteristic distance is smaller than a preset distance threshold, taking the graph node corresponding to the minimum value of the node characteristic distance to determine as the graph node matched with the track node. And if the minimum value of the node characteristic distance is greater than or equal to the preset distance threshold value, judging that no matched graph node exists between the track node and the road section graph model, and creating a graph node corresponding to the track node in the road section graph model as the graph node matched with the track node.
Matching the first directed connection edge with the second directed connection edge includes: and matching according to the two track nodes corresponding to the first directed connection edge and the two graph nodes corresponding to the second directed connection edge. And if the two track nodes corresponding to the first directed connection edge are respectively matched with the two graph nodes corresponding to the second directed connection edge, matching the first directed connection edge with the second directed connection edge. In one embodiment, if the two track nodes corresponding to the first directed connecting edge are a1 → B1, the two graph nodes corresponding to the pronunciation of the second directed connecting edge are a → B, the track node a1 matches the graph node a, and the track node B1 matches the graph node B, then the first directed connecting edge matches the second directed connecting edge.
The weight of the second directed connecting edge represents the weight of the second directed connecting edge in the road section graph model; in one embodiment, the weight of the second directed connecting edge is determined according to directed connecting edges between the trace nodes in the historical node connection sequence, and after the graph nodes in the road segment graph model are updated based on the node characteristic distance between the trace nodes and the graph nodes, the weight of the directed connecting edge between two adjacent graph nodes corresponding to the two adjacent trace nodes is + 1; in a specific embodiment, the step of determining the weight of the second directed connecting edge may refer to the description in the previous embodiment, and is not described herein again.
Step S232, using the first directed connection edge whose weight does not exceed the preset weight threshold as an abnormal connection edge, and determining an abnormal proportion of all the first directed connection edges of the abnormal connection edge in the node connection sequence.
In one embodiment, the preset weight threshold may be set to a fixed value according to actual conditions, such as to be set to a numerical value of 3, 5, 10, etc.; or the preset weight threshold may also be set as a dynamic threshold according to an actual situation, for example, weights corresponding to all second directed connecting edges in the statistical road segment graph model are counted, all weights are sorted according to sizes, and a median of all weights or a weight corresponding to a position of a third quartile from large to small is set as the preset weight threshold, so that along with time change, a weight value corresponding to the second directed connecting edge in the road segment graph model gradually increases, and the road segment graph model still meets the actual situation when the travel track to be identified is subjected to the abnormal judgment, and the accuracy of the road segment graph model for identifying the abnormal track of the travel track data can be still ensured without adjusting the weight threshold.
And step S233, when the abnormal proportion exceeds the preset proportion, determining that the travel track data is abnormal.
The preset ratio can be set according to actual conditions, such as 40%, 50% and the like. In this embodiment, if the abnormal proportion of the abnormal connecting edges in all the first directed connecting edges in the node connecting sequence exceeds the preset proportion, it is determined that the node connecting sequence is an abnormal node connecting sequence, that is, the data of the travel track to be identified is an abnormal track. In a specific embodiment, assuming that the preset percentage is set to be 40%, the number of the first directed connecting edges included in the node connection sequence corresponding to the travel track data is 10, and the number of the abnormal connecting edges is 5, the abnormal percentage is 50%, and the preset percentage is 40% exceeded, it is determined that the travel track data is abnormal.
In the above embodiment, the process of identifying the abnormal trajectory of the travel trajectory data corresponding to the node connection sequence is performed before the road segment graph model is updated based on the node connection sequence; in another embodiment, the process of performing the abnormal trajectory identification on the travel trajectory data corresponding to the node connection sequence may also be performed after updating the road segment graph model based on the node connection sequence, in this embodiment, since when updating the road segment graph model based on the node connection sequence, the graph nodes corresponding to the track nodes in the node connection sequence in the road segment graph model have been already determined, that is, the matching process in the corresponding step S231 has been completed in updating the road segment graph model based on the node connection sequence, in this embodiment, the abnormal identification only needs to be performed according to the graph nodes and the second directed connecting edges that are matched at the time of updating.
The application also provides an application scene, and the application scene applies the abnormal track identification method. In the present embodiment, the description is made with the monitoring probe capturing the trajectory of an object appearing in a target road segment within the coverage, such as a vehicle.
In a road video monitoring system, identifying and analyzing abnormal advancing tracks on roads is one of necessary functional links of the whole system. The traditional track analysis and classification method focuses on the geometric shape information of the object track, and ignores the relation between the track and the road topological structure. The road topological structure in the view field of the monitoring equipment changes along with the change of the installation place and the view angle of the monitoring equipment, which brings a huge training data updating problem to the algorithm for model learning based on off-line data, namely, object motion track data corresponding to a new road topological structure needs to be collected again, and especially the track data of some abnormal events are more difficult to obtain in a short time.
The embodiment provides an abnormal track identification method, which learns the track in the view domain of a monitoring device into a directed weighted graph in an online form to represent through the road track modeling of online unsupervised directed weighted graph learning, and not only retains the road topological structure information, but also represents the extended path of the track and the weight of the path passed by the extended path through the connection relation between nodes and directed weighted edges. By analyzing the graph model, the abnormal track can be analyzed and distinguished by the weight distribution of the node and the edge sequence on the premise of combining the road topological structure information. The method can be applied to detecting the illegal running of the vehicle under the high-level monitoring scene of road traffic, such as snake running for frequently changing lanes, reverse running of a main road and other abnormal running tracks different from normal running tracks.
Specifically, the application of the abnormal trajectory identification method in the application scenario is as follows:
(1) road track modeling based on online unsupervised directed weighted graph learning
The graph model is an algorithm structure for expressing the connection relation of nodes. In the trajectory analysis and classification, a motion trajectory of an object can be regarded as a sequence of a plurality of nodes connected together, each node represents a local segment characteristic (such as position, length, curvature, and the like) of the trajectory, and an edge represents a connection order between the nodes.
After a plurality of tracks are represented in the form of graph nodes and directed connecting edges, the nodes are clustered, and the vehicle passing frequency statistics is carried out on the edges of the connecting nodes to obtain the weight of the edges. The clustering algorithm of the graph nodes is as follows:
Input, namely, graph nodes G { }, node connection relations E { }, node connection sequences X { X1, X2, … xn }, and n { [ 1 ] corresponding to the plurality of pieces of travel trajectory data
At this point, the algorithm 1 completes inputting a node connection sequence corresponding to the travel track data of one object, constructs or updates a graph model of the current road, and when a plurality of objects are in the view domain, processes the node connection sequence of each object one by one, and then completes graph model algorithm learning of all object tracks.
In the algorithm 1, G is initially an empty set, and in order to reduce the learning cost of the algorithm in initially constructing a graph model, the prior knowledge of the lane lines in the target road segment may be used to extract the initial track node feature information (node positions) of the road as the initial track node set of G. As shown in fig. 9, the black solid points between the lane lines are the initial track nodes, and the initial track nodes have only position information at the initial time and no curvature and track segment length information.
Based on the above algorithm 1, after a certain intersection is autonomously learned for a certain period of time, a graph model of the following roads is obtained, as shown in fig. 10(1), fig. 10(2), fig. 10(3), and fig. 10(4), where a black dot is a graph node learned by the graph model, a triangle represents a direction of a directed connecting edge, a pointed angle represents a direction of a directed connecting edge, a black line segment represents a directed connecting edge on the graph, numerals (i), (ii), and … represent graph node identifiers, and numerals 1, 2, and … represent weights of the directed connecting edges. In other embodiments, the graph node identifier and the weight of the directed connecting side may be represented by numbers of different colors (for example, black numbers represent the node identifier and blue numbers represent the weight of the side), or shapes of different colors may represent the graph node and the direction of the directed connecting side (for example, black points represent the graph node and green points represent the direction of the directed connecting side, where the direction of the directed connecting side may be that a graph node farther from the green point position points to a graph node closer to the green point position); in other embodiments, information such as graph nodes, graph node identifiers, weights of directed connecting edges, and directions of directed connecting edges may be represented in the graph in other ways.
(2) Abnormal track behavior identification method based on directed weighted graph
The abnormal track is a general name different from the normal track, and has the characteristics of large intra-class dispersion degree and difficult collection. By clustering a trajectory into the graph model, a trajectory can be represented as a series of graph node connection sequences and a weight sequence of directed connection edges.
Assuming that a piece of travel track data to be identified exists in the scene of fig. 10(4) and is already clustered in the graph model of the scene, after the piece of travel track data to be identified is criticized with the graph node set in the road segment graph model and the second directed connecting edge between the graph nodes in the graph node set, the travel track data to be identified may be represented as a graph node connecting sequence G ═ 4, 8, 11, 9, 10, 14, and a weight sequence of the directed connecting edges is represented as E ═ 2, 12, 1, 9, 8. Assuming that the preset weight threshold is 3, the preset percentage is 40%: as can be seen from the above-mentioned node and edge sequences, 2 of the 5 second directed connecting edges matched in the travel track data to be recognized are abnormal directed connecting edges (edge weight < preset weight threshold 3), reach a preset percentage of 40%, and determine that the travel track data to be recognized is an abnormal behavior track.
In a specific embodiment, in the case of a temporary road condition, such as temporary setting of roadblocks, traffic control, and the like, the road segment graph model still maintains the judgment logic before the temporary road condition occurs, and if the temporary road condition occurs, the judgment logic of the road segment graph model will be affected, for example, a lane change in the lane 1 before the temporary road condition occurs will be determined as an abnormal track, but after the temporary road condition occurs, the lane change in the lane 1 should be allowed according to the actual situation and should be determined as a normal track. In actual situations, after the temporary road condition occurs, the number of the situations that the object is likely to change lanes in the lane 1 is increased, the road section graph model learns through the traveling track data containing lane change in the lane 1, when the corresponding weight of the directional connecting side corresponding to lane change in the lane 1 exceeds a preset weight threshold value and the road section graph model identifies the abnormal track of lane change in the lane 1, the identification of the directional connecting side corresponding to lane change in the lane 1 is switched from the original abnormal connecting side (before the temporary road condition occurs) to the normal connecting side (after the temporary road condition occurs), some tracks which are considered as abnormal in the past may also occur frequently and become normal tracks, and after the temporary road condition ends, the traveling track data of lane change in the lane 1 is not increased greatly, then the weight corresponding to the corresponding directional connecting edge is gradually reduced to be smaller than the preset weight threshold, that is, "lane change in lane 1" is classified as an abnormal track again. Since the graph model is always learned online, whether a track is abnormal or not in the actual traffic road condition is usually determined according to the frequency of the track. The advantage of learning all the time is that the normal track is continuously accumulated in the edge weight of the graph model, and the graph model edge weight of the abnormal track is lower, so as to be distinguished. Therefore, when the temporary road condition occurs, the model does not need to be retrained to carry out abnormal track recognition, but the temporary advancing track in the road section is increased due to the occurrence of the temporary road condition along with the passing of time, and the road section graph model can also learn the actual road condition in real time.
In one embodiment, in the embodiment where the temporary road condition occurs, a part of the tracks may be misjudged, a preset time period and an abnormal track threshold may be set at this time, when the number of abnormal tracks detected at the same position in the preset time period exceeds the abnormal track threshold, prompt information is generated, the prompt information is sent to related personnel (such as a traffic police and the like), and feedback information of the related personnel for the prompt information is received, so as to determine whether the part of the abnormal tracks needs to adjust the judgment result. The preset time period and the abnormal track threshold value can be set to be any setting according to actual conditions.
The abnormal track identification method is based on monocular camera data, and carries out time-series representation on the obstacles by sensing the positions and the types of the obstacles on the road in the image and tracking the track. A road driving topological structure diagram in a current visual field scene is constructed through the learning of an online unsupervised directed weighted graph, graph nodes and directed connecting edges represent the extending direction of a track, and the weights of the directed connecting edges represent the frequency distribution relation of a vehicle driving track. By analyzing the node sequence and the edge weight relation of the vehicle track in the directed weighted graph, the abnormal behavior track of the vehicle on the road can be mined. The track information of the vehicle is modeled by utilizing the directed weighted graph model, and the topological structure information of the road is learned into the model due to the online learning model. On one hand, the defect that a road structure is not referred when the track is judged to be abnormal by a traditional method is overcome, on the other hand, the track representation method of a graph structure is adopted, the judgment of the abnormal track is converted into the matching problem of a graph node connection sequence, and the complexity of the model is simplified.
It should be understood that, although the steps in the flowcharts involved in the above embodiments are shown in order as indicated by the arrows, the steps are not necessarily performed in order as indicated by the arrows. The steps are not performed in the exact order shown and described, and may be performed in other orders, unless explicitly stated otherwise. Moreover, at least a part of the steps in each flowchart involved in the above embodiments may include multiple steps or multiple stages, which are not necessarily performed at the same time, but may be performed at different times, and the order of performing the steps or stages is not necessarily sequential, but may be performed alternately or alternately with other steps or at least a part of the steps or stages in other steps.
In one embodiment, as shown in fig. 11, an abnormal trajectory recognition apparatus is provided, which may be a part of a computer device using a software module or a hardware module, or a combination of the two modules, and specifically includes: a trajectory data acquisition module 1110, a feature extraction module 1120, and an anomaly identification module 1130, wherein:
a track data obtaining module 1110, configured to obtain travel track data to be identified, where the travel track data is acquired at a target road segment;
the feature extraction module 1120 is configured to perform feature extraction on the travel track data to obtain a node connection sequence corresponding to the travel track data; the node connection sequence comprises track nodes corresponding to the traveling track data and first directed connection edges among the track nodes;
the anomaly identification module 1130 is configured to perform anomaly identification on a node connection sequence corresponding to the travel track data based on graph nodes in the road segment graph model of the target road segment and second directed connection edges between the graph nodes, and determine whether the travel track data is abnormal; the road section graph model of the target road section is generated based on a historical node connection sequence corresponding to historical travel track data acquired from the target road section.
The abnormal track recognition device acquires traveling track data to be recognized, which is acquired at a target road section, and performs feature extraction on the traveling track data to obtain a corresponding node connection sequence, which comprises track nodes of a traveling track and a first directed connection edge between the track nodes; and performing abnormity identification on the node connection sequence based on a second directed connection edge between the graph node and the graph node in the road section graph model of the target road section, and determining whether the travel track data is abnormal. The graph model is generated based on historical node connection sequences corresponding to historical travel track data acquired from a target road section. The method converts the identification of the traveling track data collected on the target road section into the matching problem of the node connection sequence in the corresponding graph model, can realize the automatic abnormal identification of the traveling track data based on the model with simpler complexity, and reduces the complexity of the abnormal track identification.
In one embodiment, the above apparatus further comprises: and the first updating module is used for updating the road section graph model according to the node connection sequence to obtain a new road section graph model.
In one embodiment, the first updating module of the apparatus comprises: the data acquisition unit is used for acquiring a graph node set of a road segment graph model of the target road segment and a second directed connecting edge between graph nodes in the graph node set; the characteristic distance calculation unit is used for determining a node characteristic distance between each track node in the node connection sequence and each graph node in the graph node set of the road section graph model; the first update module is further configured to: and updating a graph node set of the road section graph model based on the node characteristic distance, and obtaining a new road section graph model by using a second directed connecting edge between graph nodes in the updated graph node set.
Further, in one embodiment, the first updating module of the apparatus includes: the node aggregation unit is used for carrying out node aggregation processing on the target track nodes and the target graph nodes which meet the node aggregation condition based on the node characteristic distance and updating the target graph nodes; judging that a node aggregation condition is met when the node characteristic distance between the track node and the graph node is smaller than a preset distance threshold; the graph node creating unit is used for creating graph nodes corresponding to the non-target track nodes in the road section graph model aiming at the non-target track nodes which do not meet the node aggregation condition with any graph node; the first update module is further configured to: and updating the graph node set of the road section graph model and a second directed connecting edge between each graph node in the updated graph node set based on at least one of the graph nodes corresponding to the updated target graph node and the non-target track node to obtain a new road section graph model.
In one embodiment, the characteristic distance calculating unit of the above apparatus is further configured to: for each track node in the node connection sequence, determining a node characteristic distance between the track node and a graph node in the road section graph model based on the node characteristics of the track node and the node characteristics of the graph node in the road section graph model; wherein the node characteristics of each node include at least one of a node position, a curvature of the node position in the corresponding travel trajectory data, and a trajectory length represented by the node position.
In one embodiment, the above apparatus further comprises: the historical track acquisition module is used for acquiring historical traveling track data acquired at a target road section; the historical node connection sequence determining module is used for obtaining a corresponding historical node connection sequence based on historical travel track data; a second update module to: and updating an initial graph node set in the initial road section graph model of the target road section and a second directed connecting edge between each initial graph node in the initial graph node set based on the historical node connection sequence to obtain the road section graph model of the target road section.
In one embodiment, the above apparatus further comprises: the road section identification acquisition module is used for acquiring road section identification data of a target road section; the initial data determining module is used for taking the road section track nodes corresponding to the road section identification data as initial graph nodes to obtain an initial graph node set, and obtaining second directed connecting edges between the initial graph nodes in the initial graph node set based on first directed connecting edges between the road section track nodes; and the initial road section graph model determining module is used for obtaining an initial road section graph model of the target road section based on the initial graph node set and the second directed connecting edges between the initial graph nodes in the initial graph node set.
In one embodiment, the above apparatus further comprises: and the filtering module is used for carrying out outlier sample-sampling filtering on the graph nodes in the road segment graph model and the second directed connecting edges between the graph nodes to obtain the filtered second directed connecting edges between the graph nodes and the graph nodes in the road segment graph model.
In one embodiment, the anomaly identification module 1130 includes: the matching unit is used for matching the track nodes in the node connection sequence and the first directed connecting edges between the track nodes with the graph nodes in the road section graph model of the target road section and the second directed connecting edges between the graph nodes respectively, and determining the weight corresponding to the second directed connecting edges matched with the first directed connecting edges aiming at each first directed connecting edge; the abnormal proportion determining unit is used for taking the first directed connecting edges with the weights not exceeding the preset weight threshold value as abnormal connecting edges and determining the abnormal proportion of all the first directed connecting edges of the abnormal connecting edges in the node connecting sequence; and the judging unit is used for determining that the traveling track data is abnormal when the abnormal ratio exceeds the preset ratio.
For a specific embodiment of the abnormal trajectory identification device, reference may be made to the above embodiments of the abnormal trajectory identification method, which are not described herein again. The modules in the abnormal trajectory recognition device can be wholly or partially realized by software, hardware and a combination thereof. The modules can be embedded in a hardware form or independent from a processor in the computer device, and can also be stored in a memory in the computer device in a software form, so that the processor can call and execute operations corresponding to the modules.
In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as shown in fig. 12. The computer device includes a processor 1210, a memory 1220 (not shown), and a network interface 1230, which are connected by a system bus. Wherein the processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium 1221, an internal memory 1222. The non-volatile storage medium 1221 stores an operating system 12211, a computer program 12212, and a database 12213. The internal memory 1222 provides an environment for the operation of an operating system and computer programs in the nonvolatile storage medium 1221. The database 12213 of the computer device is used to store the generated graph model. The network interface 1230 of the computer apparatus is used for communicating with an external terminal through a network connection. The computer program is executed by a processor to implement an abnormal trajectory recognition method.
Those skilled in the art will appreciate that the architecture shown in fig. 12 is merely a block diagram of some of the structures associated with the disclosed aspects and is not intended to limit the computing devices to which the disclosed aspects apply, as particular computing devices may include more or less components than those shown, or may combine certain components, or have a different arrangement of components.
In one embodiment, a computer device is further provided, which includes a memory and a processor, the memory stores a computer program, and the processor implements the steps of the above method embodiments when executing the computer program.
In an embodiment, a computer-readable storage medium is provided, in which a computer program is stored which, when being executed by a processor, carries out the steps of the above-mentioned method embodiments.
In one embodiment, a computer program product or computer program is provided that includes computer instructions stored in a computer-readable storage medium. The computer instructions are read by a processor of a computer device from a computer-readable storage medium, and the computer instructions are executed by the processor to cause the computer device to perform the steps in the above-mentioned method embodiments.
It will be understood by those skilled in the art that all or part of the processes of the methods of the embodiments described above can be implemented by hardware related to instructions of a computer program, which can be stored in a non-volatile computer-readable storage medium, and when executed, can include the processes of the embodiments of the methods described above. Any reference to memory, storage, database or other medium used in the embodiments provided herein can include at least one of non-volatile and volatile memory. Non-volatile Memory may include Read-Only Memory (ROM), magnetic tape, floppy disk, flash Memory, optical storage, or the like. Volatile Memory can include Random Access Memory (RAM) or external cache Memory. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM), among others.
The technical features of the above embodiments can be arbitrarily combined, and for the sake of brevity, all possible combinations of the technical features in the above embodiments are not described, but should be considered as the scope of the present specification as long as there is no contradiction between the combinations of the technical features.
The above examples only express several embodiments of the present application, and the description thereof is more specific and detailed, but not construed as limiting the scope of the invention. It should be noted that, for a person skilled in the art, several variations and modifications can be made without departing from the concept of the present application, which falls within the scope of protection of the present application. Therefore, the protection scope of the present patent shall be subject to the appended claims.
Claims (10)
1. An abnormal track identification method is characterized by comprising the following steps:
acquiring traveling track data to be identified, wherein the traveling track data is acquired in a target road section;
extracting the characteristics of the traveling track data to obtain a node connection sequence corresponding to the traveling track data; the node connection sequence comprises track nodes corresponding to the traveling track data and first directed connection edges among the track nodes;
performing anomaly identification on a node connection sequence corresponding to the travel track data based on graph nodes in the road section graph model of the target road section and second directed connection edges between the graph nodes, and determining whether the travel track data is abnormal; the road section graph model of the target road section is generated based on a historical node connection sequence corresponding to historical travel track data acquired from the target road section.
2. The abnormal trajectory recognition method according to claim 1, wherein after the feature extraction is performed on the travel trajectory data to obtain a node connection sequence corresponding to the travel trajectory data, the method further comprises:
and updating the road section graph model according to the node connection sequence to obtain a new road section graph model.
3. The abnormal trajectory recognition method according to claim 2, wherein the updating the road segment graph model according to the node connection sequence to obtain a new road segment graph model comprises:
acquiring a graph node set of a road segment graph model of the target road segment and a second directed connecting edge between graph nodes in the graph node set;
determining, for each track node in the node connection sequence, a node characteristic distance between the track node and a graph node in a graph node set of the road segment graph model;
and updating a graph node set of the road section graph model based on the node characteristic distance, and obtaining a new road section graph model by using a second directed connecting edge between graph nodes in the updated graph node set.
4. The abnormal track identification method according to claim 3, wherein the updating the graph node set of the road segment graph model based on the node feature distance and the second directed connecting edge between each graph node in the updated graph node set to obtain a new road segment graph model comprises:
performing node aggregation processing on the target track nodes and the target graph nodes which meet the node aggregation condition based on the node characteristic distance, and updating the target graph nodes; judging that the node aggregation condition is met when the node characteristic distance between the track node and the graph node is smaller than a preset distance threshold;
aiming at non-target track nodes which do not meet node aggregation conditions with any graph node, creating graph nodes corresponding to the non-target track nodes in the road section graph model;
and updating the graph node set of the road section graph model and a second directed connecting edge between graph nodes in the updated graph node set based on at least one of the updated target graph node and the graph node corresponding to the non-target track node to obtain a new road section graph model.
5. The abnormal trajectory identification method according to claim 3 or 4, wherein the determining, for each trajectory node in the node connection sequence, a node feature distance between the trajectory node and a graph node in a graph node set of the road segment graph model comprises:
for each track node in the node connection sequence, determining a node feature distance between the track node and a graph node in the road segment graph model based on the node feature of the track node and the node feature of the graph node in the road segment graph model; wherein the node characteristics of each node include at least one of a node position, a curvature of the node position in the corresponding travel trajectory data, and a trajectory length represented by the node position.
6. The abnormal trajectory recognition method of claim 1, further comprising:
acquiring historical traveling track data acquired at the target road section;
obtaining a corresponding historical node connection sequence based on the historical travel track data;
and updating an initial graph node set in the initial graph model of the target road section and a second directed connecting edge between each initial graph node in the initial graph node set based on the historical node connection sequence to obtain the road section graph model of the target road section.
7. The abnormal trajectory recognition method according to claim 6, wherein the initial road segment map model of the target road segment is obtained by:
acquiring road section identification data of the target road section;
taking the road section track nodes corresponding to the road section identification data as initial graph nodes to obtain an initial graph node set, and obtaining second directed connecting edges between the initial graph nodes in the initial graph node set on the basis of first directed connecting edges between the road section track nodes;
and obtaining an initial road section graph model of the target road section based on the initial graph node set and a second directed connecting edge between each initial graph node in the initial graph node set.
8. The abnormal trajectory identification method according to claim 6 or 7, wherein after the updating of the initial graph node set in the initial segment graph model of the target segment based on the historical node connection sequence and the second directed connecting edge between the initial graph nodes in the initial graph node set to obtain the segment graph model of the target segment, the method further comprises:
performing outlier sample filtering on the graph nodes in the road segment graph model and second directed connecting edges between the graph nodes to obtain second directed connecting edges between the graph nodes and the graph nodes in the filtered road segment graph model;
performing anomaly identification on a node connection sequence corresponding to the travel track data based on graph nodes in the road segment graph model of the target road segment and second directed connection edges between the graph nodes, and determining whether the travel track data is abnormal, including: and performing abnormity identification on a node connection sequence corresponding to the travel track data based on graph nodes in the filtered road segment graph model and second directed connection edges between the graph nodes, and determining whether the travel track data is abnormal.
9. The abnormal trajectory identification method according to claim 1, wherein performing abnormal identification on a node connection sequence corresponding to the travel trajectory data based on graph nodes in the road segment graph model of the target road segment and second directed connection edges between the graph nodes to determine whether the travel trajectory data is abnormal comprises:
matching the track nodes in the node connection sequence and first directed connection edges between the track nodes with the graph nodes in the road section graph model of the target road section and second directed connection edges between the graph nodes, and determining the weight corresponding to the second directed connection edges matched with the first directed connection edges for each first directed connection edge;
taking the first directed connection edges with the weight not exceeding a preset weight threshold value as abnormal connection edges, and determining the abnormal proportion of all the first directed connection edges of the abnormal connection edges in the node connection sequence;
and when the abnormal proportion exceeds a preset proportion, determining that the travel track data is abnormal.
10. An abnormal trajectory recognition apparatus, characterized in that the apparatus comprises:
the track data acquisition module is used for acquiring the traveling track data to be identified, wherein the traveling track data is acquired at a target road section;
the characteristic extraction module is used for extracting the characteristics of the traveling track data to obtain a node connection sequence corresponding to the traveling track data; the node connection sequence comprises track nodes corresponding to the traveling track data and first directed connection edges among the track nodes;
the abnormity identification module is used for carrying out abnormity identification on a node connection sequence corresponding to the traveling track data based on graph nodes in the road section graph model of the target road section and second directed connection edges between the graph nodes, and determining whether the traveling track data is abnormal or not; the road section graph model of the target road section is generated based on a historical node connection sequence corresponding to historical travel track data acquired from the target road section.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110482956.XA CN113762043A (en) | 2021-04-30 | 2021-04-30 | Abnormal track identification method and device |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110482956.XA CN113762043A (en) | 2021-04-30 | 2021-04-30 | Abnormal track identification method and device |
Publications (1)
Publication Number | Publication Date |
---|---|
CN113762043A true CN113762043A (en) | 2021-12-07 |
Family
ID=78786959
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202110482956.XA Pending CN113762043A (en) | 2021-04-30 | 2021-04-30 | Abnormal track identification method and device |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN113762043A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114117261A (en) * | 2022-01-29 | 2022-03-01 | 腾讯科技(深圳)有限公司 | Track detection method and device, electronic equipment and storage medium |
-
2021
- 2021-04-30 CN CN202110482956.XA patent/CN113762043A/en active Pending
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114117261A (en) * | 2022-01-29 | 2022-03-01 | 腾讯科技(深圳)有限公司 | Track detection method and device, electronic equipment and storage medium |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN114970321A (en) | Scene flow digital twinning method and system based on dynamic trajectory flow | |
CN107145862B (en) | Multi-feature matching multi-target tracking method based on Hough forest | |
CN113326719A (en) | Method, equipment and system for target tracking | |
CN103235933A (en) | Vehicle abnormal behavior detection method based on Hidden Markov Model | |
EP3690744B1 (en) | Method for integrating driving images acquired from vehicles performing cooperative driving and driving image integrating device using same | |
US9760783B2 (en) | Vehicle occupancy detection using passenger to driver feature distance | |
CN103246896A (en) | Robust real-time vehicle detection and tracking method | |
CN112163446A (en) | Obstacle detection method and device, electronic equipment and storage medium | |
CN114913386A (en) | Training method of multi-target tracking model and multi-target tracking method | |
CN113762044A (en) | Road recognition method, road recognition device, computer equipment and storage medium | |
CN111666860A (en) | Vehicle track tracking method integrating license plate information and vehicle characteristics | |
CN110738366B (en) | Outdoor blind zone behavior prediction method | |
Lu et al. | An efficient network for multi-scale and overlapped wildlife detection | |
CN112347983B (en) | Lane line detection processing method, lane line detection processing device, computer equipment and storage medium | |
CN113762043A (en) | Abnormal track identification method and device | |
CN111951543A (en) | Flow prediction method and device | |
CN113807457A (en) | Method, device and equipment for determining road network characterization information and storage medium | |
CN117115752A (en) | Expressway video monitoring method and system | |
CN112509321A (en) | Unmanned aerial vehicle-based driving control method and system for urban complex traffic situation and readable storage medium | |
CN111860383A (en) | Group abnormal behavior identification method, device, equipment and storage medium | |
CN115311867B (en) | Tunnel scene positioning method and device, computer equipment and storage medium | |
KR102682309B1 (en) | System and Method for Estimating Microscopic Traffic Parameters from UAV Video using Multiple Object Tracking of Deep Learning-based | |
Grigoropoulos et al. | Detection and classification of bicyclist group behavior for automated vehicle applications | |
CN113703015B (en) | Data processing method, device, equipment and medium | |
CN114169623A (en) | Power equipment fault analysis method and device, electronic equipment and storage medium |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination |