CN117218604A - Method and device for supplementing missing passenger flow information, electronic equipment and storage medium - Google Patents
Method and device for supplementing missing passenger flow information, electronic equipment and storage medium Download PDFInfo
- Publication number
- CN117218604A CN117218604A CN202311468336.6A CN202311468336A CN117218604A CN 117218604 A CN117218604 A CN 117218604A CN 202311468336 A CN202311468336 A CN 202311468336A CN 117218604 A CN117218604 A CN 117218604A
- Authority
- CN
- China
- Prior art keywords
- passenger flow
- area
- camera
- historical
- nodes
- 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.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 82
- 230000001502 supplementing effect Effects 0.000 title claims abstract description 56
- 238000003860 storage Methods 0.000 title claims abstract description 14
- 238000012549 training Methods 0.000 claims abstract description 91
- 238000013528 artificial neural network Methods 0.000 claims abstract description 86
- 238000010586 diagram Methods 0.000 claims abstract description 86
- 230000000007 visual effect Effects 0.000 claims abstract description 68
- 238000012545 processing Methods 0.000 claims abstract description 13
- 239000013598 vector Substances 0.000 claims description 27
- 238000004891 communication Methods 0.000 claims description 26
- 239000011159 matrix material Substances 0.000 claims description 20
- 230000008569 process Effects 0.000 claims description 19
- 238000004590 computer program Methods 0.000 claims description 13
- 230000006870 function Effects 0.000 claims description 12
- 238000010606 normalization Methods 0.000 claims description 12
- 238000012544 monitoring process Methods 0.000 description 15
- 230000007246 mechanism Effects 0.000 description 5
- 238000004422 calculation algorithm Methods 0.000 description 4
- 230000000295 complement effect Effects 0.000 description 4
- 238000004220 aggregation Methods 0.000 description 3
- 238000005457 optimization Methods 0.000 description 3
- 239000013589 supplement Substances 0.000 description 3
- 230000002123 temporal effect Effects 0.000 description 3
- 230000009466 transformation Effects 0.000 description 3
- 230000002776 aggregation Effects 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 2
- 238000001514 detection method Methods 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 239000000047 product Substances 0.000 description 2
- 230000008859 change Effects 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000012937 correction Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000009826 distribution Methods 0.000 description 1
- 230000010365 information processing Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000000306 recurrent effect Effects 0.000 description 1
- 230000001172 regenerating effect Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000009469 supplementation Effects 0.000 description 1
Landscapes
- Image Analysis (AREA)
Abstract
The invention provides a method, a device, electronic equipment and a storage medium for supplementing missing passenger flow information, belonging to the field of data processing, wherein the method comprises the following steps: the method comprises the steps of acquiring the passenger flow of a region in a camera visual field range in a target region at a certain moment, inputting the passenger flow into a space-time diagram neural network of which the training of the target region is completed, and complementing the passenger flow of a region outside the camera visual field range at the moment, wherein the training step of the space-time diagram neural network comprises the following steps: obtaining a three-dimensional model of a target area to construct an undirected graph model of the target area; acquiring historical passenger flow characteristics of an area in the visual field range of the camera; determining estimated historical passenger flow characteristics of an area outside the camera view range based on an undirected graph model of the target area and the historical passenger flow characteristics of the area in the camera view range; and training the space-time diagram neural network by taking the historical passenger flow characteristics of the area in the camera view range and the estimated historical passenger flow characteristics of the area outside the camera view range as training samples.
Description
Technical Field
The present invention relates to the field of data processing, and in particular, to a method, an apparatus, an electronic device, and a storage medium for supplementing missing passenger flow information.
Background
In a subway station, cameras are typically arranged to monitor the passenger flow at each region of interest in real time. However, the cameras and their viewing angle of the arrangement may be difficult to cover or cover for a long time all areas inside the subway station, which directly results in some passenger flow information missing in time or space, and the relevant information is therefore not available.
By establishing a space-time diagram neural network to learn the relation between time and space of historical passenger flow data, missing information can be complemented in real time, however, in the related art, a traditional subway passenger flow monitoring or predicting system based on the space-time diagram neural network only focuses on the whole dimension of the whole subway network, such as focusing on the passenger's incoming and outgoing passenger flow data, so that the monitored or predicted passenger flow data cannot accurately reflect to a specific position in a subway station. The related art monitoring means for the passenger flow in the subway station are all based on available data, and if the corresponding data are unavailable for a short time or a long time, the corresponding monitoring function cannot be realized and used, namely the system of the related art cannot complement or predict the missing data.
Disclosure of Invention
The invention provides a method, a device, electronic equipment and a storage medium for supplementing missing passenger flow information, which are used for solving the defect that monitored or predicted passenger flow data in the prior art cannot accurately reflect to a specific position in a subway station and cannot supplement or predict missing data, and realizing real-time acquisition and supplement of passenger flow information at each position in a single subway station.
The invention provides a method for supplementing missing passenger flow information, which comprises the following steps:
acquiring the passenger flow of a region in the visual field range of the camera in the target region at a certain moment;
inputting the passenger flow of the region in the camera view field range in the target region at the certain moment into a space-time diagram neural network of the target region after training is completed, and complementing the passenger flow of the region outside the camera view field range in the target region at the certain moment;
the training step of the space-time diagram neural network comprises the following steps:
acquiring a three-dimensional model of the target area;
constructing an undirected graph model of the target area based on the three-dimensional model of the target area and the area in the camera view field of the target area, wherein the undirected graph model of the target area comprises nodes corresponding to the area in the camera view field, nodes corresponding to the area outside the camera view field and communication relations between the nodes corresponding to the area in the camera view field and the nodes corresponding to the area outside the camera view field;
Acquiring historical passenger flow characteristics of an area in the visual field range of the camera;
determining estimated historical passenger flow characteristics of an area outside the camera visual field range based on the undirected graph model of the target area and the historical passenger flow characteristics of the area in the camera visual field range;
and training the space-time diagram neural network by taking the historical passenger flow characteristics of the area in the camera view range and the estimated historical passenger flow characteristics of the area outside the camera view range as training samples.
According to the method for supplementing missing passenger flow information provided by the invention, the undirected graph model of the target area is constructed based on the three-dimensional model of the target area and the area in the visual field range of the camera in the target area, and the method comprises the following steps:
determining three-dimensional positions of a plurality of nodes in a three-dimensional model of the target area, wherein the nodes comprise areas in the visual field range of all cameras in the target area and at least one area outside the visual field range of the cameras;
establishing effective edges among the nodes in the three-dimensional model of the target area based on the communication relation among the nodes;
And based on the three-dimensional positions of a plurality of nodes in the three-dimensional model of the target area and the effective edges among the plurality of nodes, constructing an undirected graph model of the target area by using an adjacency matrix.
According to the method for supplementing missing passenger flow information provided by the invention, the method for determining the estimated historical passenger flow characteristics of the area outside the camera visual field range based on the undirected graph model of the target area and the historical passenger flow characteristics of the area inside the camera visual field range comprises the following steps:
for historical passenger flow characteristicsCarrying out normalization processing to obtain information of unavailable passenger flow data items, wherein the historical passenger flow characteristics are->The method comprises available passenger flow data items and unavailable passenger flow data items, wherein the available passenger flow data items are historical passenger flow characteristics of areas in the visual field range of the camera, and the unavailable passenger flow data item information is estimated historical passenger flow characteristics of areas outside the visual field range of the camera.
According to the method for supplementing missing passenger flow information provided by the invention, the historical passenger flow characteristics are comparedCarrying out normalization processing to obtain non-available passenger flow data item information, wherein the normalization processing comprises the following steps:
For historical passenger flow characteristics at a certain momentUsing binarized adjacency matrixSum matrix->Calculating normalized adjacency matrix->By multiple use->Update feature vector +.>And reassigning the available passenger flow volume data item part to +.>Up to the feature vector->Reaching a convergence state, wherein the unavailable passenger flow volume data item part is the estimated historical passenger flow volume characteristic of the area outside the visual field range of the camera at the certain moment, and the historical passenger flow volume characteristic is->The historical available passenger flow data items at a certain moment are passenger flow characteristics of areas in the camera visual field range at the certain moment.
According to the method for supplementing missing passenger flow information provided by the invention, the space-time diagram neural network comprises an encoder and a decoder, the historical passenger flow characteristics of the area in the camera view range and the estimated historical passenger flow characteristics of the area outside the camera view range are used as training samples, and the space-time diagram neural network is trained, and the method comprises the following steps:
performing a first process on the training sample loop until a loss function converges;
And taking the space-time diagram neural network after the last first process is completed as a space-time diagram neural network after training is completed.
Wherein the first process comprises:
extracting the characteristics of the historical passenger flow characteristics of a plurality of nodes in the undirected graph model at a certain moment in a sample sequence in the space dimension and the characteristics of the historical passenger flow characteristics of the plurality of nodes in the undirected graph model at the certain moment in the time dimension by using the encoder;
acquiring, by the decoder, a first-time completed passenger flow feature vector of a plurality of nodes at the certain moment based on a time dimension of historical passenger flow features of the plurality of nodes at the certain moment in the undirected graph model;
the decoder obtains the passenger flow characteristic vector of which the plurality of nodes are finally complemented at a certain moment based on the passenger flow characteristic vector of which the plurality of nodes are initially complemented at the certain moment in the undirected graph model;
calculating a loss function based on the passenger flow characteristic vector of the plurality of nodes which are finally complemented at the certain momentWherein->For the final complemented traffic feature vector of the nodes in the undirected graph model at time t +. >For the available traffic characteristics of a plurality of nodes in said undirected graph model at said time t,/>Representing all parameters to be trained in said encoder and said decoder,/for each of said parameters to be trained>A sequence length for the training samples;
the weight matrix and the bias vector in the space-time diagram neural network are adjusted through back propagation.
According to the method for supplementing missing passenger flow information provided by the invention, after taking the historical passenger flow characteristics of the area in the camera view range and the estimated historical passenger flow characteristics of the area outside the camera view range as training samples, the training step of the space-time diagram neural network further comprises the following steps:
and randomly removing historical available passenger flow characteristics of areas in the visual field range of one or more cameras in the training samples, and training the space-time diagram neural network in a Mini-Batch mode based on the removed training sample data.
The invention provides a method for supplementing missing passenger flow information, which further comprises the following steps:
continuously acquiring new passenger flow data of a region in the visual field range of the camera in the target region;
and continuously optimizing the space-time diagram neural network of the target area based on the new passenger flow data of the area in the visual field range of the camera in the target area.
The invention also provides a device for supplementing the missing passenger flow information, which comprises:
the acquisition module is used for acquiring the passenger flow of the area in the camera view field range in the target area at a certain moment;
the completion module is used for inputting the passenger flow of the region in the camera view range in the target region at the certain moment into the space-time diagram neural network of the target region after training is completed, and completing the passenger flow of the region outside the camera view range in the target region at the certain moment;
the training step of the space-time diagram neural network comprises the following steps:
acquiring a three-dimensional model of the target area;
constructing an undirected graph model of the target area based on the three-dimensional model of the target area and the area in the camera view field of the target area, wherein the undirected graph model of the target area comprises nodes corresponding to the area in the camera view field, nodes corresponding to the area outside the camera view field and communication relations between the nodes corresponding to the area in the camera view field and the nodes corresponding to the area outside the camera view field;
acquiring historical passenger flow characteristics of an area in the visual field range of the camera;
Determining estimated historical passenger flow characteristics of an area outside the camera visual field range based on the undirected graph model of the target area and the historical passenger flow characteristics of the area in the camera visual field range;
and training the space-time diagram neural network by taking the historical passenger flow characteristics of the area in the camera view range and the estimated historical passenger flow characteristics of the area outside the camera view range as training samples.
The invention also provides an electronic device comprising a memory, a processor and a computer program stored on the memory and capable of running on the processor, wherein the processor realizes the method for complementing the missing passenger flow information according to any one of the above when executing the program.
The present invention also provides a non-transitory computer readable storage medium having stored thereon a computer program which, when executed by a processor, implements a method of supplementing missing traffic information as described in any of the above.
The invention also provides a computer program product comprising a computer program which when executed by a processor implements a method of supplementing missing traffic information as described in any of the above.
According to the method, the device, the electronic equipment and the storage medium for supplementing the missing passenger flow information, the subway station graph model is built based on the building model of the subway station to establish the communication relation between the nodes corresponding to the camera-free areas and the nodes corresponding to the camera-free areas, the historical passenger flow data of the camera-free areas are obtained based on the historical passenger flow data of the camera-free areas and the communication relation, the time-space diagram neural network is established to learn the relation between the historical passenger flow data and the space, and the missing node characteristics can be supplemented, so that the purpose of supplementing the passenger flow information is achieved, and the accuracy of estimating the passenger flow of the camera-free areas of the subway station is improved.
Drawings
In order to more clearly illustrate the invention or the technical solutions of the prior art, the following description will briefly explain the drawings used in the embodiments or the description of the prior art, and it is obvious that the drawings in the following description are some embodiments of the invention, and other drawings can be obtained according to the drawings without inventive effort for a person skilled in the art.
FIG. 1 is a schematic flow chart of a method for supplementing missing traffic information provided by the present invention;
FIG. 2 is a schematic diagram of determining characteristics of passenger flow in a camera field of view according to the present invention;
FIG. 3 is a schematic diagram of a subway platform layer diagram model provided by the invention;
FIG. 4 is a schematic flow chart of the encoder operation provided by the present invention;
FIG. 5 is a second flow chart of the method for supplementing missing traffic information according to the present invention;
FIG. 6 is a schematic diagram of a device for supplementing missing traffic information according to the present invention;
fig. 7 is a schematic structural diagram of an electronic device provided by the present invention.
Detailed Description
For the purpose of making the objects, technical solutions and advantages of the present invention more apparent, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings, and it is apparent that the described embodiments are some embodiments of the present invention, not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
The following will be described first:
as the passenger flow of urban rail transit is increasingly aggravated, the real-time monitoring function of the passenger flow in the subway station is conducive to early warning of subway operators, and corresponding measures are taken to avoid the situation that the station is overcrowded and serious potential safety hazards are caused. In the related art, most of monitoring of the passenger flow in the subway station is based on cameras, namely, relevant information processing is performed based on monitored contents, and passenger flow information of a non-camera coverage area or a camera blind area is unknown. In addition, most systems or products for subway passenger flow prediction are from a macroscopic point of view, for example, passenger arrival and departure flow of each station is counted and station distribution of the whole line is combined to give future passenger flow prediction conditions of each station.
The conventional subway passenger flow volume monitoring or predicting system based on the space-time diagram neural network only focuses on the whole dimension of the whole subway network, such as the passenger in-and-out passenger flow volume data of passengers, so that the monitored or predicted passenger flow volume data cannot accurately reflect to a specific position in a subway station.
The monitoring means of the passenger flow in the subway station in the related art are all based on available data, such as data from a camera. If the corresponding data is not available for a short time or a long time, the corresponding monitoring function is not available and usable. In other words, the related art system cannot support the complementation or prediction of missing data.
The method for supplementing missing traffic information provided by the present invention is described below with reference to fig. 1 to 5.
Fig. 1 is a schematic flow chart of a method for supplementing missing passenger flow information, as shown in fig. 1, the method includes the following steps:
step 100, obtaining the passenger flow of a region in a camera view field range in a target region at a certain moment;
alternatively, the target area may be any area where prediction of passenger flow is required, such as a subway station, or a railway station, or an airport, etc., which is not limited by the present invention.
Optionally, the target area may include at least one camera, and each camera may capture an image in the field of view to obtain the passenger flow volume in the field of view.
Optionally, the fields of view of the multiple cameras in the target area may overlap or not overlap, and in the case that the fields of view overlap, the actual passenger flow volume in the overlapping area needs to be determined.
Optionally, the field of view of the camera may be a range of a maximum space that can be photographed by the camera, and the area within the field of view of the camera is an area within the maximum space that can be photographed by the camera, and the area outside the field of view of the camera is an area outside the maximum space that can be photographed by the camera.
Optionally, the field of view of cameras of different specifications is different.
Step 110, inputting the passenger flow of the region in the camera view field range in the target region at the certain moment into a space-time diagram neural network of the trained target region, and complementing the passenger flow of the region outside the camera view field range in the target region at the certain moment;
the training step of the space-time diagram neural network comprises the following steps:
acquiring a three-dimensional model of the target area;
Constructing an undirected graph model of the target area based on the three-dimensional model of the target area and the area in the camera view field of the target area, wherein the undirected graph model of the target area comprises nodes corresponding to the area in the camera view field, nodes corresponding to the area outside the camera view field and communication relations between the nodes corresponding to the area in the camera view field and the nodes corresponding to the area outside the camera view field;
acquiring historical passenger flow characteristics of an area in the visual field range of the camera;
determining estimated historical passenger flow characteristics of an area outside the camera visual field range based on the undirected graph model of the target area and the historical passenger flow characteristics of the area in the camera visual field range;
and training the space-time diagram neural network by taking the historical passenger flow characteristics of the area in the camera view range and the estimated historical passenger flow characteristics of the area outside the camera view range as training samples.
Alternatively, the three-dimensional model of the target area may be a three-dimensional building model.
For example, in the case where the target area is an a subway station, a three-dimensional building model of the a subway station may be acquired, and in the case where the target area is a B railway station, a building model of the B railway station may be acquired.
Alternatively, the three-dimensional model of the target area may be derived from a building information model (Building Information Modeling, BIM).
Alternatively, after the three-dimensional model of the target area is acquired, the positions of all cameras currently deployed in the target area in the three-dimensional model may be determined.
Alternatively, the cameras may be deployed in different floors of the target area, such as at a hall floor, a platform floor, a ground floor, or an escalator.
Optionally, for some areas where passenger flow characteristics are required to be obtained, but camera blind areas or areas where no cameras are arranged exist, the positions of the areas in the three-dimensional model are required to be determined, so that the communication relationship between the areas outside the camera view range and the areas outside the camera view range is determined conveniently.
Optionally, an undirected graph model can be constructed based on the three-dimensional model of the target area, the spatial positions of all cameras and the area where the passenger flow characteristics are required to be obtained, but the blind areas of the cameras exist or the cameras are not arranged, so that the communication relation among the areas is determined.
Optionally, historical passenger flow data can be collected through actually deployed cameras, and historical passenger flow characteristics of corresponding positions are generated.
Alternatively, an object detection algorithm that can identify pedestrians, such as YOLO (You Only Look Once) algorithm, or other object detection algorithm, may be used to obtain the number of pedestrians in the image captured by the camera at time t, as a primary passenger flow feature of the area of the camera in the field of view at time t.
Alternatively, an effective visual field range can be extracted from the visual field range of each deployed monitoring camera, and the number of people in the visual field range at the current moment is taken as the passenger flow characteristic.
Fig. 2 is a schematic diagram of determining a passenger flow characteristic in a camera view range, and as shown in fig. 2, an effective circular view range can be determined in the camera view range, and the number of people in the circular view range is taken as the passenger flow characteristic.
Alternatively, for some areas where passenger flow characteristics are to be obtained, but where there is a camera dead zone or where no camera is arranged, the number of people can be set to zero.
Optionally, after the historical passenger flow characteristics of the regions in the camera view range are obtained, the passenger flow characteristics of all the regions in the undirected graph model at the current moment can be expressed as a matrix Which represents the characteristic data of the passenger flow volume of all areas in the undirected graph model at the current moment.
Alternatively, the sequence data length of one training sample may be set toFor the current sequence data, the passenger flow characteristics of one training sample can be recorded as a set: />These sequence samples may be extracted from a historical passenger flow volume dataset.
Alternatively, if data at a certain time in the sample corresponds to a case where the monitoring camera is not available, its corresponding value may be set to zero due to the data loss.
Optionally, in order to initialize some missing data more accurately, so that the neural network can learn effectively from the training process, existing available data needs to be used for complementing unknown unavailable data through spatial relationships, namely, based on an undirected graph model of a target area and historical passenger flow characteristics of an area in the camera view field, determining estimated historical passenger flow characteristics of an area outside the camera view field.
Optionally, the historical passenger flow characteristics of the area in the camera view field range and the estimated historical passenger flow characteristics of the area outside the camera view field range can be used as training samples, the space-time diagram neural network is trained, and the missing node characteristics can be complemented through the trained space-time diagram neural network, so that the passenger flow information complementing purpose is achieved.
Alternatively, continuous optimization of existing space-time diagram neural networks may be performed by continuous acquisition of new passenger flow characteristic data.
According to the method for supplementing the missing passenger flow information, the subway station graph model is built based on the building model of the subway station to establish the communication relation between the nodes corresponding to the camera-free areas and the nodes corresponding to the camera-free areas, the historical passenger flow data of the camera-free areas are obtained based on the historical passenger flow data of the camera-free areas and the communication relation, the time-space graph neural network is built to learn the relation between the historical passenger flow data and the space, and missing node characteristics can be supplemented, so that the purpose of supplementing the passenger flow information is achieved, and the accuracy of estimating the passenger flow of the camera-free areas of the subway station is improved.
Optionally, the constructing an undirected graph model of the target area based on the three-dimensional model of the target area and an area in a camera view range in the target area includes:
determining three-dimensional positions of a plurality of nodes in a three-dimensional model of the target area, wherein the nodes comprise areas in the visual field range of all cameras in the target area and at least one area outside the visual field range of the cameras;
Establishing effective edges among the nodes in the three-dimensional model of the target area based on the communication relation among the nodes;
and based on the three-dimensional positions of a plurality of nodes in the three-dimensional model of the target area and the effective edges among the plurality of nodes, constructing an undirected graph model of the target area by using an adjacency matrix.
Alternatively, each monitoring camera monitoring position and each region to be complemented with passenger flow volume can be used as one node in the undirected graph model, and the collection of all nodes is recorded asThe total number of nodes is marked +.>。
Optionally, after determining the three-dimensional positions of all the nodes, connectivity among the nodes needs to be determined according to an actual building model, and the communication relation among the nodes needs to be checked and corrected by combining the actual passenger flow movement direction, so that effective edges among the nodes are established.
Alternatively, if there is a direct occlusion of the building wall between the links of two nodes, no active edge is established.
Alternatively, if there is a manual traffic guiding measure, such as a guardrail, between the links of the two nodes, no active edge is established.
Alternatively, the nodes across floors may be connected by constructing the active edges with the nodes at the escalator or stairway.
Alternatively, the set of all valid edges can be written as。
Alternatively, the calculation may be performed for the construction based on the actual spatial distance and connectivity between the respective monitoring camerasAdjacency matrix for building undirected graph model。
Optionally, the adjacency matrixBased on the European spatial distance between nodes +.>Calculations are performed but require additional correction using gaussian kernels: />。
Alternatively, the finalized undirected graph model can be written as。
Fig. 3 is a schematic diagram of a subway platform layer graph model provided by the present invention, as shown in fig. 3, in an embodiment of the present invention, the constructed subway platform layer graph model is shown as a graph, which includes 14 nodes and 14 effective edges.
According to the method for supplementing the missing passenger flow information, provided by the invention, the connection relation among all areas is accurately established by accurately spatially modeling the passenger flow data in the subway station, so that the information of the areas with missing passenger flow data can be conveniently supplemented based on the connection relation.
Optionally, the determining the estimated historical passenger flow characteristic of the area outside the camera field of view based on the undirected graph model of the target area and the historical passenger flow characteristic of the area within the camera field of view includes:
For historical passenger flow characteristicsCarrying out normalization processing to obtain information of unavailable passenger flow data items, wherein the historical passenger flow characteristics are->The method comprises available passenger flow data items and unavailable passenger flow data items, wherein the available passenger flow data items are historical passenger flow characteristics of areas in the visual field range of the camera, and the unavailable passenger flow data item information is estimated historical passenger flow characteristics of areas outside the visual field range of the camera.
Optionally, in order to initialize some missing data more accurately, so that the neural network can learn effectively from the training process, existing available data needs to be used for complementing unknown unavailable data through spatial relationships, namely, based on an undirected graph model of a target area and historical passenger flow characteristics of an area in the camera view field, determining estimated historical passenger flow characteristics of an area outside the camera view field.
Alternatively, the unknown unavailable data may be passenger flow data of an area outside the field of view of the camera, or data corresponding to when the camera is unavailable at a certain moment, and may be data in a time dimension, or data in a space dimension.
Alternatively, to determine the estimated historical traffic characteristics of the area outside the camera field of view, the historical traffic characteristics may be determinedAnd carrying out normalization processing to obtain the information of the unavailable passenger flow data item.
For example, the characteristics of the passenger flow volume at the time of the history A can beAnd carrying out normalization processing to obtain the unavailable passenger flow data item information at the time A.
The method for supplementing the missing passenger flow information provided by the invention is characterized by historical passenger flowAnd carrying out normalization processing to obtain information of unavailable passenger flow data items so as to obtain training samples and train the space-time diagram neural network.
Optionally, the calendarHistory passenger flow characteristicsCarrying out normalization processing to obtain non-available passenger flow data item information, wherein the normalization processing comprises the following steps:
for historical passenger flow characteristics at a certain momentAdjacent matrix using binarization +.>Sum matrix->Calculating normalized adjacency matrix->By multiple use->Updating feature vectorsAnd reassign the available traffic data portion to +.>Up to the feature vector->Reaching a convergence state, wherein the unavailable passenger flow volume data part is the estimated historical passenger flow volume characteristic of the area outside the visual field range of the camera at the certain moment, and the historical passenger flow volume characteristic is- >The historical available passenger flow data items at a certain moment are passenger flow characteristics of areas in the camera visual field range at the certain moment.
Optionally, in order to initialize some missing data more accurately, so that the neural network can learn effectively from the training process, existing available data needs to be used for complementing unknown unavailable data through spatial relationships, namely, based on an undirected graph model of a target area and historical passenger flow characteristics of an area in the camera view field, determining estimated historical passenger flow characteristics of an area outside the camera view field.
Alternatively, the unknown unavailable data may be passenger flow data of an area outside the field of view of the camera, or data corresponding to when the camera is unavailable at a certain moment, and may be data in a time dimension, or data in a space dimension.
Alternatively, each data of the sequence of training samples may be selectedAccording to the established undirected graph model, a binarized adjacency matrix +.>Degree matrix->Computing normalized adjacency matrix。
Alternatively, for some unavailable data, such as passenger flow data of areas outside the field of view of the camera, the data can be continuously used Update feature vector +.>And reassign the available traffic data portion to +.>Up to the feature vector->Reaching the convergence state, can be identifiedAnd estimating the passenger flow data at the current moment.
The method for supplementing the missing passenger flow information provided by the invention is based on the normalized adjacency matrix, and uses the existing available data to supplement the unknown unavailable data through the spatial relationship.
Optionally, the space-time diagram neural network includes an encoder and a decoder, and the training of the space-time diagram neural network by using the historical passenger flow characteristics of the area within the camera view range and the estimated historical passenger flow characteristics of the area outside the camera view range as training samples includes:
performing a first process on the training sample loop until a loss function converges;
taking the space-time diagram neural network after the last first process is completed as a space-time diagram neural network after training is completed; wherein the first process comprises:
extracting the characteristics of the historical passenger flow characteristics of a plurality of nodes in the undirected graph model at a certain moment in a sample sequence in the space dimension and the characteristics of the historical passenger flow characteristics of the plurality of nodes in the undirected graph model at the certain moment in the time dimension by using the encoder;
Acquiring, by the decoder, a first-time completed passenger flow feature vector of a plurality of nodes at the certain moment based on a time dimension of historical passenger flow features of the plurality of nodes at the certain moment in the undirected graph model;
the decoder obtains the passenger flow characteristic vector of which the plurality of nodes are finally complemented at a certain moment based on the passenger flow characteristic vector of which the plurality of nodes are initially complemented at the certain moment in the undirected graph model;
calculating a loss function based on the passenger flow characteristic vector of the plurality of nodes which are finally complemented at the certain momentWherein->Is saidFinal complemented traffic feature vector of multiple nodes at time t in undirected graph model, +.>For the available traffic characteristics of a plurality of nodes in said undirected graph model at said time t,/>Representing all parameters to be trained in said encoder and said decoder,a sequence length for the training samples;
the weight matrix and the bias vector in the space-time diagram neural network are adjusted through back propagation.
Optionally, in the encoding process, in order to extract the features of all data in the spatial dimension, any node at each moment can be calculated by using a graph self-attention mechanism when feature aggregation of nodes is performed Self-attention parameter between neighboring nodes>Which represents node->For node->Is a function of (a) and (b).
Alternatively, the process may be carried out in a single-stage,is required to train a shared weight matrix of a graph network>Feedforward neural network for self-attention mechanism (noted +.>) Execution of->And normalization using a softmax function: /> 。
Alternatively, the node may be connected by a nodeAll neighbor nodes->The characteristics are weighted and summed to finally obtain the characteristics of the node after aggregation>。
Optionally, during the encoding process, in order to extract the features of all data in the time dimension, the GRU units are used in combination with a graph self-attention mechanism, at each instantAggregation feature for each node in the graph model>Execute the graph convolution operation +.>Instead of the traditional matrix multiplication in the gating loop unit (Gate Recurrent Unit, GRU):
；
；
；
；
wherein, 、/>and->In order to use the parameters of the neural network to be trained when the GRU units are used, the optimal values of the parameters need to be obtained through the training process of the neural network.
FIG. 4 is a schematic diagram of the operation of the encoder according to the present invention, as shown in FIG. 4, where the encoder needs to initialize the unknown unavailable data in the training samples, and extract the features of all the data in the spatial dimension and the features in the temporal dimension using the graph annotation mechanism and the gating loop.
Optionally, after the encoder encodes the features of the historical traffic features of the plurality of nodes in the spatial dimension and the temporal features at each time instant, the decoder needs to decode the features of the encoded historical traffic features in the spatial dimension and the temporal features.
Alternatively, to decode the characteristics of the encoded historical traffic characteristics in the time dimension, the decoder may calculate the hidden state of all nodes through the GRU at the previous timeMerging into vector->And calculating the total output characteristics of all nodes at the current moment by using a linear transformation layer>Wherein->Is of linear variationWeight matrix in layer change, +.>The bias vector is the parameter to be trained.
Alternatively, can be based onCalculating an initially complemented data feature vector +.>. Wherein (1)>Is a mask at the current time, which represents +.>Which nodes of the data are not available, i.e. +.>Otherwise->，/>Representing multiplication between elements->Is->Is the inverse of the matrix, i.e.)>When a certain element takes 0, the element is +.>1->When a certain element takes 1, the element is +.>Taking 0.
Alternatively, to decode the features of the encoded historical traffic features in the spatial dimension, the decoder may use the self-attention mechanism of the graph to vector the initially completed data features In combination with the spatial structure of the undirected graph model, a feedforward neural network is used again to calculate the self-attention parameter +.>By combining the final aggregate characteristics of each node, the aggregate characteristic representation of each node at decoding time can be obtained>。
Alternatively, the decoder may represent the aggregate characteristics of all nodesHidden state with encoder outputCombining and calculating the total output characteristics of all nodes at the current moment again by using the linear transformation layerWherein->For the weight matrix in the linear transformation layer, +.>The bias vector is the parameter to be trained.
Alternatively, the decoder may be based on the total output characteristics of all nodes at the current timeCalculating the data feature vector which is finally complemented +.>。
Alternatively, during the training process, a plurality of sequences may first be extracted from the historical passenger flow characteristic dataset as the training dataset.
Alternatively, each sequence data sample should have a set lengthThe sequence data samples can be denoted as set +.>。/>
Alternatively, the loss function may be defined as the reconstruction loss of all data, i.e. not only focusing on missing data, but also on the prediction accuracy of existing data, i.e. Wherein->For the passenger flow characteristics of the nodes in the undirected graph model which are complemented after model reasoning is carried out at the moment t, the passenger flow characteristics are +.>For the initially available traffic characteristics of a plurality of nodes in the undirected graph model at that moment>Representing all parameters to be trained in said encoder and said decoder,/for each of said parameters to be trained>Is the sequence length of the training samples.
Alternatively, the back propagation may be achieved by optimization algorithms such as the chain law or gradient descent.
Alternatively, after training of the space-time diagram neural network is completed, the space-time diagram neural network may be used for real-time information completion.
According to the method for supplementing the missing passenger flow information, the time-space diagram neural network is established to learn the relation between the historical passenger flow data and the space, and the trained time-space diagram neural network can be used for supplementing the missing node characteristics, so that the passenger flow information is supplemented.
Optionally, after the training samples include the historical passenger flow characteristics of the area within the camera view range and the estimated historical passenger flow characteristics of the area outside the camera view range, the training step of the space-time diagram neural network further includes:
And randomly removing historical available passenger flow characteristics of areas in the visual field range of one or more cameras in the training samples, and training the space-time diagram neural network in a Mini-Batch mode based on the removed training sample data.
Optionally, in order for the space-time neural network to have more powerful generalization and information replenishment capabilities, in addition to some nodes that do not themselves contain available data, it is also necessary to consider randomly passing some available data through a mask at the current time during the training processErasing is performed, and training is performed by using a Mini-Batch mode.
Optionally, the process of randomly selecting data and training is as follows:
for i in{1,2,…execution of
Randomly selecting a time t and obtaining corresponding sequence data
for j in{1，2，…Execution of
Randomly generating a missProportion r（0，1）
Randomly generating a series of masks based on the r values
Regenerating sequence data from a mask
end
Training using current Mini-Batch sequence data
end
The method for complementing the missing passenger flow information provided by the invention randomly passes some available data through the mask at the current momentThe erasing is carried out, and training is carried out by using a Mini-Batch mode, so that the space-time diagram neural network has stronger generalization and information supplementation capability.
Optionally, the method further comprises:
continuously acquiring new passenger flow data of a region in the visual field range of the camera in the target region;
and continuously optimizing the space-time diagram neural network of the target area based on the new passenger flow data of the area in the visual field range of the camera in the target area.
Optionally, after training of the space-time diagram neural network is completed, passenger flow data of the area within the camera field of view in the target area can be obtained continuously.
Alternatively, training can be continued on the trained space-time diagram neural network based on the continuously obtained new data, and model performance of the space-time diagram neural network can be continuously optimized.
Fig. 5 is a second flow chart of the method for supplementing missing passenger flow information provided by the present invention, as shown in fig. 5, in one embodiment of the present invention, the method for supplementing missing passenger flow information of a subway station includes the following steps:
1. based on the building information model BIM of the subway station, all monitoring cameras in the subway station and all areas which are not provided with cameras or are in the blind areas of the cameras and need to estimate the passenger flow, the modeling process of the passenger flow undirected graph model in the subway station is completed.
2. Historical passenger flow data are collected through actually deployed monitoring cameras, the number of people is selected as passenger flow characteristics, and passenger flow characteristics of corresponding positions are generated. For the area where the camera is not arranged or is in the blind area of the camera, the corresponding passenger flow characteristic needs to be temporarily set to be zero.
3. Training the constructed space-time diagram neural network, wherein the trained space-time diagram neural network can be used for supplementing the missing passenger flow information, and in addition, continuous training optimization of the existing space-time diagram neural network is executed through continuous acquisition of new passenger flow characteristic data.
4. The trained space-time diagram neural network can be used for real-time complementation of the missing passenger flow information.
The method for supplementing the missing passenger flow information can input the historical passenger flow data of the target area into the space-time diagram neural network so as to obtain the passenger flow information which is output by the space-time diagram neural network and is supplemented by the target area.
The device for supplementing the missing passenger flow information provided by the invention is described below, and the device for supplementing the missing passenger flow information described below and the method for supplementing the missing passenger flow information described above can be correspondingly referred to each other.
Fig. 6 is a schematic structural diagram of a device 600 for supplementing missing passenger flow information provided by the present invention, and as shown in fig. 6, the device 600 for supplementing missing passenger flow information provided by the present invention includes an obtaining module 610 and a supplementing module 620, where:
the acquiring module 610 is configured to acquire a passenger flow volume of an area in a camera view range in a target area at a certain moment;
A complement module 620, configured to input, to the space-time diagram neural network of the target area where training is completed, the passenger flow volume of the area in the camera view field in the target area at the certain moment, and complement the passenger flow volume of the area outside the camera view field in the target area at the certain moment;
the training step of the space-time diagram neural network comprises the following steps:
acquiring a three-dimensional model of the target area;
constructing an undirected graph model of the target area based on the three-dimensional model of the target area and the area in the camera view field of the target area, wherein the undirected graph model of the target area comprises nodes corresponding to the area in the camera view field, nodes corresponding to the area outside the camera view field and communication relations between the nodes corresponding to the area in the camera view field and the nodes corresponding to the area outside the camera view field;
acquiring historical passenger flow characteristics of an area in the visual field range of the camera;
determining estimated historical passenger flow characteristics of an area outside the camera visual field range based on the undirected graph model of the target area and the historical passenger flow characteristics of the area in the camera visual field range;
And training the space-time diagram neural network by taking the historical passenger flow characteristics of the area in the camera view range and the estimated historical passenger flow characteristics of the area outside the camera view range as training samples.
According to the device for supplementing missing passenger flow information, the subway station graph model is built based on the building model of the subway station to establish the communication relation between the nodes corresponding to the camera-free areas and the nodes corresponding to the camera-free areas, the historical passenger flow data of the camera-free areas are obtained based on the historical passenger flow data of the camera-free areas and the communication relation, the time-space graph neural network is built to learn the relation between the historical passenger flow data and the space, missing node characteristics can be supplemented, the purpose of supplementing the passenger flow information is achieved, and accuracy of estimating the passenger flow of the camera-free areas of the subway station is improved.
It can be understood that the device for supplementing the missing passenger flow information provided by the present invention corresponds to the method for supplementing the missing passenger flow information provided by the above embodiments, and the relevant technical features of the device for supplementing the missing passenger flow information provided by the present invention may refer to the relevant technical features of the method for supplementing the missing passenger flow information provided by the above embodiments, which are not described herein again.
Fig. 7 illustrates a physical schematic diagram of an electronic device, as shown in fig. 7, which may include: processor 710, communication interface (Communications Interface) 720, memory 730, and communication bus 740, wherein processor 710, communication interface 720, memory 730 communicate with each other via communication bus 740. Processor 710 may invoke logic instructions in memory 730 to perform a complement method of missing traffic information, the method comprising: acquiring the passenger flow of a region in the visual field range of the camera in the target region at a certain moment; inputting the passenger flow of the region in the camera view field range in the target region at the certain moment into a space-time diagram neural network of the target region after training is completed, and complementing the passenger flow of the region outside the camera view field range in the target region at the certain moment; the training step of the space-time diagram neural network comprises the following steps: acquiring a three-dimensional model of the target area; constructing an undirected graph model of the target area based on the three-dimensional model of the target area and the area in the camera view field of the target area, wherein the undirected graph model of the target area comprises nodes corresponding to the area in the camera view field, nodes corresponding to the area outside the camera view field and communication relations between the nodes corresponding to the area in the camera view field and the nodes corresponding to the area outside the camera view field; acquiring historical passenger flow characteristics of an area in the visual field range of the camera; determining estimated historical passenger flow characteristics of an area outside the camera visual field range based on the undirected graph model of the target area and the historical passenger flow characteristics of the area in the camera visual field range; and training the space-time diagram neural network by taking the historical passenger flow characteristics of the area in the camera view range and the estimated historical passenger flow characteristics of the area outside the camera view range as training samples.
Further, the logic instructions in the memory 730 described above may be implemented in the form of software functional units and may be stored in a computer readable storage medium when sold or used as a stand alone product. Based on this understanding, the technical solution of the present invention may be embodied essentially or in a part contributing to the prior art or in a part of the technical solution, in the form of a software product stored in a storage medium, comprising several instructions for causing a computer device (which may be a personal computer, a server, a network device, etc.) to perform all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a random access Memory (RAM, random Access Memory), a magnetic disk, or an optical disk, or other various media capable of storing program codes.
In another aspect, the present invention also provides a computer program product, the computer program product comprising a computer program, the computer program being storable on a non-transitory computer readable storage medium, the computer program, when executed by a processor, being capable of performing a method of supplementing missing traffic information provided by the methods described above, the method comprising: acquiring the passenger flow of a region in the visual field range of the camera in the target region at a certain moment; inputting the passenger flow of the region in the camera view field range in the target region at the certain moment into a space-time diagram neural network of the target region after training is completed, and complementing the passenger flow of the region outside the camera view field range in the target region at the certain moment; the training step of the space-time diagram neural network comprises the following steps: acquiring a three-dimensional model of the target area; constructing an undirected graph model of the target area based on the three-dimensional model of the target area and the area in the camera view field of the target area, wherein the undirected graph model of the target area comprises nodes corresponding to the area in the camera view field, nodes corresponding to the area outside the camera view field and communication relations between the nodes corresponding to the area in the camera view field and the nodes corresponding to the area outside the camera view field; acquiring historical passenger flow characteristics of an area in the visual field range of the camera; determining estimated historical passenger flow characteristics of an area outside the camera visual field range based on the undirected graph model of the target area and the historical passenger flow characteristics of the area in the camera visual field range; and training the space-time diagram neural network by taking the historical passenger flow characteristics of the area in the camera view range and the estimated historical passenger flow characteristics of the area outside the camera view range as training samples.
In yet another aspect, the present invention also provides a non-transitory computer readable storage medium having stored thereon a computer program which, when executed by a processor, is implemented to perform a method of supplementing missing traffic information provided by the above methods, the method comprising: acquiring the passenger flow of a region in the visual field range of the camera in the target region at a certain moment; inputting the passenger flow of the region in the camera view field range in the target region at the certain moment into a space-time diagram neural network of the target region after training is completed, and complementing the passenger flow of the region outside the camera view field range in the target region at the certain moment; the training step of the space-time diagram neural network comprises the following steps: acquiring a three-dimensional model of the target area; constructing an undirected graph model of the target area based on the three-dimensional model of the target area and the area in the camera view field of the target area, wherein the undirected graph model of the target area comprises nodes corresponding to the area in the camera view field, nodes corresponding to the area outside the camera view field and communication relations between the nodes corresponding to the area in the camera view field and the nodes corresponding to the area outside the camera view field; acquiring historical passenger flow characteristics of an area in the visual field range of the camera; determining estimated historical passenger flow characteristics of an area outside the camera visual field range based on the undirected graph model of the target area and the historical passenger flow characteristics of the area in the camera visual field range; and training the space-time diagram neural network by taking the historical passenger flow characteristics of the area in the camera view range and the estimated historical passenger flow characteristics of the area outside the camera view range as training samples.
The apparatus embodiments described above are merely illustrative, wherein the elements illustrated as separate elements may or may not be physically separate, and the elements shown as elements may or may not be physical elements, may be located in one place, or may be distributed over a plurality of network elements. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art will understand and implement the present invention without undue burden.
From the above description of the embodiments, it will be apparent to those skilled in the art that the embodiments may be implemented by means of software plus necessary general hardware platforms, or of course may be implemented by means of hardware. Based on this understanding, the foregoing technical solution may be embodied essentially or in a part contributing to the prior art in the form of a software product, which may be stored in a computer readable storage medium, such as ROM/RAM, a magnetic disk, an optical disk, etc., including several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute the method described in the respective embodiments or some parts of the embodiments.
Finally, it should be noted that: the above embodiments are only for illustrating the technical solution of the present invention, and are not limiting; 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 technical features thereof can be replaced by equivalents; such modifications and substitutions do not depart from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims (10)
1. A method for supplementing missing traffic information, comprising:
acquiring the passenger flow of a region in the visual field range of the camera in the target region at a certain moment;
inputting the passenger flow of the region in the camera view field range in the target region at the certain moment into a space-time diagram neural network of the target region after training is completed, and complementing the passenger flow of the region outside the camera view field range in the target region at the certain moment;
the training step of the space-time diagram neural network comprises the following steps:
acquiring a three-dimensional model of the target area;
constructing an undirected graph model of the target area based on the three-dimensional model of the target area and the area in the camera view field of the target area, wherein the undirected graph model of the target area comprises nodes corresponding to the area in the camera view field, nodes corresponding to the area outside the camera view field and communication relations between the nodes corresponding to the area in the camera view field and the nodes corresponding to the area outside the camera view field;
Acquiring historical passenger flow characteristics of an area in the visual field range of the camera;
determining estimated historical passenger flow characteristics of an area outside the camera visual field range based on the undirected graph model of the target area and the historical passenger flow characteristics of the area in the camera visual field range;
and training the space-time diagram neural network by taking the historical passenger flow characteristics of the area in the camera view range and the estimated historical passenger flow characteristics of the area outside the camera view range as training samples.
2. The method for supplementing missing passenger flow information according to claim 1, wherein the constructing an undirected graph model of the target area based on the three-dimensional model of the target area and an area within a camera field of view in the target area comprises:
determining three-dimensional positions of a plurality of nodes in a three-dimensional model of the target area, wherein the nodes comprise areas in the visual field range of all cameras in the target area and at least one area outside the visual field range of the cameras;
establishing effective edges among the nodes in the three-dimensional model of the target area based on the communication relation among the nodes;
And based on the three-dimensional positions of a plurality of nodes in the three-dimensional model of the target area and the effective edges among the plurality of nodes, constructing an undirected graph model of the target area by using an adjacency matrix.
3. The method of supplementing missing traffic information according to claim 1, wherein the determining the estimated historical traffic characteristics of the area outside the camera field of view based on the undirected graph model of the target area and the historical traffic characteristics of the area within the camera field of view comprises:
for historical passenger flow characteristicsCarrying out normalization processing to obtain information of unavailable passenger flow data items, wherein the historical passenger flow characteristics are->The method comprises available passenger flow data items and unavailable passenger flow data items, wherein the available passenger flow data items are historical passenger flow characteristics of areas in the visual field range of the camera, and the unavailable passenger flow data item information is estimated historical passenger flow characteristics of areas outside the visual field range of the camera.
4. A method of supplementing missing traffic information according to claim 3, wherein said pair of historical traffic characteristicsNormalization processing is carried out to obtain unavailable passenger flow volume Data item information, comprising:
for historical passenger flow characteristics at a certain momentAdjacent matrix using binarization +.>Sum matrix->Calculating normalized adjacency matrix->By multiple use->Updating feature vectorsAnd reassigning the available passenger flow volume data item part to +.>Up to the feature vector->Reaching a convergence state, wherein the unavailable passenger flow volume data item part is the estimated historical passenger flow volume characteristic of the area outside the visual field range of the camera at the certain moment, and the historical passenger flow volume characteristic is->The historical available passenger flow data items at a certain moment are passenger flow characteristics of areas in the camera visual field range at the certain moment.
5. The method of supplementing missing traffic information according to claim 4, wherein the space-time diagram neural network includes an encoder and a decoder, wherein training the space-time diagram neural network using historical traffic characteristics of regions within the camera field of view and estimated historical traffic characteristics of regions outside the camera field of view as training samples includes:
Performing a first process on the training sample loop until a loss function converges;
taking the space-time diagram neural network after the last first process is completed as a space-time diagram neural network after training is completed;
wherein the first process comprises:
extracting the characteristics of the historical passenger flow characteristics of a plurality of nodes in the undirected graph model at a certain moment in a sample sequence in the space dimension and the characteristics of the historical passenger flow characteristics of the plurality of nodes in the undirected graph model at the certain moment in the time dimension by using the encoder;
acquiring, by the decoder, a first-time completed passenger flow feature vector of a plurality of nodes at the certain moment based on a time dimension of historical passenger flow features of the plurality of nodes at the certain moment in the undirected graph model;
the decoder obtains the passenger flow characteristic vector of which the plurality of nodes are finally complemented at a certain moment based on the passenger flow characteristic vector of which the plurality of nodes are initially complemented at the certain moment in the undirected graph model;
calculating a loss function based on the passenger flow characteristic vector of the plurality of nodes which are finally complemented at the certain momentWherein- >For the final complemented traffic feature vector of the nodes in the undirected graph model at time t +.>For the available traffic characteristics of a plurality of nodes in said undirected graph model at said time t,/>Representing all parameters to be trained in said encoder and said decoder,/for each of said parameters to be trained>A sequence length for the training samples;
the weight matrix and the bias vector in the space-time diagram neural network are adjusted through back propagation.
6. The method for supplementing missing traffic information according to claim 5, wherein after taking the historical traffic characteristics of the area within the camera field of view and the estimated historical traffic characteristics of the area outside the camera field of view as training samples, the training step of the space-time diagram neural network further comprises:
and randomly removing historical available passenger flow characteristics of areas in the visual field range of one or more cameras in the training samples, and training the space-time diagram neural network in a Mini-Batch mode based on the removed training sample data.
7. The method of supplementing missing passenger flow information of claim 1, wherein the method further comprises:
Continuously acquiring new passenger flow data of a region in the visual field range of the camera in the target region;
and continuously optimizing the space-time diagram neural network of the target area based on the new passenger flow data of the area in the visual field range of the camera in the target area.
8. A completion apparatus for deleting traffic information, the apparatus comprising:
the acquisition module is used for acquiring the passenger flow of the area in the camera view field range in the target area at a certain moment;
the completion module is used for inputting the passenger flow of the region in the camera view range in the target region at the certain moment into the space-time diagram neural network of the target region after training is completed, and completing the passenger flow of the region outside the camera view range in the target region at the certain moment;
the training step of the space-time diagram neural network comprises the following steps:
acquiring a three-dimensional model of the target area;
constructing an undirected graph model of the target area based on the three-dimensional model of the target area and the area in the camera view field of the target area, wherein the undirected graph model of the target area comprises nodes corresponding to the area in the camera view field, nodes corresponding to the area outside the camera view field and communication relations between the nodes corresponding to the area in the camera view field and the nodes corresponding to the area outside the camera view field;
Acquiring historical passenger flow characteristics of an area in the visual field range of the camera;
determining estimated historical passenger flow characteristics of an area outside the camera visual field range based on the undirected graph model of the target area and the historical passenger flow characteristics of the area in the camera visual field range;
and training the space-time diagram neural network by taking the historical passenger flow characteristics of the area in the camera view range and the estimated historical passenger flow characteristics of the area outside the camera view range as training samples.
9. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor implements a method of supplementing missing traffic information according to any of claims 1 to 7 when the program is executed by the processor.
10. A non-transitory computer readable storage medium having stored thereon a computer program, wherein the computer program when executed by a processor implements a method of supplementing missing traffic information according to any of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311468336.6A CN117218604B (en) | 2023-11-07 | 2023-11-07 | Method and device for supplementing missing passenger flow information, electronic equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311468336.6A CN117218604B (en) | 2023-11-07 | 2023-11-07 | Method and device for supplementing missing passenger flow information, electronic equipment and storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN117218604A true CN117218604A (en) | 2023-12-12 |
| CN117218604B CN117218604B (en) | 2024-04-12 |
Family
ID=89039291
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311468336.6A Active CN117218604B (en) | 2023-11-07 | 2023-11-07 | Method and device for supplementing missing passenger flow information, electronic equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117218604B (en) |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112183862A (en) * | 2020-09-29 | 2021-01-05 | 长春理工大学 | Traffic flow prediction method and system for urban road network |
| CN114092866A (en) * | 2020-07-30 | 2022-02-25 | 上海虹桥枢纽交通中心建设发展有限公司 | Method for predicting space-time distribution of airport passenger flow |
| CN115600022A (en) * | 2022-10-17 | 2023-01-13 | 京东城市（北京）数字科技有限公司（Cn） | Training and processing method, device and medium of spatio-temporal data processing model |
| US20230021868A1 (en) * | 2021-07-23 | 2023-01-26 | Ancestry.Com Operations Inc. | Data-sharding for efficient record search |
| CN115936762A (en) * | 2022-11-21 | 2023-04-07 | 中国联合网络通信集团有限公司 | Scenic spot pedestrian flow prediction method, device, equipment and storage medium |
| CN116050554A (en) * | 2021-10-28 | 2023-05-02 | 中移(苏州)软件技术有限公司 | Scenic spot passenger flow volume prediction method and device, computing equipment and storage medium |
-
2023
- 2023-11-07 CN CN202311468336.6A patent/CN117218604B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114092866A (en) * | 2020-07-30 | 2022-02-25 | 上海虹桥枢纽交通中心建设发展有限公司 | Method for predicting space-time distribution of airport passenger flow |
| CN112183862A (en) * | 2020-09-29 | 2021-01-05 | 长春理工大学 | Traffic flow prediction method and system for urban road network |
| US20230021868A1 (en) * | 2021-07-23 | 2023-01-26 | Ancestry.Com Operations Inc. | Data-sharding for efficient record search |
| CN116050554A (en) * | 2021-10-28 | 2023-05-02 | 中移(苏州)软件技术有限公司 | Scenic spot passenger flow volume prediction method and device, computing equipment and storage medium |
| CN115600022A (en) * | 2022-10-17 | 2023-01-13 | 京东城市（北京）数字科技有限公司（Cn） | Training and processing method, device and medium of spatio-temporal data processing model |
| CN115936762A (en) * | 2022-11-21 | 2023-04-07 | 中国联合网络通信集团有限公司 | Scenic spot pedestrian flow prediction method, device, equipment and storage medium |
Non-Patent Citations (2)
| Title |
|---|
| 殷礼胜;李胜;唐圣期;何怡刚;: "基于改进的路网聚类结合PSOWNN的交通流预测", 微电子学与计算机, no. 01 * |
| 秦继伟;吐尔根・依布拉音;张辉国;: "一种基于高斯无向图的e-Learner情感调节模型", ***仿真学报, no. 05 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN117218604B (en) | 2024-04-12 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Cho et al. | A neural-based crowd estimation by hybrid global learning algorithm | |
| Wang et al. | Semi-supervised semantic segmentation network for surface crack detection | |
| Forkan et al. | CorrDetector: A framework for structural corrosion detection from drone images using ensemble deep learning | |
| CN105844234B (en) | Method and equipment for counting people based on head and shoulder detection | |
| CA3080050A1 (en) | Training tree-based machine-learning modeling algorithms for predicting outputs and generating explanatory data | |
| CN107220603A (en) | Vehicle checking method and device based on deep learning | |
| CN113298789A (en) | Insulator defect detection method and system, electronic device and readable storage medium | |
| CN110826429A (en) | Scenic spot video-based method and system for automatically monitoring travel emergency | |
| Guo et al. | Evaluation-oriented façade defects detection using rule-based deep learning method | |
| CN112784954A (en) | Method and device for determining neural network | |
| CN113781519A (en) | Target tracking method and target tracking device | |
| Chang et al. | Active attentional sampling for speed-up of background subtraction | |
| CN106612385A (en) | Video detection method and video detection device | |
| Połap | Neuro-heuristic analysis of surveillance video in a centralized IoT system | |
| Martani et al. | Pedestrian monitoring techniques for crowd-flow prediction | |
| Kong et al. | Fdflownet: Fast optical flow estimation using a deep lightweight network | |
| Zhao et al. | Exploiting spatial-temporal correlations for video anomaly detection | |
| Wang et al. | Multitype damage detection of container using CNN based on transfer learning | |
| CN117218604B (en) | Method and device for supplementing missing passenger flow information, electronic equipment and storage medium | |
| CN111950518B (en) | Video image enhancement method for violent behavior recognition | |
| Monaco et al. | Improving wildfire severity classification of deep learning U-nets from satellite images | |
| KR102563346B1 (en) | System for monitoring of structural and method ithereof | |
| CN106682581B (en) | Luggage identification method and equipment | |
| CN114781696A (en) | City road network-oriented model-free accident influence range prediction method | |
| Palit et al. | A Multi-Graph Convolutional Neural Network Model for Short-Term Prediction of Turning Movements at Signalized Intersections |
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 |