Passenger flow statistical method and device, electronic equipment and storage medium

Abstract

This invention provides a passenger flow statistics method, comprising: acquiring monitoring data of a target vehicle within a specified time period, and extracting passenger boarding and alighting behavior information and passenger information from the monitoring data. The passenger boarding and alighting behavior information includes passenger boarding behavior and passenger alighting behavior, with each passenger boarding and alighting behavior information corresponding to one passenger information entry; and counting the number of passengers in the target vehicle based on the passenger boarding and alighting behavior information and passenger information at each station to obtain the passenger flow statistics result for the target vehicle. This method can extract passenger boarding and alighting behavior information and passenger information from the monitoring data, and counts passenger flow in the target vehicle based solely on these information, eliminating errors in counting drivers and passengers. This improves the accuracy of passenger count statistics for public transportation vehicles, thereby improving the accuracy of passenger flow statistics and consequently improving the accuracy of subsequent passenger flow and origin-destination (OD) analysis results.

Description

Technical Field

[0001] This invention relates to the fields of artificial intelligence and intelligent public transportation, and in particular to a passenger flow statistics method, device, electronic device, and storage medium. Background Technology

[0002] With the development of artificial intelligence technology, accurately acquiring passenger flow data, including both boarding and alighting information, and OD (origin-destination, referring to the starting and ending points of a passenger's journey, specifically the starting and ending stations in bus passenger flow) information is a core issue in intelligent public transportation systems. Existing solutions utilize infrared sensor technology to count passenger flow, but this technology lacks the ability to distinguish between passengers and drivers / conductors. The boarding and alighting behavior of drivers / conductors is included in the general passenger data for that trip, resulting in significant errors in passenger flow statistics and affecting the accuracy of subsequent passenger flow and OD analysis. Summary of the Invention

[0003] This invention provides a passenger flow statistics method that can extract passenger boarding and alighting behavior information and passenger information from monitoring data. It counts passenger flow of a target vehicle based solely on passenger boarding and alighting behavior information and passenger information, eliminating the possibility of incorrect counting of drivers and conductors. This can improve the accuracy of passenger count statistics for public transportation vehicles, thereby improving the accuracy of passenger flow statistics and the accuracy of subsequent passenger flow and OD analysis results.

[0004] In a first aspect, embodiments of the present invention provide a passenger flow statistics method, the method comprising:

[0005] Acquire surveillance images of the target vehicle within a specified time period, and extract passenger boarding and alighting behavior information and passenger information from the surveillance images. The passenger boarding and alighting behavior information includes passenger boarding behavior and passenger alighting behavior, and each passenger boarding and alighting behavior information corresponds to one passenger information.

[0006] Based on the passengers' boarding and alighting behavior information and the passenger information, the number of passengers in the target vehicle is counted to obtain the passenger flow statistics of the target vehicle.

[0007] Optionally, extracting passenger boarding and alighting behavior information and passenger information from the surveillance image includes:

[0008] For each station, a pre-trained target detection model is used to perform target detection on the monitoring image to obtain the detection results of people getting on and off the vehicle;

[0009] The detection results of the passengers getting on and off the bus are matched by a pre-trained multi-target tracking model to obtain the trajectory results of the passengers getting on and off the bus.

[0010] Extract the identification information of the passengers getting on and off the vehicle from the monitoring image, which corresponds to the trajectory results of the passengers getting on and off the vehicle. Each trajectory result of the passengers getting on and off the vehicle corresponds to one identification information of the passengers getting on and off the vehicle.

[0011] Based on the trajectory results of the passengers getting on and off the vehicle and their identification information, passenger information on getting on and off the vehicle and passenger information are obtained.

[0012] Optionally, the persons getting on and off the vehicle include drivers and passengers. The step of obtaining passenger boarding / alighting behavior information and passenger information based on the trajectory results and identification information of the persons getting on and off the vehicle includes:

[0013] Based on the relationship between the trajectory results of the passengers getting on and off the bus and the starting station and the ending station, the trajectory results and driver and passenger information of the drivers and passengers are obtained, and each trajectory result of the drivers and passengers corresponds to one piece of driver and passenger information.

[0014] Based on the trajectory results of the passengers getting on and off the bus and the identification information of the passengers getting on and off the bus, the trajectory results of the drivers and passengers at each station are removed, and the driver and passengers information is obtained.

[0015] Based on the passenger's trajectory, the passenger's boarding and alighting behavior information is calculated.

[0016] Optionally, calculating the passenger's boarding and alighting behavior information based on the passenger's trajectory results includes:

[0017] Calculate the congestion status of each station;

[0018] When the congestion level of the station to be counted is greater than a preset value, the trajectory results of the passengers corresponding to the station to be counted are deduplicated to obtain the trajectory results of the passengers after deduplication.

[0019] Based on the deduplicated passenger trajectory results, the passenger's boarding and alighting behavior information is calculated.

[0020] Optionally, calculating the congestion status of each station includes:

[0021] The monitoring images are processed using a pre-trained site congestion model to obtain congestion scores for each site.

[0022] The congestion status of each station is calculated based on the congestion score of each station, the trajectory results of passengers at each station, and the trajectory results of passengers getting on and off at each station.

[0023] Optionally, calculating the passenger's boarding and alighting behavior information based on the deduplicated passenger trajectory results includes:

[0024] When the station to be counted is a non-crowded station, the trajectory result of the deduplicated passenger is calculated and the first passage relationship of the first designated area is calculated. Based on the first passage relationship, it is determined whether the behavior corresponding to the trajectory result of the deduplicated passenger is boarding behavior or alighting behavior. The first designated area is set at the first designated position of the door.

[0025] When the station to be counted is a crowded station, the trajectory result of the deduplicated passenger is calculated and the second passage relationship of the second designated area is calculated. Based on the second passage relationship, it is determined whether the behavior corresponding to the trajectory result of the deduplicated passenger is getting off the bus or getting off the bus. The second designated area is set at the second designated position of the door, and the area of ​​the second designated area is smaller than the area of ​​the first designated area.

[0026] Optionally, the step of counting the number of passengers in the target vehicle based on the passengers' boarding and alighting behavior information and the passenger information to obtain the passenger flow statistics of the target vehicle includes:

[0027] Based on the passenger information, the passenger's boarding behavior is matched with the passenger's alighting behavior;

[0028] The number of successful matches between a passenger's boarding behavior and a passenger's alighting behavior is calculated, and the number of successful matches is used as the passenger flow statistics result of the target vehicle.

[0029] Secondly, embodiments of the present invention also include a passenger flow counting device, the device comprising:

[0030] The acquisition module is used to acquire monitoring images of the target vehicle within a specified time period, and extract passenger boarding and alighting behavior information and passenger information from the monitoring images. The passenger boarding and alighting behavior information includes passenger boarding behavior and passenger alighting behavior, and each passenger boarding and alighting behavior information corresponds to one passenger information.

[0031] The statistics module is used to count the number of passengers in the target vehicle based on the passengers' boarding and alighting behavior information and the passenger information, and to obtain the passenger flow statistics result of the target vehicle.

[0032] Thirdly, embodiments of the present invention provide an electronic device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps in the passenger flow statistics method provided in embodiments of the present invention.

[0033] Fourthly, embodiments of the present invention provide a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps in the passenger flow statistics method provided in the embodiments of the present invention.

[0034] In this embodiment of the invention, monitoring data of a target vehicle within a specified time period is acquired, and passenger boarding and alighting behavior information and passenger information are extracted from the monitoring data. The passenger boarding and alighting behavior information includes passenger boarding behavior and passenger alighting behavior, with each passenger boarding and alighting behavior information corresponding to one passenger information entry. Based on the passenger boarding and alighting behavior information at each station and the passenger information, the number of passengers in the target vehicle is counted to obtain the passenger flow statistics for the target vehicle. This method can extract passenger boarding and alighting behavior information and passenger information from the monitoring data, and counts the passenger flow of the target vehicle based solely on these information, eliminating errors in driver and conductor counts. This improves the accuracy of passenger count statistics for public transportation vehicles, thereby improving the accuracy of passenger flow statistics and consequently enhancing the accuracy of subsequent passenger flow and origin-destination (OD) analysis results. Attached Figure Description

[0035] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0036] Figure 1 This is a flowchart of a passenger flow statistics method provided in an embodiment of the present invention;

[0037] Figure 2 This is a schematic diagram illustrating the setting of a designated area according to an embodiment of the present invention;

[0038] Figure 3 This is a schematic diagram of the structure of a passenger flow statistics device provided in an embodiment of the present invention;

[0039] Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0040] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0041] Please see Figure 1 , Figure 1 This is a flowchart of a passenger flow statistics method provided in an embodiment of the present invention, such as... Figure 1 As shown, it includes the following steps:

[0042] 101. Obtain monitoring data of the target vehicle within a specified time period, and extract passenger boarding and alighting behavior information and passenger information from the monitoring data.

[0043] In this embodiment of the invention, the specified time period can be a statistical cycle of the target vehicle, such as the time to complete a route, a day, a month, or the data clearing cycle.

[0044] The aforementioned surveillance images can be images captured by surveillance cameras installed inside the vehicle. These surveillance images can also be referred to as surveillance data or surveillance video, or surveillance data obtained by swiping a card. The surveillance area of ​​the surveillance camera is the door area inside the vehicle.

[0045] It should be noted that during the operation of public transportation vehicles, the people inside the vehicles can be divided into drivers and conductors and passengers. Drivers and conductors are the people who provide public transportation services through public transportation vehicles, while passengers are the people who enjoy public transportation services. For passenger flow statistics, drivers and conductors should not be included in passenger flow.

[0046] In one possible embodiment, facial recognition can be performed on the surveillance images to distinguish between drivers and passengers. It is only necessary to record the facial features of the drivers and passengers, and compare the facial features of the people in the vehicle with those of the drivers and passengers. Based on the comparison results, drivers and passengers can be distinguished.

[0047] The aforementioned passenger boarding and alighting behavior information includes both boarding and alighting behaviors, with each entry corresponding to one passenger profile. Boarding behavior can be the act of a passenger moving from the vehicle into the vehicle; that is, if a passenger was not present in the vehicle before a certain moment but appears after that moment, then that behavior constitutes boarding. Conversely, if a passenger was present in the vehicle before a certain moment but is no longer present after that moment, then that behavior constitutes alighting.

[0048] In this embodiment of the invention, boarding and alighting behaviors can be determined by the passenger's trajectory, which includes the passenger's spatiotemporal information. By using the passenger's spatiotemporal information, the spatial location of the passenger at different times can be obtained, and the passenger's boarding and alighting behavior information can be extracted.

[0049] Specifically, in the steps of extracting passenger boarding and alighting behavior information and passenger information from surveillance images, for each station, a pre-trained object detection model can be used to perform object detection on the surveillance images to obtain the detection results of boarding and alighting personnel; a pre-trained multi-object tracking model can be used to match the detection results of boarding and alighting personnel to obtain the trajectory results of boarding and alighting personnel; the identification information of boarding and alighting personnel corresponding to the trajectory results of boarding and alighting personnel can be extracted from the surveillance images, with each trajectory result of boarding and alighting personnel corresponding to one identification information of boarding and alighting personnel; based on the trajectory results of boarding and alighting personnel and the identification information of boarding and alighting personnel, the passenger boarding and alighting behavior information and passenger information can be obtained.

[0050] In this embodiment of the invention, the target vehicle opens its doors when passing a station, allowing passengers to disembark and passengers to board. Monitoring images captured during the duration the doors are open are processed using a pre-trained target detection model. This pre-trained target detection model can be a convolutional neural network-based target detection algorithm, such as Faster R-CNN, Mask R-CNN, RFCN, YOLO, YOLO2, SSD, or DSSD. The pre-trained target detection model can perform real-time detection of passengers boarding and alighting.

[0051] After obtaining the real-time detection results of passengers getting on and off the vehicle, a pre-trained multi-object tracking model can be used to match these detection results to obtain the trajectory results of the passengers. This multi-object tracking model can track passengers getting on and off the vehicle, thus forming their trajectory results. These trajectory results include the starting point, ending point, direction, and length of the trajectory. This multi-object tracking model only matches the detection results of passengers getting on and off the vehicle, thus eliminating the need to track other people inside the vehicle. The pre-trained multi-object tracking model can be derived from existing multi-object tracking algorithms, such as deep learning multi-object tracking algorithms, multi-object tracking algorithms based on Siamese symmetric networks, and target tracking based on fully connected Siamese-FC networks.

[0052] After obtaining the trajectory results of passengers getting on and off the vehicle, for each trajectory result, the identification information of the corresponding passengers getting on and off the vehicle can be extracted from the monitoring image. The identification information of passengers getting on and off the vehicle can be the identification features of the personnel, such as facial features, body features, etc.

[0053] Among passengers getting on and off the vehicle, if a passenger's trajectory shows them passing through a door from outside the vehicle to inside, that passenger is identified as a boarding passenger, and their behavior is confirmed as boarding. Conversely, if a passenger's trajectory shows them passing through a door from inside the vehicle to outside, that passenger is identified as a disembarking passenger, and their behavior is confirmed as disembarking. After removing the driver and conductor from the list of passengers getting on and off the vehicle, we can obtain the passengers' boarding and disembarking behavior information, as well as the passenger information itself.

[0054] By performing target detection and multi-target tracking on surveillance images, the trajectory results of people getting on and off the bus can be obtained, which can more accurately determine the people getting on and off the bus, thereby improving the accuracy of passenger flow statistics.

[0055] Specifically, in the steps of obtaining passenger boarding / alighting behavior information and passenger information based on the trajectory results and identification information of boarding / alighting passengers, the trajectory results and driver / passenger information of the driver / passenger can be obtained based on the relationship between the trajectory results of boarding / alighting passengers and the starting and ending stations, with each driver / passenger's trajectory result corresponding to one driver / passenger information record; the trajectory results and driver / passenger information of each station are removed from the trajectory results and identification information of boarding / alighting passengers to obtain the passenger's trajectory results and passenger information; and the passenger's boarding / alighting behavior information is calculated based on the passenger's trajectory results.

[0056] In one possible implementation, passengers disembarking at the starting station and passengers boarding at the terminal station can be identified as drivers and conductors. Passengers who board and then disembark at the same station can be considered drivers and conductors since they did not receive any transportation service.

[0057] Furthermore, we can define the total boarding and alighting records as set A, where each record includes a corresponding boarding / alighting trajectory. Let the records of drivers and passengers to be removed be set E, which includes their trajectory results. Set E is initialized to empty. Let the similarity threshold for the identification features be t. Specific frames can be extracted from the monitoring images corresponding to the trajectory results as representative frames for subsequent feature extraction. Take the first boarding record at the starting station and compare its recording time with the card-swiping data time series to determine if a card was swiped. If no card was swiped, the record is considered a driver / passenger, added to set E, and removed from set A. Add all alighting records from the starting station to set E and remove the corresponding records from set A. Add all boarding records from the destination station to set E and remove the corresponding records from set A. For other stations, the ReID model can be used to calculate the identification features of all records in set A, denoted as feature set F. ACalculate the identifying features for all records in set E, denoted as feature set F. E Take a record 'a' from set A, and use the cosine distance formula to calculate the similarity score between the corresponding features of record 'a' and the corresponding features of all records in set E. Specifically, this can be done from feature set F. A Extract the recognition features corresponding to record a from the set F, and combine the recognition features corresponding to record a with the feature set F. E Calculate the similarity score. If any record in set E has a similarity score greater than the threshold t with the identification features of record a, then record a is considered to be a driver / passenger, and is added to set E and removed from set A. After there are no driver / passenger records in set A, return sets A and E. Set A is the set of passenger boarding / alighting records after removing driver / passenger records, and set E is the set of driver / passenger boarding / alighting records.

[0058] Specifically, in the step of calculating passenger boarding and alighting behavior information based on passenger trajectory results, the congestion status of each station can be calculated; when the congestion status of the station to be counted is greater than a preset status value, the trajectory results of the passengers corresponding to the station to be counted are deduplicated to obtain the deduplicated passenger trajectory results; based on the deduplicated passenger trajectory results, the passenger boarding and alighting behavior information is calculated.

[0059] It should be noted that at some crowded stops, some passengers may disembark and then re-board to facilitate other passengers' boarding and alighting. For example, someone standing by the door before reaching their destination might disembark if other passengers need to get off, but there isn't enough space at the door. In such cases, this disembarking and re-boarding behavior should not be counted as a single passenger flow, as this would result in double counting. By deduplicating passenger boarding and alighting behavior data at crowded stops, the accuracy of passenger boarding and alighting information can be improved, thereby enhancing the accuracy of passenger flow statistics.

[0060] More specifically, the monitoring images can be processed using a pre-trained station congestion model to obtain the congestion score for each station; based on the congestion score of each station, the trajectory results of passengers at each station, and the trajectory results of people getting on and off at each station, the congestion status of each station can be calculated.

[0061] To determine the congestion status of a bus stop, surveillance images of the target vehicle within a preset time frame before the doors open can be acquired. A target detection and tracking algorithm is used to detect and track passengers in the video, outputting the number of passenger trajectories detected and tracked within that preset time frame. A pre-trained station congestion model is then used to process the surveillance images within the preset time frame before the doors open, outputting a station congestion score. The station congestion score is calculated as: Station Congestion Score * (Number of Passenger Trajectories / Maximum Number of Trajectories). The maximum number of trajectories can be a preset value, determined based on the target vehicle's maximum passenger capacity. If the number of trajectories counted before the doors open exceeds this value, it indicates congestion. The greater the excess, the higher the ratio, indicating greater congestion. If the station's congestion score is greater than the preset value (e.g., greater than 0.5), the station is considered congested; otherwise, it is considered non-congested.

[0062] By calculating the congestion level of each station to determine whether a station is congested, and deduplicating the trajectory results of passengers at congested stations, the accuracy of passenger boarding and alighting behavior information can be increased. The trajectory results of passengers at non-congested stations are not deduplicated to avoid unnecessary calculations, thereby improving the calculation speed.

[0063] In one possible embodiment, the step of deduplicating the trajectory results of passengers corresponding to congested stations includes deduplication within the same station and deduplication across different stations. In deduplication within the same station, surveillance images of passenger boarding and alighting records are taken, and the ReID model is used to extract the identification features corresponding to each boarding and alighting record. The feature similarity score between the identification features corresponding to all records is calculated pairwise. If the feature similarity score of two records is greater than a preset threshold, then these two records can be considered to have originated from the same passenger. If two records are both boarding records, the later record is removed, and the earlier record is retained. If two records are both alighting records, the earlier record is removed, and the later record is retained. If two records are one boarding record and one alighting record, then both records are removed.

[0064] In cross-station deduplication, when a station is identified as a congested station, passenger boarding and alighting deduplication is performed between consecutive congested stations until the next non-congested station is identified. Boarding and alighting records from these congested stations are collected, and the ReID model is used to extract the corresponding identification features for each record. The similarity score between all boarding records is calculated pairwise. If the feature similarity score between two records is greater than a preset threshold, the latter record is removed first, and the former is retained. Similarly, the feature similarity score between all alighting records is calculated pairwise. If the feature similarity score between two records is greater than a preset threshold, the former record is removed first, and the latter is retained.

[0065] It should be noted that the feature similarity between two records is calculated using the corresponding identification features of the two records.

[0066] In one possible embodiment, during the step of obtaining boarding and alighting records, a monitoring image of the period from when the door opens to when it closes can be acquired, and a target detection and tracking algorithm can be used to detect and track passengers in the video to obtain the trajectory results of passengers during the period from when the door opens to when it closes.

[0067] Specifically, in calculating passenger boarding and alighting behavior information based on the deduplicated passenger trajectory results, when the station to be counted is a non-crowded station, the deduplicated passenger trajectory results are calculated with a first passage relationship in a first designated area, and the behavior corresponding to the deduplicated passenger trajectory results is determined to be boarding or alighting based on the first passage relationship. The first designated area is set at a first designated position of the vehicle door. When the station to be counted is a crowded station, the deduplicated passenger trajectory results are calculated with a second passage relationship in a second designated area, and the behavior corresponding to the deduplicated passenger trajectory results is determined to be alighting or alighting based on the second passage relationship. The second designated area is set at a second designated position of the vehicle door, and the area of ​​the second designated area is smaller than the area of ​​the first designated area.

[0068] Furthermore, double lanes for getting on and off can be installed inside the car doors, such as... Figure 2 As shown, Figure 2 This is a schematic diagram illustrating the setting of a designated area according to an embodiment of the present invention. The non-congested station boarding / alighting lines (L1, L2), i.e., the first designated area, are the area between L1 and L2. The congested station boarding / alighting lines (L1, L3), i.e., the second designated area, are the area between L1 and L3. The starting position of the trajectory is determined. When the trajectory crosses both lines from top to bottom, it is determined to be a boarding record; when the trajectory crosses both lines from bottom to top, it is determined to be an alighting record.

[0069] 102. Based on passenger boarding and alighting behavior information and passenger information, count the number of passengers in the target vehicle to obtain the passenger flow statistics of the target vehicle.

[0070] In this embodiment of the invention, after the target vehicle completes a route, the number of passengers on the target vehicle can be counted based on the passengers' boarding and alighting behavior information and passenger information during that route, thereby obtaining the passenger flow statistics of the target vehicle. Passenger flow can be interpreted as the number of passengers who alight after receiving a complete transportation service from boarding. The above passenger flow statistics can be used for subsequent passenger flow analysis and OD analysis.

[0071] Specifically, passenger boarding behavior can be matched with passenger alighting behavior based on passenger information; the number of successful matches between passenger boarding behavior and passenger alighting behavior can be calculated, and the number of successful matches can be used as the passenger flow statistics result for the target vehicle.

[0072] Since a complete transportation service begins with boarding and ends with alighting, passenger boarding and alighting behaviors can be matched based on passenger information. Each successful match increments the passenger count and passenger flow statistics by 1.

[0073] In this embodiment of the invention, monitoring data of the target vehicle within a specified time period is acquired, and passenger boarding and alighting behavior information and passenger information are extracted from the monitoring data. Passenger boarding and alighting behavior information includes passenger boarding and alighting behaviors, with each passenger boarding / alighting behavior corresponding to one passenger information entry. Based on the passenger boarding / alighting behavior information and passenger information at each stop, the number of passengers in the target vehicle is counted to obtain the passenger flow statistics for the target vehicle. This method, which extracts passenger boarding / alighting behavior information and passenger information from the monitoring data and uses only this information to count the passenger flow of the target vehicle, eliminates errors in driver and conductor counts, improves the accuracy of passenger count statistics for public transportation vehicles, and thus improves the accuracy of passenger flow statistics. This, in turn, enhances the accuracy of subsequent passenger flow and origin-destination (OD) analysis results.

[0074] It should be noted that the passenger flow statistics method provided in this embodiment of the invention can be applied to devices such as mobile phones, monitors, computers, and servers that can perform passenger flow statistics.

[0075] Please see Figure 3 , Figure 3 This is a schematic diagram of the structure of a passenger flow statistics device provided in an embodiment of the present invention, as shown below. Figure 3 As shown, the device includes:

[0076] The acquisition module 301 is used to acquire monitoring images of the target vehicle within a specified time period, and extract passenger boarding and alighting behavior information and passenger information from the monitoring images. The passenger boarding and alighting behavior information includes passenger boarding behavior and passenger alighting behavior, and each passenger boarding and alighting behavior information corresponds to one passenger information.

[0077] The statistics module 302 is used to count the number of passengers in the target vehicle based on the passenger boarding and alighting behavior information and the passenger information, and to obtain the passenger flow statistics result of the target vehicle.

[0078] Optionally, module 301 includes:

[0079] The detection submodule is used to perform target detection on the monitoring image for each station using a pre-trained target detection model to obtain the detection results of people getting on and off the vehicle;

[0080] The matching submodule is used to match the detection results of the passengers getting on and off the vehicle using a pre-trained multi-target tracking model to obtain the trajectory results of the passengers getting on and off the vehicle.

[0081] The first identification submodule is used to extract identification information of the passengers getting on and off the vehicle from the monitoring image, corresponding to the trajectory results of the passengers getting on and off the vehicle. Each trajectory result of the passengers getting on and off the vehicle corresponds to one piece of identification information of the passengers getting on and off the vehicle.

[0082] The second identification submodule is used to obtain passenger boarding and alighting behavior information and passenger information based on the trajectory results of the boarding and alighting personnel and the identification information of the boarding and alighting personnel.

[0083] Optionally, the persons getting on and off the vehicle include drivers and passengers, and the second identification submodule includes:

[0084] The corresponding unit is used to obtain the trajectory results of the driver and passenger and the driver and passenger information based on the relationship between the trajectory results of the passengers getting on and off the bus and the starting station and the ending station. Each trajectory result of the driver and passenger corresponds to one piece of driver and passenger information.

[0085] The elimination unit is used to eliminate the trajectory results and driver and passenger information of each station from the trajectory results of the passengers getting on and off the bus and the identification information of the passengers getting on and off the bus, so as to obtain the trajectory results and passenger information of the passenger.

[0086] The calculation unit is used to calculate the passenger's boarding and alighting behavior information based on the passenger's trajectory results.

[0087] Optional computing units include:

[0088] The first calculation subunit is used to calculate the congestion status of each station;

[0089] The deduplication subunit is used to deduplicate the trajectory results of passengers corresponding to the station to be counted when the congestion status of the station to be counted is greater than a preset status value, so as to obtain the deduplicated trajectory results of passengers.

[0090] The second calculation subunit is used to calculate the passenger's boarding and alighting behavior information based on the deduplicated passenger trajectory results.

[0091] Optionally, the first calculation subunit is further configured to process the monitoring image using a pre-trained station congestion model to obtain a congestion score for each station; and to calculate the congestion status of each station based on the congestion scores of each station, the trajectory results of passengers at each station, and the trajectory results of passengers getting on and off at each station.

[0092] Optionally, the second calculation subunit is further configured to: when the station to be counted is a non-crowded station, calculate the trajectory result of the deduplicated passenger and the first passage relationship of the first designated area, and determine whether the behavior corresponding to the trajectory result of the deduplicated passenger is boarding or alighting based on the first passage relationship, wherein the first designated area is set at the first designated position of the door; when the station to be counted is a crowded station, calculate the trajectory result of the deduplicated passenger and the second passage relationship of the second designated area, and determine whether the behavior corresponding to the trajectory result of the deduplicated passenger is alighting or alighting based on the second passage relationship, wherein the second designated area is set at the second designated position of the door, and the area of ​​the second designated area is smaller than the area of ​​the first designated area.

[0093] Optional, the statistics module 302 includes:

[0094] The matching submodule is used to match the passenger's boarding behavior with the passenger's alighting behavior based on the passenger information.

[0095] The calculation submodule is used to calculate the number of times the passenger's boarding behavior and the passenger's alighting behavior are successfully matched, and the number of successful matches is used as the passenger flow statistics result of the target vehicle.

[0096] It should be noted that the passenger flow statistics device provided in this embodiment of the invention can be applied to devices such as mobile phones, monitors, computers, and servers that can perform passenger flow statistics.

[0097] The passenger flow counting device provided in this embodiment of the invention can implement all the processes implemented by the passenger flow counting method in the above-described method embodiments, and can achieve the same beneficial effects. To avoid repetition, it will not be described again here.

[0098] See Figure 4 , Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention, such as... Figure 4 As shown, it includes: a memory 402, a processor 401, and a computer program stored in the memory 402 and executable on the processor 401, wherein:

[0099] The processor 401 is used to call the computer program stored in the memory 402 and perform the following steps:

[0100] Acquire surveillance images of the target vehicle within a specified time period, and extract passenger boarding and alighting behavior information and passenger information from the surveillance images. The passenger boarding and alighting behavior information includes passenger boarding behavior and passenger alighting behavior, and each passenger boarding and alighting behavior information corresponds to one passenger information.

[0101] Based on the passengers' boarding and alighting behavior information and the passenger information, the number of passengers in the target vehicle is counted to obtain the passenger flow statistics of the target vehicle.

[0102] Optionally, the process of extracting passenger boarding and alighting behavior information and passenger information from the monitoring image executed by processor 401 includes:

[0103] For each station, a pre-trained target detection model is used to perform target detection on the monitoring image to obtain the detection results of people getting on and off the vehicle;

[0104] The detection results of the passengers getting on and off the bus are matched by a pre-trained multi-target tracking model to obtain the trajectory results of the passengers getting on and off the bus.

[0105] Extract the identification information of the passengers getting on and off the vehicle from the monitoring image, which corresponds to the trajectory results of the passengers getting on and off the vehicle. Each trajectory result of the passengers getting on and off the vehicle corresponds to one identification information of the passengers getting on and off the vehicle.

[0106] Based on the trajectory results of the passengers getting on and off the vehicle and their identification information, passenger information on getting on and off the vehicle and passenger information are obtained.

[0107] Optionally, the persons getting on and off the vehicle include drivers and passengers. The step executed by processor 401 to obtain passenger boarding and alighting behavior information and passenger information based on the trajectory results and identification information of the persons getting on and off the vehicle includes:

[0108] Based on the relationship between the trajectory results of the passengers getting on and off the bus and the starting station and the ending station, the trajectory results and driver and passenger information of the drivers and passengers are obtained, and each trajectory result of the drivers and passengers corresponds to one piece of driver and passenger information.

[0109] Based on the trajectory results of the passengers getting on and off the bus and the identification information of the passengers getting on and off the bus, the trajectory results of the drivers and passengers at each station are removed, and the driver and passengers information is obtained.

[0110] Based on the passenger's trajectory, the passenger's boarding and alighting behavior information is calculated.

[0111] Optionally, the step of processor 401 calculating the passenger's boarding and alighting behavior information based on the passenger's trajectory results includes:

[0112] Calculate the congestion status of each station;

[0113] When the congestion level of the station to be counted is greater than a preset value, the trajectory results of the passengers corresponding to the station to be counted are deduplicated to obtain the trajectory results of the passengers after deduplication.

[0114] Based on the deduplicated passenger trajectory results, the passenger's boarding and alighting behavior information is calculated.

[0115] Optionally, the calculation of the congestion status of each site performed by processor 401 includes:

[0116] The monitoring images are processed using a pre-trained site congestion model to obtain congestion scores for each site.

[0117] The congestion status of each station is calculated based on the congestion score of each station, the trajectory results of passengers at each station, and the trajectory results of passengers getting on and off at each station.

[0118] Optionally, the step of processor 401 calculating the passenger's boarding and alighting behavior information based on the deduplicated passenger trajectory results includes:

[0119] When the station to be counted is a non-crowded station, the trajectory result of the deduplicated passenger is calculated and the first passage relationship of the first designated area is calculated. Based on the first passage relationship, it is determined whether the behavior corresponding to the trajectory result of the deduplicated passenger is boarding behavior or alighting behavior. The first designated area is set at the first designated position of the door.

[0120] When the station to be counted is a crowded station, the trajectory result of the deduplicated passenger is calculated and the second passage relationship of the second designated area is calculated. Based on the second passage relationship, it is determined whether the behavior corresponding to the trajectory result of the deduplicated passenger is getting off the bus or getting off the bus. The second designated area is set at the second designated position of the door, and the area of ​​the second designated area is smaller than the area of ​​the first designated area.

[0121] Optionally, the step of processor 401 performing passenger count statistics on the target vehicle based on the passenger boarding and alighting behavior information and the passenger information to obtain the passenger flow statistics result of the target vehicle includes:

[0122] Based on the passenger information, the passenger's boarding behavior is matched with the passenger's alighting behavior;

[0123] The number of successful matches between a passenger's boarding behavior and a passenger's alighting behavior is calculated, and the number of successful matches is used as the passenger flow statistics result of the target vehicle.

[0124] It should be noted that the aforementioned electronic devices can be mobile phones, monitors, computers, servers, and other devices that can be used for passenger flow statistics.

[0125] The electronic device provided in this embodiment of the invention can implement each process of the passenger flow statistics method in the above method embodiment, and can achieve the same beneficial effect. To avoid repetition, it will not be described again here.

[0126] This invention also provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, it implements the various processes of the passenger flow statistics method provided in this invention and achieves the same technical effect. To avoid repetition, it will not be described again here.

[0127] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM), etc.

[0128] The above description discloses only preferred embodiments of the present invention and should not be construed as limiting the scope of the present invention. Therefore, equivalent variations made in accordance with the claims of the present invention are still within the scope of the present invention.

Claims

1. A passenger flow statistics method for public transportation vehicles, characterized in that, Includes the following steps: Acquire surveillance images of the target vehicle within a specified time period, and extract passenger boarding and alighting behavior information and passenger information from the surveillance images. The passenger boarding and alighting behavior information includes passenger boarding behavior and passenger alighting behavior, and each passenger boarding and alighting behavior information corresponds to one passenger information. Based on the passenger boarding and alighting behavior information and the passenger information, the number of passengers in the target vehicle is counted to obtain the passenger flow statistics of the target vehicle. Extracting passenger boarding and alighting behavior information from the surveillance images includes: Calculate the congestion status of each station; When the congestion level of the station to be counted exceeds a preset value, the trajectory results of passengers corresponding to the station to be counted are deduplicated to obtain the deduplicated trajectory results of passengers. The deduplication steps for the trajectory results of passengers corresponding to the station to be counted include deduplication within the same station and deduplication across different stations. In the deduplication process at the same station, the monitoring images of passenger boarding and alighting records at the station are taken, and the identification features corresponding to each boarding and alighting record are extracted. The feature similarity score between the identification features corresponding to all records is calculated pairwise. If the feature similarity score of two records is greater than the preset threshold, if the two records are both boarding records, the later one is removed and the earlier one is retained. If the two records are both alighting records, the earlier one is removed and the later one is retained. If the two records are one boarding record and one alighting record, both records are removed. In the deduplication process at different stations, boarding and alighting records of consecutively crowded stations are taken, and the identification features corresponding to each record are extracted. The similarity score between all boarding records is calculated pairwise. If the feature similarity score between two records is greater than a preset threshold, the later record is removed and the earlier record is retained. Similarly, the feature similarity score between all alighting records is calculated pairwise. If the feature similarity score between two records is greater than a preset threshold, the earlier record is removed and the later record is retained. Based on the deduplicated passenger trajectory results, the passenger's boarding and alighting behavior information is calculated.

2. The method as described in claim 1, characterized in that, The step of extracting passenger boarding and alighting behavior information and passenger information from the surveillance images includes: For each station, a pre-trained target detection model is used to perform target detection on the monitoring image to obtain the detection results of people getting on and off the vehicle; The detection results of the passengers getting on and off the bus are matched by a pre-trained multi-target tracking model to obtain the trajectory results of the passengers getting on and off the bus. Extract the identification information of the passengers getting on and off the vehicle from the monitoring image, which corresponds to the trajectory results of the passengers getting on and off the vehicle. Each trajectory result of the passengers getting on and off the vehicle corresponds to one identification information of the passengers getting on and off the vehicle. Based on the trajectory results of the passengers getting on and off the vehicle and their identification information, passenger information on getting on and off the vehicle and passenger information are obtained.

3. The method as described in claim 2, characterized in that, The persons getting on and off the vehicle include drivers and passengers. The process of obtaining passenger boarding and alighting behavior information and passenger information based on the trajectory results and identification information of these persons includes: Based on the relationship between the trajectory results of the passengers getting on and off the bus and the starting station and the ending station, the trajectory results and driver and passenger information of the drivers and passengers are obtained, and each trajectory result of the drivers and passengers corresponds to one piece of driver and passenger information. Based on the trajectory results of the passengers getting on and off the bus and the identification information of the passengers getting on and off the bus, the trajectory results of the drivers and passengers at each station are removed, and the driver and passengers information is obtained. Based on the passenger's trajectory, the passenger's boarding and alighting behavior information is calculated.

4. The method as described in claim 3, characterized in that, The calculation of the congestion status of each station includes: The monitoring images are processed using a pre-trained site congestion model to obtain congestion scores for each site. The congestion status of each station is calculated based on the congestion score of each station, the trajectory results of passengers at each station, and the trajectory results of passengers getting on and off at each station.

5. The method as described in claim 3, characterized in that, The step of calculating the passenger's boarding and alighting behavior information based on the deduplicated passenger trajectory results includes: When the station to be counted is a non-crowded station, the trajectory result of the deduplicated passenger is calculated and the first passage relationship of the first designated area is calculated. Based on the first passage relationship, it is determined whether the behavior corresponding to the trajectory result of the deduplicated passenger is boarding behavior or alighting behavior. The first designated area is set at the first designated position of the door. When the station to be counted is a crowded station, the trajectory result of the deduplicated passenger is calculated and the second passage relationship of the second designated area is calculated. Based on the second passage relationship, it is determined whether the behavior corresponding to the trajectory result of the deduplicated passenger is getting off the bus or getting off the bus. The second designated area is set at the second designated position of the door, and the area of ​​the second designated area is smaller than the area of ​​the first designated area.

6. The method according to any one of claims 1 to 5, characterized in that, The step of counting the number of passengers in the target vehicle based on the passengers' boarding and alighting behavior information and the passenger information to obtain the passenger flow statistics of the target vehicle includes: Based on the passenger information, the passenger's boarding behavior is matched with the passenger's alighting behavior; The number of successful matches between a passenger's boarding behavior and a passenger's alighting behavior is calculated, and the number of successful matches is used as the passenger flow statistics result of the target vehicle.

7. A passenger flow counting device, characterized in that, The device includes: The acquisition module is used to acquire monitoring images of the target vehicle within a specified time period, and extract passenger boarding and alighting behavior information and passenger information from the monitoring images. The passenger boarding and alighting behavior information includes passenger boarding behavior and passenger alighting behavior, and each passenger boarding and alighting behavior information corresponds to one passenger information. The statistics module is used to count the number of passengers in the target vehicle based on the passengers' boarding and alighting behavior information and the passenger information, and to obtain the passenger flow statistics result of the target vehicle. Extracting passenger boarding and alighting behavior information from the surveillance images includes: Calculate the congestion status of each station; When the congestion level of the station to be counted exceeds a preset value, the trajectory results of passengers corresponding to the station to be counted are deduplicated to obtain the deduplicated trajectory results of passengers. The deduplication steps for the trajectory results of passengers corresponding to the station to be counted include deduplication within the same station and deduplication across different stations. In the deduplication process at the same station, the monitoring images of passenger boarding and alighting records at the station are taken, and the identification features corresponding to each boarding and alighting record are extracted. The feature similarity score between the identification features corresponding to all records is calculated pairwise. If the feature similarity score of two records is greater than the preset threshold, if the two records are both boarding records, the later one is removed and the earlier one is retained. If the two records are both alighting records, the earlier one is removed and the later one is retained. If the two records are one boarding record and one alighting record, both records are removed. In the deduplication process at different stations, boarding and alighting records of consecutively crowded stations are taken, and the identification features corresponding to each record are extracted. The similarity score between all boarding records is calculated pairwise. If the feature similarity score between two records is greater than a preset threshold, the later record is removed and the earlier record is retained. Similarly, the feature similarity score between all alighting records is calculated pairwise. If the feature similarity score between two records is greater than a preset threshold, the earlier record is removed and the later record is retained. Based on the deduplicated passenger trajectory results, the passenger's boarding and alighting behavior information is calculated.

8. An electronic device, characterized in that, include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, implements the steps in the passenger flow statistics method as described in any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the steps of the passenger flow statistics method as described in any one of claims 1 to 6.

Images