CN112327337B - Intersection reconstruction method, device, equipment and storage medium - Google Patents
Intersection reconstruction method, device, equipment and storage medium Download PDFInfo
- Publication number
- CN112327337B CN112327337B CN202011205254.9A CN202011205254A CN112327337B CN 112327337 B CN112327337 B CN 112327337B CN 202011205254 A CN202011205254 A CN 202011205254A CN 112327337 B CN112327337 B CN 112327337B
- Authority
- CN
- China
- Prior art keywords
- intersection
- information
- determining
- track information
- target area
- 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.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 68
- 238000013136 deep learning model Methods 0.000 claims abstract description 23
- 238000010586 diagram Methods 0.000 claims description 51
- 238000001514 detection method Methods 0.000 claims description 42
- 238000013527 convolutional neural network Methods 0.000 claims description 25
- 238000013528 artificial neural network Methods 0.000 claims description 20
- 230000010365 information processing Effects 0.000 claims description 16
- 238000004590 computer program Methods 0.000 claims description 8
- 238000004891 communication Methods 0.000 abstract description 15
- 238000003709 image segmentation Methods 0.000 description 6
- 238000010276 construction Methods 0.000 description 4
- 238000013135 deep learning Methods 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 4
- 238000012545 processing Methods 0.000 description 4
- 241000272183 Geococcyx californianus Species 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 230000000877 morphologic effect Effects 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 238000012549 training Methods 0.000 description 2
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 238000000802 evaporation-induced self-assembly Methods 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000002093 peripheral effect Effects 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S19/00—Satellite radio beacon positioning systems; Determining position, velocity or attitude using signals transmitted by such systems
- G01S19/38—Determining a navigation solution using signals transmitted by a satellite radio beacon positioning system
- G01S19/39—Determining a navigation solution using signals transmitted by a satellite radio beacon positioning system the satellite radio beacon positioning system transmitting time-stamped messages, e.g. GPS [Global Positioning System], GLONASS [Global Orbiting Navigation Satellite System] or GALILEO
- G01S19/393—Trajectory determination or predictive tracking, e.g. Kalman filtering
-
- 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
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Radar, Positioning & Navigation (AREA)
- Remote Sensing (AREA)
- General Physics & Mathematics (AREA)
- Biophysics (AREA)
- Molecular Biology (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Computer Networks & Wireless Communication (AREA)
- Traffic Control Systems (AREA)
- Navigation (AREA)
Abstract
The embodiment of the disclosure provides an intersection reconstruction method, an intersection reconstruction device, intersection reconstruction equipment and a storage medium. The method comprises the following steps: acquiring track information corresponding to the intersection to be processed; determining the type and/or the access point information of the intersection through a deep learning model according to the track information; and reconstructing the intersection according to the type and/or the access point information of the intersection. The method and the device can solve the problem that the accuracy of intersection reconstruction is low due to the fact that a plurality of noise road sections exist at the intersection reconstructed by the prior art or important topological communication is lost.
Description
Technical Field
The embodiment of the disclosure relates to the technical field of maps, in particular to a method, a device, equipment and a storage medium for reconstructing an intersection.
Background
Along with the continuous development of positioning technology and continuous improvement of travel convenience, the application of navigation functions is also becoming wider and wider. Accurate electronic maps play an important role in navigation applications, but traditional ways of manually drawing electronic maps are very labor and material consuming and inefficient.
With the wide use of positioning sensors such as GPS (Global Positioning System ), road network automatic reconstruction technology (MAP INFERENCE) based on large-scale GPS track information will greatly reduce the manufacturing cost of electronic maps. The existing reconstruction technology processes track information by a clustering method, so that the reconstruction of the intersection is realized. However, the intersection reconstructed by the method often does not conform to the structure of the real intersection, and a lot of noise road sections exist or important topological communication is lost, so that the accuracy of intersection reconstruction is low.
Disclosure of Invention
The embodiment of the disclosure provides an intersection reconstruction method, device, equipment and storage medium, which can solve the problem that the accuracy of intersection reconstruction is lower because a plurality of noise road sections exist or important topological communication is lost in an intersection reconstructed by the prior art.
In a first aspect, an embodiment of the present disclosure provides an intersection reconstruction method, including: acquiring track information corresponding to the intersection to be processed; determining the type and/or the access point information of the intersection through a deep learning model according to the track information; and reconstructing the intersection according to the type and/or the access point information of the intersection.
In a second aspect, an embodiment of the present disclosure provides an intersection reconstruction device, including: the track information acquisition module is used for acquiring track information corresponding to the intersection to be processed; the track information processing module is used for determining the type and/or the access point information of the intersection through a deep learning model according to the track information; and the intersection reconstruction module is used for reconstructing the intersection according to the type and/or the access point information of the intersection.
In a third aspect, an embodiment of the present disclosure provides an electronic device, including:
A memory;
A processor; and
A computer program;
wherein the computer program is stored in the memory and configured to be executed by the processor to implement the method according to the first aspect.
In a fourth aspect, embodiments of the present disclosure provide a computer-readable storage medium having stored thereon a computer program for execution by a processor to implement the method of the first aspect.
According to the intersection reconstruction method, device, equipment and storage medium, the type and/or the access point information of the intersection are determined through the deep learning model according to the track information by acquiring the track information corresponding to the intersection to be processed, so that intersection topology communication is guaranteed, then the intersection is reconstructed according to the type and/or the access point information of the intersection, a more real and accurate intersection topology structure is generated, and accuracy of intersection reconstruction is improved.
Drawings
Fig. 1 is a schematic view of an application scenario of an intersection reconstruction method provided by an embodiment of the present disclosure;
fig. 2 is a schematic flow chart of an intersection reconstruction method according to an embodiment of the disclosure;
Fig. 3 is a schematic flow chart of an intersection reconstruction method according to another embodiment of the disclosure;
FIG. 4 is a schematic diagram of a topology map provided by a further embodiment of the present disclosure;
Fig. 5 is a schematic structural diagram of an intersection reconstruction device according to an embodiment of the present disclosure;
fig. 6 is a schematic structural diagram of an electronic device according to an embodiment of the disclosure.
Specific embodiments of the present disclosure have been shown by way of the above drawings and will be described in more detail below. These drawings and the written description are not intended to limit the scope of the disclosed concepts in any way, but rather to illustrate the disclosed concepts to those skilled in the art by reference to specific embodiments.
Detailed Description
Reference will now be made in detail to exemplary embodiments, examples of which are illustrated in the accompanying drawings. When the following description refers to the accompanying drawings, the same numbers in different drawings refer to the same or similar elements, unless otherwise indicated. The implementations described in the following exemplary examples are not representative of all implementations consistent with the present disclosure. Rather, they are merely examples of apparatus and methods consistent with some aspects of the present disclosure as detailed in the accompanying claims.
In the prior art, the existing reconstruction technology processes track information by a clustering method, so that the reconstruction of the intersection is realized. However, the intersection reconstructed by the method often does not conform to the structure of the real intersection, and a lot of noise road sections exist or important topological communication is lost, so that the accuracy of intersection reconstruction is low.
In order to solve the technical problems, the technical concept of the present disclosure is to provide a multi-task deep learning framework, combine CNN and GCN to learn the shape and the topology characteristics of the intersection at the same time, predict the category and the entry/exit point to which the intersection belongs, and then reconstruct the intersection based on the type of the intersection and the entry/exit point of the intersection, wherein the reconstructed intersection not only accords with the real intersection structure, but also ensures the topology communication, and improves the accuracy of intersection reconstruction.
Fig. 1 is an application scenario schematic diagram of an intersection reconstruction method provided by an embodiment of the present disclosure. As shown in fig. 1, first, a track of a target area is acquired, and a range frame where each intersection is located is determined according to the track. For each intersection, a topological feature map is constructed by a graph convolution neural network (Graph Convolutional Network, GCN) with the original trajectory (i.e., trajectory information) as input. The topology feature map is passed through convolutional neural network (Convolutional Neural Networks, CNN) to obtain the type of intersection, and at the same time passed through target detection algorithm to obtain access point, where the target detection algorithm can be an improved one-dimensional target detection algorithm. The intersection can be reconstructed according to the type and the access point of the intersection. The intersection reconstructed by the method accords with the real intersection structure, and has no many noise road sections or loses important topological communication, thereby improving the accuracy of intersection reconstruction.
Specifically, the intersection reconstruction method provided by the disclosure aims to solve the technical problems in the prior art.
The following describes the technical solutions of the present disclosure and how the technical solutions of the present disclosure solve the above technical problems in detail with specific embodiments. The following embodiments may be combined with each other, and the same or similar concepts or processes may not be described in detail in some embodiments. Embodiments of the present disclosure will be described below with reference to the accompanying drawings.
Fig. 2 is a schematic flow chart of an intersection reconstruction method according to an embodiment of the disclosure. Aiming at the technical problems in the prior art, the embodiment of the disclosure provides an intersection reconstruction method, which comprises the following specific steps:
S201, obtaining track information corresponding to the intersection to be processed.
The execution body of the embodiment of the disclosure may be a server, and the server may collect track information of the GPS track through the collecting device, where the track information may include a speed and a direction angle of the track point.
In practical application, a road network reconstruction is performed on a region, a GPS track of an actual road network corresponding to the region needs to be acquired, then a range frame where each intersection is located is identified for the region, and the region corresponding to the electronic map is drawn or reconstructed according to track information for the range frame where each intersection is located.
Specifically, how to obtain the track information corresponding to the intersection to be processed can be realized through the following steps:
and a1, acquiring track information corresponding to a target area to be reconstructed in the electronic map.
And a2, determining a range frame of each intersection in the target area according to the track information corresponding to the target area.
And a3, determining track information corresponding to each intersection according to the range frame where each intersection is located.
The intersection to be processed is any intersection in the target area.
In the embodiment of the disclosure, the track information corresponding to the intersection to be processed may be obtained by intersection detection. In the crossing detection, the whole crossing is divided into two parts of road sections and crossing. The intersection detection can comprise two parts, namely space track feature map construction and detection of an intersection range frame by using an image target detection algorithm, such as an SSD algorithm.
Specifically, track information corresponding to a target area to be reconstructed in an electronic map is firstly obtained by using a GPS sensor, then a space track feature map is constructed based on the track information corresponding to the target area, a range frame where each intersection is located in the target area is detected based on the space track feature map, and then track information of the range frame where each intersection to be processed is located, namely track information corresponding to the intersection to be processed, is determined according to the track information. The intersection to be processed here is any intersection in the target area.
Optionally, how to determine the range frame where each intersection is located in the target area may be implemented by the following steps:
and b1, carrying out grid division on the target area, and determining the two-dimensional spatial characteristics of the target area according to the track information in each grid.
And b2, determining a range frame of each intersection of the target area through a target detection algorithm according to the two-dimensional spatial characteristics of the target area.
In the embodiment of the disclosure, the target area is subjected to grid division, for example, (5 m) grid division is performed, the number of track points, average direction angles and average speeds of the GPS falling into each grid are counted to form two-dimensional spatial features of the target area, then the SSD algorithm, namely the target detection algorithm, is used for detecting each intersection of the target area according to the two-dimensional spatial features of the target area, and the range frame of each intersection of the target area is determined. The average direction angle is an average value of direction angles of the track points of the GPS falling in each grid, and the average speed is an average value of speeds of the track points of the GPS falling in each grid.
After the target area is meshed, the continuous information of the track can be determined through the track points of the GPS according to the range frame where each determined intersection is located, for example, which grid is accessed to pass through and then which grid is accessed out.
S202, determining the type and/or the access point information of the intersection through a deep learning model according to the track information.
In the embodiment of the disclosure, after the range frame of each intersection is obtained, for each intersection, classification and access points of the intersection can be predicted simultaneously through a built multi-task deep learning network, namely, the type of the intersection is identified, and the position of the access point of the intersection is checked.
Specifically, by training the GCN model and the CNN model, the graph convolution neural network and the convolution neural network simultaneously learn the type of the intersection and the topological feature graph to construct a multi-task deep learning model. Wherein GCN models the track continuity characteristic, CNN carries on the classification prediction of crossing. The input quantity of the GCN model is track information of the crossing to be processed and track continuity information of the crossing to be processed, a topological feature map is output through weighted modeling of the GCN, then the topological feature map is used as the input quantity of the CNN model, crossing classification is carried out to obtain the type of the crossing to be processed, and according to the topological feature map, an in-out point of the crossing to be processed is identified through a target detection algorithm, such as an improved one-dimensional target detection method, so as to obtain in-out point information, wherein the in-out point information can comprise position information of the crossing to be processed, where each in-out point is located, and whether the in-out point is a communication point or a non-communication point.
Optionally, according to the track information, determining the type and/or the access point information of the intersection through a deep learning model may be implemented in the following manner: and constructing a topological feature diagram of the intersection through a deep learning model according to the track information, and determining the type and/or the access point information of the intersection according to the topological feature diagram.
The topological feature map can be constructed through a map convolution neural network according to the track information; the type of the intersection and/or the information of the access point can be determined according to the topological feature map.
Specifically, for intersection classification, a complete continuous GPS trajectory may be used to construct an intersection topological feature map, and GCN is used to model trajectory continuity features, wherein the GCN is used to learn the topological connectivity differences of intersections. The CNN is combined to conduct intersection classification prediction, and the CNN learns morphological differences, so that the task of intersection classification is achieved, and the task of intersection classification has the topological connectivity of identifying intersections.
Aiming at the detection of the access points, the information of the access points of the intersections is determined through a target detection algorithm according to the topological feature diagram. The implementation mode is as follows: and detecting the outermost grid of the intersection through a target detection algorithm according to the topological feature diagram, and determining the information of the access point of the intersection.
The access point information comprises the position and connectivity of the access point of the intersection.
Specifically, a target detection algorithm is used for the outermost grid based on the topological feature map, image segmentation is performed in the form of a sliding window, and the positions of the in-out points of the intersections and the connectivity of the in-out points of the intersections are predicted.
S203, reconstructing the intersection according to the type and/or the access point information of the intersection.
In the embodiment of the disclosure, the actual intersection structure can be determined according to the type of the intersection and the continuity information of the track, so that the intersection reconstruction is realized; the actual intersection structure can be determined according to the position of the access point of the intersection and the connectivity of the access point and the continuity information of the track, so that the intersection reconstruction is realized; the actual intersection structure can be determined by combining the type of the intersection and the position of the access point, so that the intersection reconstruction is realized.
Specifically, referring to fig. 3, fig. 3 is a schematic flow chart of an intersection reconstruction method according to another embodiment of the disclosure. The embodiment of the present disclosure describes S203 in detail on the basis of the above embodiment. Reconstructing the intersection according to the type and/or the access point information of the intersection may include:
s301, acquiring an intersection template corresponding to the type of the intersection.
In the embodiment of the disclosure, the used intersection templates can be determined according to the types of intersections, and the intersection templates can comprise templates corresponding to the types of various intersections, namely, one intersection template corresponds to one intersection type. The types of intersections may include T-shapes, Y-shapes, cross-shapes, X-shapes, offset, annular shapes, and the like.
S302, according to the information of the access points of the intersections, determining the corresponding relation between the access points of the intersections and the access points of the intersection templates.
In this embodiment, the information of the access point of the intersection includes the position of the access point and the connectivity of the access point, and according to the connectivity of the access point, the position of the access point of the intersection template can be determined on the intersection template, so as to form the corresponding relationship between the access point of the intersection and the access point of the intersection template.
S303, reconstructing the intersection according to the determined corresponding relation.
In the embodiment of the disclosure, the used intersection template can be determined according to the type of the intersection, then the access point of the intersection template and the detected access point are subjected to one-to-one matching correspondence, the final intersection reconstruction result is obtained through the stretching and deformation, and the intersection reconstruction is realized, so that the deformable priori intersection template is used, the priori information of the intersection construction standard in reality is fully utilized, and more real and accurate intersection topological expression is generated compared with the method for directly reconstructing the intersection by using trajectory clustering and the like in the prior art. The method solves the problems that in the prior art, a CNN method similar to image segmentation is used for extracting the road center line, the extracted road center line is represented by pixel points, for example, the pixel points can be represented as roads or non-roads, but road network topology cannot be constructed based on the pixel points, the same intersection topology problem as the track clustering method exists, namely the intersection topology is not in line with a real intersection structure, a plurality of noise road sections exist or important topology communication is lost, and the like.
Meanwhile, the method is different from the prior art that the track is converted into the thermodynamic diagram (because the thermodynamic diagram is difficult to distinguish whether the intersection is connected or not, such as a connected-T intersection or a non-connected-T intersection, so that the intersection reconstructed by the existing method has topology connection errors), and the continuity information of the track is ignored, so that the accuracy of identifying the topology connectivity of the intersection is low. According to the method, the type and/or the access point information of the intersection are determined through a deep learning model according to the track information, so that the topological communication of the intersection is guaranteed, then the intersection is rebuilt according to the type and/or the access point information of the intersection, a more real and accurate intersection topological structure is generated, and the accuracy of intersection rebuilding is improved.
Optionally, on the basis of the foregoing embodiments, embodiments of the present disclosure describe how to construct a topology feature map in detail. According to the track information, constructing a topological feature diagram of the intersection through a diagram convolutional neural network, wherein the topological feature diagram can be realized through the following steps:
and c1, determining the two-dimensional space characteristics of the intersection according to the track information of the intersection.
In the embodiment of the disclosure, firstly, the target area is subjected to grid division, for example, grid division is performed by (5 m x 5 m), and the number of track points, average direction angles and average speed of the GPS falling into each grid are counted; and then based on the determined range frame of each intersection, counting grid information contained in the range frame of each intersection, such as the number of grids divided by the intersection, the number of track points corresponding to each grid, the average direction angle and the average speed, and forming two-dimensional space features of the intersection based on the grid information.
And c2, constructing a topological feature map of the intersection through a graph convolution neural network according to the two-dimensional spatial feature of the intersection and the continuity of the track information in the intersection.
Wherein the intersection is divided into a plurality of grids, and the two-dimensional spatial characteristics of the intersection comprise at least one of the following information of each grid in the intersection: the number of trajectory information within the grid, the average direction angle, the average speed.
In the embodiment of the disclosure, two-dimensional spatial characteristics of the intersection and continuity of track information in the intersection are used as input quantities of a trained graph convolution neural network, topology communication differences of the intersection are learned through the graph convolution neural network, track continuity characteristics of the intersection are output, and a topology characteristic graph is further constructed. Referring to fig. 4, a part of the grids in the intersection is shown, wherein, taking the grid 8 as an example, a topological feature map of the grid 8 is passed.
Wherein, the GPS track is a track passing through the grid 1 and the grid 8, and the topological characteristics corresponding to the grids 1 to 8 in the topological characteristic diagram are 0.1; two tracks pass through the grid 2 and the grid 8 in the GPS track, and the topological characteristic corresponding to the 2 to 8 in the topological characteristic diagram is 0.2; two tracks pass through the grid 3 and the grid 8 in the GPS track, and the topological characteristics corresponding to 3 to 8 in the topological characteristic diagram are 0.2; the GPS track is track-free through the grid 4 and the grid 8, and the topological features 4 to 8 in the topological feature map are not topological features; three tracks pass through the grid 5 and the grid 8 in the GPS track, and the topological characteristic corresponding to 5 to 8 in the topological characteristic diagram is 0.3; one track passes through the grid 6 and the grid 8 in the GPS track, and the topological characteristic corresponding to 6 to 8 in the topological characteristic diagram is 0.1; the GPS track passes through the no tracks of the grid 7 and the grid 8, so that the topology features 7 to 8 in the topology feature map are free; and if the GPS track passes through the mesh 9 and the mesh 8, no topological feature exists from 9 to 8 in the topological feature map, and so on, a topological feature map is formed for each mesh, so that the topological feature map of the whole intersection is formed.
Then, based on the topological feature diagram of the whole intersection, for the detection of the access point of the intersection, the method of target detection is used for the outermost points (such as 1,2, 3, 6 and 8 in the example of fig. 4), and the positions of the access points of the intersection and the connectivity of the access points are predicted by carrying out image segmentation through a sliding window.
Optionally, after reconstructing each intersection of the target area, a road section between each intersection of the target area may be determined according to the track information of the target area, so as to form a complete road network.
Specifically, according to the track information of the target area, a RoadRunner method is used to realize road section connection between intersections, and finally a complete road network is formed.
Therefore, the road network structure with the topological connectivity is real and accurate and is formed by adopting the road junction and road network reconstruction frame based on deep learning to realize road junction detection, road junction classification, road junction identification, road junction reconstruction and road segment reconstruction.
Fig. 5 is a schematic structural diagram of an intersection reconstruction device according to an embodiment of the present disclosure. The intersection reconstruction device may specifically be a server in the above embodiment. The intersection reconstruction device provided by the embodiment of the present disclosure may execute the processing flow provided by the intersection reconstruction method embodiment, as shown in fig. 5, where the intersection reconstruction device includes: the track information acquisition module 501 is used for acquiring track information corresponding to the intersection to be processed; the track information processing module 502 is configured to determine, according to the track information, a type and/or entry and exit point information of the intersection through a deep learning model; the intersection reconstruction module 503 is configured to reconstruct the intersection according to the type and/or the information of the access point of the intersection.
The track information acquisition module 501, the track information processing module 502 and the intersection reconstruction module 503 are configured to determine the type and/or the access point information of the intersection through a deep learning model according to track information corresponding to the intersection to be processed, so as to ensure the intersection topology communication, and then reconstruct the intersection according to the type and/or the access point information of the intersection, thereby generating a more real and accurate intersection topology structure and improving the accuracy of intersection reconstruction.
In the embodiment of the disclosure, the server may collect the track information of the GPS track through the collecting device, where the track information may include the speed and the direction angle of the track point.
Optionally, the track information acquisition module is specifically configured to: acquiring track information corresponding to a target area to be reconstructed in an electronic map; determining a range frame of each intersection in the target area according to the track information corresponding to the target area; determining track information corresponding to each intersection according to the range frame of each intersection; the intersection to be processed is any intersection in the target area.
In the embodiment of the disclosure, the track information corresponding to the intersection to be processed may be obtained by intersection detection. In the crossing detection, the whole crossing is divided into two parts of road sections and crossing. The intersection detection can comprise two parts, namely space track feature map construction and detection of an intersection range frame by using an image target detection algorithm, such as an SSD algorithm.
Specifically, firstly, a track information acquisition module acquires track information corresponding to a target area to be reconstructed in an electronic map by using a GPS sensor, then constructs a space track feature map based on the track information corresponding to the target area, detects a range frame where each intersection is located in the target area based on the space track feature map, and then determines track information of the range frame where each intersection to be processed is located, namely track information corresponding to the intersection to be processed, according to the track information. The intersection to be processed here is any intersection in the target area.
Optionally, how to determine the range frame of each intersection in the target area can be implemented by the track information acquisition module; the track information acquisition module is realized and is also specifically used for: dividing the target area into grids, and determining the two-dimensional space characteristics of the target area according to track information in each grid; and determining a range frame of each intersection of the target area through a target detection algorithm according to the two-dimensional space characteristics of the target area.
In the embodiment of the disclosure, firstly, the target area is subjected to grid division, for example, (5 m) grid division is performed, the number of track points, average direction angles and average speeds of the GPS falling into each grid are counted to form two-dimensional space features of the target area, then, according to the two-dimensional space features of the target area, the SSD algorithm, namely a target detection algorithm, is used for detecting all intersections of the target area, and the range frame of all intersections of the target area is determined. The average direction angle is an average value of direction angles of the track points of the GPS falling in each grid, and the average speed is an average value of speeds of the track points of the GPS falling in each grid.
After the target area is meshed, the continuous information of the track can be determined through the track points of the GPS according to the range frame where each determined intersection is located, for example, which grid is accessed to pass through and then which grid is accessed out.
Optionally, the deep learning model is obtained by simultaneously learning the type of the intersection and the topological feature map through a graph convolution neural network and a convolution neural network.
Optionally, the track information processing module is specifically configured to: and constructing a topological feature diagram of the intersection through a deep learning model according to the track information, and determining the type and/or the access point information of the intersection according to the topological feature diagram.
In the embodiment of the disclosure, after the range frame of each intersection is obtained, for each intersection, the track information processing module can predict classification and access points of the intersection at the same time through a built multi-task deep learning network, namely, the type of the intersection is identified, and the position of the access point of the intersection is checked.
Specifically, by training the GCN model and the CNN model, the graph convolution neural network and the convolution neural network simultaneously learn the type of the intersection and the topological feature graph to construct a multi-task deep learning model. Wherein GCN models the track continuity characteristic, CNN carries on the classification prediction of crossing. The input quantity of the GCN model is track information of the to-be-processed intersection and track continuity information of the to-be-processed intersection, a topological feature map is output through weighted modeling of the GCN, then the topological feature map is used as the input quantity of the CNN model, intersection classification is carried out to obtain the type of the to-be-processed intersection, and an in-and-out point of the to-be-processed intersection is identified through a target detection algorithm according to the topological feature map to obtain in-and-out point information, wherein the in-and-out point information can comprise position information of the to-be-processed intersection where each in-and-out point is located and whether the in-and-out point is a connected point or a non-connected point.
Optionally, the track information processing module is further specifically configured to: constructing a topological feature diagram of the intersection through a graph convolutional neural network according to the track information; and determining the type and/or the access point information of the intersection according to the topological feature diagram.
Optionally, the track information processing module is further specifically configured to: determining the type of the intersection through a convolutional neural network according to the topological feature diagram; and/or determining the information of the access point of the intersection through a target detection algorithm according to the topological feature diagram.
In the embodiment of the disclosure, the topological feature map can be constructed through a map convolution neural network according to the track information; the type of the intersection and/or the information of the access point can be determined according to the topological feature map.
Specifically, for intersection classification, a complete continuous GPS trajectory may be used to construct an intersection topological feature map, and GCN is used to model trajectory continuity features, wherein the GCN is used to learn the topological connectivity differences of intersections. The CNN is combined to conduct intersection classification prediction, and the CNN learns morphological differences, so that the task of intersection classification is achieved, and the task of intersection classification has the topological connectivity of identifying intersections.
Optionally, the track information processing module is further specifically configured to: detecting the outermost grid of the intersection through a target detection algorithm according to the topological feature diagram, and determining the information of the access point of the intersection; the access point information comprises the position and connectivity of the access point of the intersection.
In the embodiment of the disclosure, for the detection of the access point, the detection of the outermost grid of the intersection according to the topological feature diagram through the track information processing module root is realized, and the access point information of the intersection is determined through a target detection algorithm.
The access point information comprises the position and connectivity of the access point of the intersection.
Specifically, a target detection algorithm is used for the outermost grid based on the topological feature map, image segmentation is performed in the form of a sliding window, and the positions of the in-out points of the intersections and the connectivity of the in-out points of the intersections are predicted.
Optionally, the track information processing module is further specifically configured to: determining the two-dimensional space characteristics of the intersection according to the track information of the intersection; constructing a topological feature map of the intersection through a graph convolution neural network according to the two-dimensional spatial feature of the intersection and the continuity of track information in the intersection; wherein the intersection is divided into a plurality of grids, and the two-dimensional spatial characteristics of the intersection comprise at least one of the following information of each grid in the intersection: the number of trajectory information within the grid, the average direction angle, the average speed.
In the embodiment of the disclosure, firstly, the target area is subjected to grid division, for example, grid division is performed by (5 m x 5 m), and the number of track points, average direction angles and average speed of the GPS falling into each grid are counted; and then based on the determined range frame of each intersection, counting grid information contained in the range frame of each intersection, such as the number of grids divided by the intersection, the number of track points corresponding to each grid, the average direction angle and the average speed, and forming two-dimensional space features of the intersection based on the grid information. Wherein the intersection is divided into a plurality of grids, and the two-dimensional spatial characteristics of the intersection comprise at least one of the following information of each grid in the intersection: the number of trajectory information within the grid, the average direction angle, the average speed.
Taking the two-dimensional spatial characteristics of the crossing and the continuity of track information in the crossing as the input quantity of a trained graph convolution neural network, learning the topological connection difference of the crossing through the graph convolution neural network, outputting the track continuity characteristics of the crossing, and further constructing a topological characteristic graph. Referring to fig. 4, a part of the grids in the intersection is shown, wherein, taking the grid 8 as an example, a topological feature map of the grid 8 is passed.
Wherein, the GPS track is a track passing through the grid 1 and the grid 8, and the topological characteristics corresponding to the grids 1 to 8 in the topological characteristic diagram are 0.1; two tracks pass through the grid 2 and the grid 8 in the GPS track, and the topological characteristic corresponding to the 2 to 8 in the topological characteristic diagram is 0.2; two tracks pass through the grid 3 and the grid 8 in the GPS track, and the topological characteristics corresponding to 3 to 8 in the topological characteristic diagram are 0.2; the GPS track is track-free through the grid 4 and the grid 8, and the topological features 4 to 8 in the topological feature map are not topological features; three tracks pass through the grid 5 and the grid 8 in the GPS track, and the topological characteristic corresponding to 5 to 8 in the topological characteristic diagram is 0.3; one track passes through the grid 6 and the grid 8 in the GPS track, and the topological characteristic corresponding to 6 to 8 in the topological characteristic diagram is 0.1; the GPS track passes through the no tracks of the grid 7 and the grid 8, so that the topology features 7 to 8 in the topology feature map are free; and if the GPS track passes through the mesh 9 and the mesh 8, no topological feature exists from 9 to 8 in the topological feature map, and so on, a topological feature map is formed for each mesh, so that the topological feature map of the whole intersection is formed.
Then, based on the topological feature diagram of the whole intersection, for the detection of the access point of the intersection, the method of target detection is used for the outermost points (such as 1,2, 3, 6 and 8 in the example of fig. 4), and the positions of the access points of the intersection and the connectivity of the access points are predicted by carrying out image segmentation through a sliding window.
Optionally, the intersection reconstruction module is specifically configured to: acquiring an intersection template corresponding to the type of the intersection; according to the information of the access points of the intersections, determining the corresponding relation between the access points of the intersections and the access points of the intersection templates; and reconstructing the intersection according to the determined corresponding relation.
In the embodiment of the disclosure, the intersection reconstruction module can determine the used intersection template according to the type of the intersection, then performs one-to-one matching on the access point of the intersection template and the detected access point, stretches the corresponding relationship, and deforms to obtain the final intersection reconstruction result to realize the reconstruction of the intersection, so that the deformable prior intersection template is used, prior information of the intersection construction standard in reality is fully utilized, and more real and accurate intersection topology expression is generated compared with a method for directly reconstructing the intersection by using track clustering and the like in the prior art. The method solves the problems that in the prior art, a CNN method similar to image segmentation is used for extracting the road center line, the extracted road center line is represented by pixel points, for example, the pixel points can be represented as roads or non-roads, but road network topology cannot be constructed based on the pixel points, the same intersection topology problem as the track clustering method exists, namely the intersection topology is not in line with a real intersection structure, a plurality of noise road sections exist or important topology communication is lost, and the like.
Meanwhile, the method is different from the prior art that the track is converted into the thermodynamic diagram (because the thermodynamic diagram is difficult to distinguish whether the intersection is connected or not, such as a connected-T intersection or a non-connected-T intersection, so that the intersection reconstructed by the existing method has topology connection errors), and the continuity information of the track is ignored, so that the accuracy of identifying the topology connectivity of the intersection is low. According to the method, the type and/or the access point information of the intersection are determined through a deep learning model according to the track information, so that the topological communication of the intersection is guaranteed, then the intersection is rebuilt according to the type and/or the access point information of the intersection, a more real and accurate intersection topological structure is generated, and the accuracy of intersection rebuilding is improved.
Optionally, the apparatus further includes: a road section determining module; and the road section determining module is used for determining road sections among the intersections of the target area according to the track information of the target area after reconstructing the intersections of the target area so as to form a complete road network.
Specifically, according to the track information of the target area, a RoadRunner method is used to realize road section connection between intersections, and finally a complete road network is formed.
The intersection reconstruction device of the embodiment shown in fig. 5 may be used to implement the technical solution of the method embodiment described in the first aspect, and its implementation principle and technical effects are similar, and are not repeated here.
Fig. 6 is a schematic structural diagram of an electronic device according to an embodiment of the disclosure. The electronic device may specifically be a server in the above-described embodiments. The electronic device provided in the embodiment of the present disclosure may execute the processing flow provided in the intersection reconstruction method embodiment, as shown in fig. 6, where the electronic device 600 provided in the embodiment includes: at least one processor 601 and a memory 602. The processor 601 and the memory 602 are connected by a bus 603.
In a specific implementation, at least one processor 601 executes computer-executable instructions stored in the memory 602, so that the at least one processor 601 performs the method in the above method embodiments.
The specific implementation process of the processor 601 may refer to the above-mentioned method embodiment, and its implementation principle and technical effects are similar, and this embodiment will not be described herein again.
In the embodiment shown in fig. 6, it should be understood that the processor may be a central processing unit (english: central Processing Unit, abbreviated as CPU), other general purpose processors, digital signal processor (english: DIGITAL SIGNAL processor, abbreviated as DSP), application-specific integrated circuit (english: application SPECIFIC INTEGRATED circuit, abbreviated as ASIC), and the like. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like. The steps of a method disclosed in connection with the present invention may be embodied directly in a hardware processor for execution, or in a combination of hardware and software modules in a processor for execution.
The memory may comprise high speed RAM memory or may further comprise non-volatile storage NVM, such as at least one disk memory.
The bus may be an industry standard architecture (Industry Standard Architecture, ISA) bus, an external device interconnect (PERIPHERAL COMPONENT INTERCONNECT, PCI) bus, or an extended industry standard architecture (Extended Industry Standard Architecture, EISA) bus, among others. The buses may be divided into address buses, data buses, control buses, etc. For ease of illustration, the buses in the drawings of the present disclosure are not limited to only one bus or to one type of bus.
In addition, the embodiment of the present disclosure also provides a computer readable storage medium having stored thereon a computer program that is executed by a processor to implement the intersection reconstruction method described in the above embodiment.
The computer readable storage medium described above may be implemented by any type of volatile or non-volatile memory device or combination thereof, such as Static Random Access Memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk, or optical disk. A readable storage medium can be any available medium that can be accessed by a general purpose or special purpose computer.
An exemplary readable storage medium is coupled to the processor such the processor can read information from, and write information to, the readable storage medium. In the alternative, the readable storage medium may be integral to the processor. The processor and the readable storage medium may reside in an Application SPECIFIC INTEGRATED Circuits (ASIC). The processor and the readable storage medium may reside as discrete components in a device.
Those of ordinary skill in the art will appreciate that: all or part of the steps for implementing the method embodiments described above may be performed by hardware associated with program instructions. The foregoing program may be stored in a computer readable storage medium. The program, when executed, performs steps including the method embodiments described above; and the aforementioned storage medium includes: various media that can store program code, such as ROM, RAM, magnetic or optical disks.
Finally, it should be noted that: the above embodiments are only for illustrating the technical solution of the present invention, and not for limiting the same; although the invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical scheme described in the foregoing embodiments can be modified or some or all of the technical features thereof can be replaced by equivalents; such modifications and substitutions do not depart from the spirit of the invention.
Claims (20)
1. An intersection reconstruction method, comprising:
Acquiring track information corresponding to the intersection to be processed;
Determining the type and/or the access point information of the intersection through a deep learning model according to the track information;
Reconstructing the intersection according to the type and/or the access point information of the intersection;
obtaining track information corresponding to the intersection to be processed, including:
acquiring track information corresponding to a target area to be reconstructed in an electronic map;
dividing the target area into grids, and determining the two-dimensional space characteristics of the target area according to track information in each grid;
Determining a range frame of each intersection of the target area through a target detection algorithm according to the two-dimensional space characteristics of the target area;
determining track information corresponding to each intersection according to the range frame of each intersection;
The intersection to be processed is any intersection in the target area.
2. The method of claim 1, wherein determining the type of intersection and/or the entry and exit point information from the trajectory information by a deep learning model comprises:
and constructing a topological feature diagram of the intersection through a deep learning model according to the track information, and determining the type and/or the access point information of the intersection according to the topological feature diagram.
3. The method according to claim 2, wherein constructing a topological feature map of the intersection by a deep learning model according to the trajectory information, and determining the type and/or the in-out point information of the intersection according to the topological feature map, comprises:
constructing a topological feature diagram of the intersection through a graph convolutional neural network according to the track information;
and determining the type and/or the access point information of the intersection according to the topological feature diagram.
4. A method according to claim 3, wherein constructing a topological feature map of the intersection from the trajectory information by a graph convolutional neural network comprises:
Determining the two-dimensional space characteristics of the intersection according to the track information of the intersection;
constructing a topological feature map of the intersection through a graph convolution neural network according to the two-dimensional spatial feature of the intersection and the continuity of track information in the intersection;
wherein the intersection is divided into a plurality of grids, and the two-dimensional spatial characteristics of the intersection comprise at least one of the following information of each grid in the intersection: the number of trajectory information within the grid, the average direction angle, the average speed.
5. The method of claim 4, wherein determining the type of intersection and/or the access point information from the topological feature map comprises:
determining the type of the intersection through a convolutional neural network according to the topological feature diagram; and/or the number of the groups of groups,
And determining the information of the access points of the intersections through a target detection algorithm according to the topological feature diagram.
6. The method of claim 5, wherein determining the entry and exit point information of the intersection by a target detection algorithm based on the topology map comprises:
Detecting the outermost grid of the intersection through a target detection algorithm according to the topological feature diagram, and determining the information of the access point of the intersection;
the access point information comprises the position and connectivity of the access point of the intersection.
7. The method of claim 1, wherein reconstructing the intersection based on the type of intersection and/or the access point information comprises:
acquiring an intersection template corresponding to the type of the intersection;
According to the information of the access points of the intersections, determining the corresponding relation between the access points of the intersections and the access points of the intersection templates;
and reconstructing the intersection according to the determined corresponding relation.
8. The method as recited in claim 1, further comprising:
After reconstructing each intersection of the target area, determining road segments among the intersections of the target area according to the track information of the target area so as to form a complete road network.
9. The method of claim 5, wherein the deep learning model is obtained by simultaneously learning the type of intersection and the topological feature map through a graph convolution neural network and a convolution neural network.
10. An intersection reconstruction device, comprising:
the track information acquisition module is used for acquiring track information corresponding to the intersection to be processed;
The track information processing module is used for determining the type and/or the access point information of the intersection through a deep learning model according to the track information;
The intersection reconstruction module is used for reconstructing the intersection according to the type and/or the access point information of the intersection;
The track information acquisition module is specifically used for acquiring track information corresponding to a target area to be reconstructed in the electronic map;
dividing the target area into grids, and determining the two-dimensional space characteristics of the target area according to track information in each grid;
Determining a range frame of each intersection of the target area through a target detection algorithm according to the two-dimensional space characteristics of the target area;
determining track information corresponding to each intersection according to the range frame of each intersection;
The intersection to be processed is any intersection in the target area.
11. The apparatus according to claim 10, wherein the track information processing module is specifically configured to:
and constructing a topological feature diagram of the intersection through a deep learning model according to the track information, and determining the type and/or the access point information of the intersection according to the topological feature diagram.
12. The apparatus of claim 11, wherein the track information processing module is further specifically configured to:
constructing a topological feature diagram of the intersection through a graph convolutional neural network according to the track information;
and determining the type and/or the access point information of the intersection according to the topological feature diagram.
13. The apparatus of claim 12, wherein the track information processing module is further specifically configured to:
Determining the two-dimensional space characteristics of the intersection according to the track information of the intersection;
constructing a topological feature map of the intersection through a graph convolution neural network according to the two-dimensional spatial feature of the intersection and the continuity of track information in the intersection;
wherein the intersection is divided into a plurality of grids, and the two-dimensional spatial characteristics of the intersection comprise at least one of the following information of each grid in the intersection: the number of trajectory information within the grid, the average direction angle, the average speed.
14. The apparatus of claim 13, wherein the track information processing module is further specifically configured to:
determining the type of the intersection through a convolutional neural network according to the topological feature diagram; and/or the number of the groups of groups,
And determining the information of the access points of the intersections through a target detection algorithm according to the topological feature diagram.
15. The apparatus of claim 14, wherein the track information processing module is further specifically configured to:
Detecting the outermost grid of the intersection through a target detection algorithm according to the topological feature diagram, and determining the information of the access point of the intersection;
the access point information comprises the position and connectivity of the access point of the intersection.
16. The device according to claim 10, wherein the intersection reconstruction module is specifically configured to:
acquiring an intersection template corresponding to the type of the intersection;
According to the information of the access points of the intersections, determining the corresponding relation between the access points of the intersections and the access points of the intersection templates;
and reconstructing the intersection according to the determined corresponding relation.
17. The apparatus of claim 10, wherein the apparatus further comprises: a road section determining module; and the road section determining module is used for determining road sections among the intersections of the target area according to the track information of the target area after reconstructing the intersections of the target area so as to form a complete road network.
18. The apparatus of claim 14, wherein the deep learning model is obtained by simultaneously learning the type of intersection and the topological feature map through a graph convolution neural network and a convolutional neural network.
19. An electronic device, comprising:
A memory;
A processor; and
A computer program;
wherein the computer program is stored in the memory and configured to be executed by the processor to implement the method of any one of claims 1-9.
20. A computer readable storage medium, on which a computer program is stored, which computer program, when being executed by a processor, implements the method according to any of claims 1-9.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011205254.9A CN112327337B (en) | 2020-11-02 | 2020-11-02 | Intersection reconstruction method, device, equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011205254.9A CN112327337B (en) | 2020-11-02 | 2020-11-02 | Intersection reconstruction method, device, equipment and storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112327337A CN112327337A (en) | 2021-02-05 |
| CN112327337B true CN112327337B (en) | 2024-04-19 |
Family
ID=74324380
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011205254.9A Active CN112327337B (en) | 2020-11-02 | 2020-11-02 | Intersection reconstruction method, device, equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112327337B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114201997B (en) * | 2022-02-17 | 2022-06-14 | 深圳依时货拉拉科技有限公司 | Intersection turning recognition method, device, equipment and storage medium |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109084798A (en) * | 2018-08-29 | 2018-12-25 | 武汉环宇智行科技有限公司 | Network issues the paths planning method at the control point with road attribute |
| CN110634291A (en) * | 2019-09-17 | 2019-12-31 | 武汉中海庭数据技术有限公司 | High-precision map topology automatic construction method and system based on crowdsourcing data |
| CN110728735A (en) * | 2019-09-17 | 2020-01-24 | 武汉中海庭数据技术有限公司 | Road-level topological layer construction method and system |
| CN110827544A (en) * | 2019-11-11 | 2020-02-21 | 重庆邮电大学 | Short-term traffic flow control method based on graph convolution recurrent neural network |
| CN111095291A (en) * | 2018-02-27 | 2020-05-01 | 辉达公司 | Real-time detection of lanes and boundaries by autonomous vehicles |
| CN111243277A (en) * | 2020-03-09 | 2020-06-05 | 山东大学 | Commuting vehicle space-time trajectory reconstruction method and system based on license plate recognition data |
| CN111540198A (en) * | 2020-04-17 | 2020-08-14 | 浙江工业大学 | Urban traffic situation recognition method based on directed graph convolution neural network |
| CN111583652A (en) * | 2020-05-21 | 2020-08-25 | 北京易华录信息技术股份有限公司 | Topological modeling method and system for traffic network |
-
2020
- 2020-11-02 CN CN202011205254.9A patent/CN112327337B/en active Active
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111095291A (en) * | 2018-02-27 | 2020-05-01 | 辉达公司 | Real-time detection of lanes and boundaries by autonomous vehicles |
| CN109084798A (en) * | 2018-08-29 | 2018-12-25 | 武汉环宇智行科技有限公司 | Network issues the paths planning method at the control point with road attribute |
| CN110634291A (en) * | 2019-09-17 | 2019-12-31 | 武汉中海庭数据技术有限公司 | High-precision map topology automatic construction method and system based on crowdsourcing data |
| CN110728735A (en) * | 2019-09-17 | 2020-01-24 | 武汉中海庭数据技术有限公司 | Road-level topological layer construction method and system |
| CN110827544A (en) * | 2019-11-11 | 2020-02-21 | 重庆邮电大学 | Short-term traffic flow control method based on graph convolution recurrent neural network |
| CN111243277A (en) * | 2020-03-09 | 2020-06-05 | 山东大学 | Commuting vehicle space-time trajectory reconstruction method and system based on license plate recognition data |
| CN111540198A (en) * | 2020-04-17 | 2020-08-14 | 浙江工业大学 | Urban traffic situation recognition method based on directed graph convolution neural network |
| CN111583652A (en) * | 2020-05-21 | 2020-08-25 | 北京易华录信息技术股份有限公司 | Topological modeling method and system for traffic network |
Non-Patent Citations (2)
| Title |
|---|
| 徐建闽等.基于梯度提升决策树的城市车辆路径链重构.华南理工大学学报( 自然科学版).2020,第48卷(第7期),第55-64页. * |
| 谭康 ; 刘建勋 ; 廖祝华 ; .一种基于GPS轨迹的道路拓扑生成方法.计算机科学.2015,(第09期),第37-40、55页. * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112327337A (en) | 2021-02-05 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN114120102A (en) | Boundary-optimized remote sensing image semantic segmentation method, device, equipment and medium | |
| CN110363049B (en) | Method and device for detecting, identifying and determining categories of graphic elements | |
| CN111694917B (en) | Vehicle abnormal track detection and model training method and device | |
| JP2018523875A (en) | Lane recognition modeling method, apparatus, storage medium and device, and lane recognition method, apparatus, storage medium and apparatus | |
| CN113065594B (en) | Road network extraction method and device based on Beidou data and remote sensing image fusion | |
| US20180365594A1 (en) | Systems and methods for generative learning | |
| CN111353580B (en) | Training method of target detection network, electronic equipment and storage medium | |
| CN112634369A (en) | Space and or graph model generation method and device, electronic equipment and storage medium | |
| He et al. | Learning scene dynamics from point cloud sequences | |
| CN115345905A (en) | Target object tracking method, device, terminal and storage medium | |
| CN112327337B (en) | Intersection reconstruction method, device, equipment and storage medium | |
| CN115424053A (en) | Small sample image identification method, device and equipment and storage medium | |
| CN115439654A (en) | Method and system for finely dividing weakly supervised farmland plots under dynamic constraint | |
| CN117083621A (en) | Detector training method, device and storage medium | |
| WO2021071422A1 (en) | Method, data processing apparatus and computer program product for determining road intersections | |
| CN112651417B (en) | License plate recognition method, device, equipment and storage medium | |
| CN105303584A (en) | Laser radar-based moving object detection method and device | |
| CN110414585B (en) | Real-time particulate matter detection method based on improved embedded platform | |
| CN111325811B (en) | Lane line data processing method and processing device | |
| CN112001453A (en) | Method and device for calculating accuracy of video event detection algorithm | |
| CN114090560A (en) | Method, device and equipment for generating center line of lane and storage medium | |
| CN116189130A (en) | Lane line segmentation method and device based on image annotation model | |
| CN113239815B (en) | Remote sensing image classification method, device and equipment based on real semantic full-network learning | |
| CN115115967A (en) | Video motion analysis method, device, equipment and medium for model creature | |
| CN111210870B (en) | Protein classification method |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |