Method and system for calculating passenger flow volume in passenger transport industry

Abstract

The invention discloses a passenger flow volume calculation method and system in the passenger transport industry, and belongs to the technical field of public transport management, and the method comprises the steps: judging whether a first snapshot triggering condition is satisfied, and if yes, sending a snapshot control instruction to a vehicle-mounted terminal, so that the vehicle-mounted terminal collects a panoramic image covering the whole area in a vehicle and uploads the panoramic image back to a server; the first snapshot triggering condition is that in a preset time period, the real-time speed is firstly smaller than a first speed threshold value and then larger than a second speed threshold value; and receiving the panoramic image, carrying out target identification processing on the panoramic image, and counting the actual passenger number. Whether the vehicle runs at a low speed or stops is judged through the first speed threshold value, if yes, whether the vehicle enters a stable running state is judged through the second speed threshold value, if yes, the vehicle completes one-time passenger getting-on and getting-off and enters a relatively stable running state, and target recognition is conducted on the panoramic image collected in the state; the accuracy of passenger flow volume statistics can be improved.

Description

Technical Field

[0001] This invention relates to the field of public transportation management technology, and in particular to a method and system for calculating passenger flow in the passenger transport industry. Background Technology

[0002] In the passenger transport industry, the statistics of passenger numbers in operating vehicles are a crucial issue. On the one hand, due to regulatory requirements, accurate real-time passenger numbers are essential for safe operation. On the other hand, the mobility of vehicle schedules and the randomness of passenger boarding and alighting locations, including situations where passengers board and alight outside of designated stations or purchase tickets outside of stations, increase the difficulty for passenger transport companies in managing vehicle license plates and fares. This leads to incomplete statistical data on fares and fares, creating management loopholes and resulting in economic losses.

[0003] In real-world scenarios, before a vehicle departs from a station, the station can accurately obtain the number of passengers boarding through the ticketing system. For passenger boarding and alighting outside the station, traditional solutions use passenger flow cameras for statistics. Typically, one camera is deployed at the front door and another at the rear door, with the front camera used for boarding and the rear camera for alighting. The difference between the two cameras' values ​​is used to count passenger flow. This method requires specific passenger boarding and alighting procedures, making it difficult to reflect actual boarding and alighting situations. It also has limitations, requiring the vehicle to have two doors. Furthermore, since drivers and attendants also board and alight during the journey, existing statistical methods struggle to effectively distinguish between drivers / attendants and passengers, further deviating from passenger flow statistics and impacting the accuracy and effectiveness of passenger flow management. Summary of the Invention

[0004] The purpose of this invention is to overcome the problems of the prior art and provide a method and system for calculating passenger flow in the passenger transport industry.

[0005] The objective of this invention is achieved through the following technical solution: a method for calculating passenger flow in the passenger transport industry, comprising the following steps: Obtain the vehicle's real-time speed; Determine whether the first capture trigger condition is met. If it is, send a capture control command to the vehicle terminal to enable the vehicle terminal to collect a panoramic image covering the entire area inside the vehicle and upload it back to the server. The first capture trigger condition is: within a preset time period, the real-time speed is first less than a first speed threshold and then greater than a second speed threshold. The first speed threshold corresponds to the maximum speed at which passengers are allowed to get on and off the vehicle, and the second speed threshold corresponds to the minimum speed at which passengers are not allowed to get on or off the vehicle. The preset time period is determined based on the time interval between one passenger pick-up and drop-off. Receive panoramic images, perform target recognition processing on the panoramic images, and count the actual number of passengers.

[0006] In one example, the first capture trigger condition is replaced with a second capture trigger condition. The second capture trigger condition is: within a preset time period, the real-time speed is first less than a first speed threshold and then greater than a second speed threshold, and the real-time location of the vehicle enters the electronic fence; where the electronic fence is the area outside the station.

[0007] In one example, prior to performing target recognition processing on the panoramic image, the method further includes: Based on the annotation information of passengers and seats in the current historical panoramic image of the vehicle, calculate the probability density distribution map of passengers and seats in the image; The current panoramic image is gridded based on the probability density distribution map, and the probability density value of each grid region is calculated. Grid regions where the probability density of passengers and seats is higher than the preset probability density threshold are identified as regions of interest.

[0008] In one example, after the region of interest is determined, the process further includes: The region of interest is expanded by boundary extension to obtain the final region of interest.

[0009] In one example, after the region of interest is determined, the process further includes: Calculate the first brightness difference between the average brightness of the passenger area and the average brightness of the background area in the current historical panoramic image of the vehicle. Calculate the second brightness difference between the average brightness of the region of interest and the average brightness of the non-region of interest in the current panoramic image; The first brightness difference and the second brightness difference are compared and processed, and the overall brightness of the area of ​​interest is compensated and adjusted based on the comparison result.

[0010] In one example, the step of compensating for the overall brightness of the region of interest based on the comparison results includes: When the second brightness difference falls within the preset brightness threshold range centered on the first brightness difference, no brightness compensation processing is required. When the second brightness difference is greater than the upper limit of the preset brightness threshold range, the brightness of the area of ​​interest is increased; When the second brightness difference is less than the lower limit of the preset brightness threshold range, the brightness of the area of ​​interest is reduced.

[0011] In one example, after counting the actual number of passengers, the process also includes: The system compares the actual number of passengers with the ticket sales data for the corresponding vehicle trip. If the actual number of passengers does not match the ticket sales data, a ticketing error message is generated.

[0012] It should be further noted that the technical features corresponding to the above examples can be combined or replaced to form new technical solutions.

[0013] This invention discloses a method for calculating passenger flow in the passenger transport industry, with an on-board terminal as the executing entity, and includes the following steps: Collect the real-time speed of the vehicle and upload it to the server; The receiving server sends a capture control command when the first capture trigger condition is met based on the real-time speed. The first capture trigger condition is: within a preset time period, the real-time speed is first less than a first speed threshold and then greater than a second speed threshold. The first speed threshold corresponds to the maximum speed at which passengers are allowed to get on and off the vehicle, and the second speed threshold corresponds to the minimum speed at which passengers are not allowed to get on or off the vehicle. The preset time period is determined based on the time interval between when a vehicle completes one passenger pick-up and drop-off. The system collects panoramic images covering the entire interior of the vehicle and uploads them to a server. The server then performs target recognition processing on the panoramic images and counts the actual number of passengers.

[0014] In one example, the method further includes: Collect the real-time location of the vehicle and upload it to the server; The receiving server receives a capture control command when the second capture trigger condition is met based on the vehicle's real-time speed and real-time location. It then collects panoramic images covering the entire area inside the vehicle and uploads them to the server. The second capture trigger condition is: within a preset time period, the real-time speed is first less than the first speed threshold and then greater than the second speed threshold, and the real-time position of the vehicle enters the electronic fence; whereby the electronic fence is the area outside the station.

[0015] The present invention also includes a passenger flow calculation system for the passenger transport industry, the system comprising an on-board terminal and a server; The vehicle-mounted terminal is used to collect the real-time speed of the vehicle and upload it to the server, or to collect the real-time speed and real-time location of the vehicle and upload it to the server. The server is used to receive the vehicle's real-time speed, or to receive the vehicle's real-time speed and real-time location. The server is also used to determine whether the first or second capture trigger condition is met. If met, a capture control command is sent to the vehicle terminal. The first capture trigger condition is: within a preset time period, the real-time speed is first less than a first speed threshold and then greater than a second speed threshold. The first speed threshold corresponds to the maximum speed at which passengers are allowed to get on and off the vehicle, and the second speed threshold corresponds to the minimum speed at which passengers are not allowed to get on or off the vehicle. The preset time period is determined based on the time interval between a vehicle completing one passenger pick-up and drop-off. The second capture trigger condition is: within a preset time period, the real-time speed is first less than the first speed threshold and then greater than the second speed threshold, and the real-time location of the vehicle enters the electronic fence. The electronic fence is the area outside the station. The vehicle-mounted terminal is also used to receive capture control commands sent by the server, trigger the acquisition of panoramic images covering the entire area inside the vehicle, and upload them back to the server. The server is also used to receive panoramic images, perform target recognition processing on the panoramic images, and count the actual number of passengers.

[0016] It should be further noted that the technical features corresponding to the above examples can be combined or replaced to form new technical solutions.

[0017] Compared with the prior art, the beneficial effects of the present invention are: 1. In one example, a first speed threshold is used to determine whether the vehicle is driving at low speed or stopped. If so, a second speed threshold is used to determine whether the vehicle has entered a stable driving state. If so, the vehicle has completed one passenger pick-up and drop-off and entered a relatively stable driving state. At this time, the people on the vehicle no longer move and are relatively stationary. Target recognition of the panoramic image collected in this state can eliminate the influence of the driver and attendants getting on and off the vehicle on passenger flow statistics, thereby truly reflecting the number of passengers in the vehicle and improving the accuracy of passenger flow statistics.

[0018] 2. In one example, by combining the spatial constraint of an electronic fence, the timing of image acquisition is further set to the area outside the station that needs to be monitored, thereby achieving refined passenger flow monitoring and improving the accuracy of passenger flow statistics and the effectiveness of management.

[0019] 3. In one example, the effective focus area for each vehicle is adaptively determined using probabilistic statistical methods, thereby effectively focusing on areas with high probability of passenger distribution and excluding invalid or interfering areas such as windows and driver's areas, providing a reliable data foundation for post-target recognition.

[0020] 4. In one example, boundary expansion processing of the region of interest can effectively cover situations where the passenger target part exceeds the region of interest due to passenger leaning, posture changes, etc., ensuring the integrity of the region of interest for valid targets and further improving the accuracy and reliability of target detection.

[0021] 5. In one example, the average brightness difference between the current image's region of interest and the region of non-interest is compared with the standard average brightness difference obtained from historical data to determine whether the current image's region of interest needs brightness adjustment, thereby improving the image quality of the region of interest and enhancing the accuracy of target recognition in complex environments (low light or overexposure environments, etc.).

[0022] 6. In one example, comparing the actual number of passengers with the corresponding ticket sales data can promptly identify abnormal ticketing situations and improve the effectiveness of ticketing management. Attached Figure Description

[0023] The specific embodiments of the present invention will be further described in detail below with reference to the accompanying drawings. The accompanying drawings are provided to provide a further understanding of the present application and constitute a part of the present application. The same reference numerals are used in these drawings to denote the same or similar parts. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation of the present application.

[0024] Figure 1 A flowchart illustrating a method provided as an example of the present invention; Figure 2 This is a schematic diagram illustrating data transmission between a server and an in-vehicle terminal, as provided in an example of the present invention. Figure 3 This is a system architecture diagram provided as an example of the present invention. Detailed Implementation

[0025] The technical solution 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.

[0026] In the description of this invention, ordinal numbers (e.g., "first and second," etc.) are used to distinguish objects and are not limited to this order, nor should they be construed as indicating or implying relative importance. Furthermore, the technical features involved in the different embodiments of the invention described below can be combined with each other as long as they do not conflict with each other.

[0027] In one example, such as Figure 1 As shown, a method for calculating passenger flow in the passenger transport industry, with the server as the executing entity, includes the following steps: S1: Obtain the vehicle's real-time speed; In step S1, the server obtains the vehicle's real-time speed by receiving real-time speed information sent by a mobile terminal inside the vehicle, such as an in-vehicle terminal. Specifically, the in-vehicle terminal integrates a speed detector. Preferably, in this example, the in-vehicle terminal integrates a Global Positioning System (GPS) module, which collects the vehicle's real-time speed and uploads it to the server.

[0028] S2: Determine whether the first capture trigger condition is met. If it is met, send a capture control command to the vehicle terminal to enable the vehicle terminal to collect panoramic images covering the entire area inside the vehicle and upload them back to the server. The first capture trigger condition is: within a preset time period, the real-time speed is first less than a first speed threshold and then greater than a second speed threshold. The first speed threshold corresponds to the maximum speed at which passengers are allowed to get on and off the vehicle, and the second speed threshold corresponds to the minimum speed at which passengers are not allowed to get on or off the vehicle. The preset time period is determined based on the time interval between a vehicle completing one passenger pick-up and drop-off. In step S2, the server determines whether a passenger pick-up or drop-off event has occurred based on the vehicle's real-time speed changes. Specifically, the server uses a first speed threshold to determine whether the vehicle is moving at a low speed or stopped. If the real-time speed is less than the first speed threshold (0-20 km / h), the vehicle is either moving at a low speed (e.g., 1 km / h) or stopped. Further, the server uses a second speed threshold (above 20 km / h) to determine whether the vehicle has entered a stable driving state. If the real-time speed is greater than the second speed threshold (e.g., 30 km / h), it indicates that the vehicle has entered a stable driving state. Thus, if the vehicle's real-time speed is first less than the first speed threshold and then greater than the second speed threshold, the process of stopping to pick up passengers, leaving, and entering a stable driving state is completed, i.e., a complete passenger pick-up or drop-off event is completed. Of course, a complete passenger pick-up or drop-off event must be completed within a preset time period (1 min-60 min), such as within 20 minutes. If the time is too long, such as 3 hours, the first capture trigger condition is not met. When the server determines that passenger pick-up or drop-off has occurred based on the vehicle's real-time speed changes, it sends a capture control command to the onboard terminal, which in turn triggers the onboard camera to capture a panoramic image of the vehicle's interior. This panoramic image covers all areas inside the vehicle and can comprehensively reflect the passengers' positions.

[0029] This example uses real-time speed as the trigger condition for passenger flow detection, thus masking the impact of passenger movement or boarding / alighting on passenger flow statistics when the vehicle is moving at low speed or stopped, thereby improving the accuracy of passenger flow statistics.

[0030] Optionally, the server supports timed triggering, such as 5 minutes after the vehicle departs or 5 minutes before it arrives at the station, which are peak times for passenger pick-up and drop-off. Of course, it can also be timed to collect panoramic images of the vehicle interior during the journey, which can promptly count the number of passengers inside the vehicle.

[0031] S3: Receive panoramic images, perform target recognition processing on the panoramic images, and count the actual number of passengers.

[0032] In step S3, the server integrates image recognition models, such as YOLO, SSD, Faster R-CNN, or their variants, deep convolutional neural network models. Optionally, YOLOv5s is trained on a large in-vehicle passenger flow dataset to identify passenger targets. By iteratively training on new datasets, a more accurate passenger target recognition model is formed, ensuring both recognition accuracy and real-time processing speed requirements.

[0033] In one example, the first capture trigger condition is replaced with a second capture trigger condition, which is: within a preset time period, the real-time speed is first less than a first speed threshold and then greater than a second speed threshold, and the real-time location of the vehicle enters the electronic fence.

[0034] In this example, the server determines whether passenger pick-up or drop-off events have occurred based on the vehicle's real-time speed and location changes. Specifically, the geofence is a key monitoring area for passenger pick-up and drop-off outside of station sites, such as highway entrances. Generally, once vehicles enter the highway, passengers will not pick up or drop off, which helps in accurately counting passenger flow. This example combines the vehicle's real-time speed with the spatial constraint of the geofence, setting the image acquisition trigger time in the key monitored area outside the station, achieving refined passenger flow monitoring and improving the accuracy of passenger flow statistics and the effectiveness of management.

[0035] In one example, after determining whether the first or second capture trigger condition is met, the process also includes: If the first or second capture trigger condition is met, capture control commands are sent to the vehicle terminal multiple times, causing the vehicle terminal to collect panoramic images covering the entire area inside the vehicle multiple times in a time-sharing manner and upload them back to the server.

[0036] This example uses a single trigger condition to collect images multiple times over a period of time, such as collecting multiple panoramic images of the vehicle interior within one minute. This allows for the acquisition of panoramic images at multiple time points within the same monitoring scenario. By comprehensively analyzing the recognition results of multiple panoramic images, such as calculating the maximum, average, and minimum passenger flow values ​​for subsequent manual verification, the random errors that may exist in a single frame image (such as momentary occlusion) can be reduced, thereby outputting more stable and reliable passenger flow statistics results.

[0037] In one example, a region of interest lookup step is included before performing target recognition processing on the panoramic image: 1) Based on the annotation information of passengers and seats in the current historical panoramic image of the vehicle, calculate the probability density distribution map of passengers and seats in the image.

[0038] The original image is divided into a specific NxM grid, where N and M are the number of rows and columns, respectively. Depending on the resolution of the original image, a grid of 50x50 or 30x30 pixels is typically used. Based on a large amount of original labeled data and subsequent actual operational data, passenger and vehicle seat targets are statistically analyzed. Specifically, generating the probability density distribution map involves: first, extracting and normalizing the center point coordinates of all labeled target bounding boxes in the historical panoramic image dataset; then, using the Gaussian kernel density estimation method to perform two-dimensional density estimation on the normalized coordinate point set; finally, discretizing the estimated continuous density function to generate a probability density matrix corresponding to the image size, and visualizing this probability density matrix to generate a probability density distribution map indicating the possible regions where targets may appear in the image.

[0039] 2) Based on the probability density distribution map, the current panoramic image is processed into a grid, the probability density value of each grid area is calculated, and the grid areas where the probability density of passengers and seats is higher than the preset probability density threshold are identified as areas of interest.

[0040] Specifically, the current panoramic image is gridded, preferably using the same grid size as when generating the probability density distribution map. The probability density value of each grid region is determined based on the probability density distribution map, representing the probability of the target appearing in each grid region. Then, a probability threshold is set, and each grid is binarized for judgment. Grids with probability density values ​​higher than the probability threshold are marked as 1, indicating a high probability area for passengers or seats, and are retained; grids with probability density values ​​lower than the probability threshold are marked as 0, indicating irrelevant background, and are discarded. Finally, all retained regions are the regions of interest.

[0041] Preferably, the region of interest is expanded to obtain the final region of interest. Since passengers may lean against something, the region of interest is safely expanded outward, for example, by 10% in the x-axis direction. This expansion effectively covers situations where the passenger target portion extends beyond the region of interest due to leaning, posture changes, etc., ensuring the integrity of the region of interest for the effective target and further improving the accuracy and reliability of target detection.

[0042] In one example, due to the influence of lighting, there is often a significant difference in brightness between the interior and exterior of the vehicle. Since cameras typically expose for the entire image before taking the picture, exposure processing is needed to enhance the target features and improve the recognition rate. Therefore, this example, after identifying the region of interest, also includes adjusting the image exposure: 1) Calculate the first brightness difference between the average brightness of the passenger area and the average brightness of the background area in the current historical panoramic image of the vehicle.

[0043] Using the labeled passenger flow dataset, the average brightness a(i) of multiple labeled passenger target boxes is calculated. Since the cameras are all horizontally placed, the image resolution is width x height, where the width direction is horizontal. The data is segmented according to the horizontal direction, and the segment size is 50 pixels. The average brightness b(i) of different segmented regions is calculated. The mean values ​​a1 and b1 of a(i) and b(i) are calculated. The absolute value c1 of the mean difference between a1 and b1 is calculated to obtain the first brightness difference value.

[0044] 2) Calculate the second brightness difference between the average brightness of the region of interest and the average brightness of the region of non-interest in the current panoramic image.

[0045] Find the region of interest in the current panoramic image, and then determine the regions of interest and non-interest in the original panoramic image. Calculate the average brightness d1 of the non-interest region according to the grid, and the average brightness e1 of the region of interest according to the grid. Calculate the absolute value f1 of the difference between the average values ​​of d1 and e1 to obtain the second brightness difference value.

[0046] 3) Compare the first brightness difference with the second brightness difference, and adjust the overall brightness of the area of ​​interest based on the comparison results.

[0047] Preferably, the relationship between the second brightness difference f1 and the first brightness difference c1 is determined, and the following brightness adjustment strategy is executed: a. When the second brightness difference falls within the preset brightness threshold range centered on the first brightness difference, such as 0.8c1≤f1≤1.2c1, no brightness compensation processing is required; b. When the second brightness difference is greater than the upper limit of the preset brightness threshold range, such as f1 > 1.2c1, the brightness of the area of ​​interest is increased, and the brightness adjustment value is δ = f1 - 1.2c1.

[0048] c. When the second brightness difference is less than the lower limit of the preset brightness threshold range, such as f1 < 0.8c1, reduce the brightness of the area of ​​interest, and the brightness adjustment value is δ = c1 - f1.

[0049] In one example, after counting the actual number of passengers, the following is also included: By comparing the actual number of passengers with the ticket sales data for the corresponding vehicle trip, and if the actual number of passengers does not match the ticket sales data, a ticketing anomaly alert is generated to promptly detect abnormal ticketing situations and improve the effectiveness of ticketing management.

[0050] Combining the above examples yields a preferred example of the present invention, in which the method includes the following steps: S1': Obtain the vehicle's real-time speed and real-time location; S2': Determine whether the second capture trigger condition is met. If it is met, send a capture control command to the vehicle terminal so that the vehicle terminal can collect panoramic images covering the entire area inside the vehicle multiple times in time and upload them back to the server. S3': Receive the panoramic image, locate the region of interest in the panoramic image, adjust the exposure of the region of interest, then perform target recognition processing, and count the actual number of passengers.

[0051] In one example, a passenger flow calculation method for the passenger transport industry, executed by an on-board terminal, includes the following steps: S10: Collect the real-time speed of the vehicle and upload it to the server; S20: Receive the capture control command sent by the server when the real-time speed determines that the first capture trigger condition is met. The first capture trigger condition is: within a preset time period, the real-time speed is first less than a first speed threshold and then greater than a second speed threshold. The first speed threshold corresponds to the maximum speed of the vehicle that allows passengers to get on and off, and the second speed threshold corresponds to the minimum speed of the vehicle that prevents passengers from getting on and off. The preset time period is determined based on the time interval between the vehicle completing one passenger pick-up and drop-off. S30: Collects panoramic images covering the entire interior of the vehicle and uploads them to the server, where the server performs target recognition processing on the panoramic images and counts the actual number of passengers.

[0052] Preferably, the method further includes: S10': Collect the vehicle's real-time speed and location and upload them to the server; S20': Receive the capture control command sent by the server when the second capture trigger condition is met based on the vehicle's real-time speed and real-time position, collect panoramic images covering the entire area inside the vehicle, and upload them to the server; S30': The second capture trigger condition is: within a preset time period, the real-time speed is first less than the first speed threshold and then greater than the second speed threshold, and the real-time position of the vehicle enters the electronic fence; wherein, the electronic fence is the area outside the station.

[0053] Combining the above examples of passenger flow calculation methods in the passenger transport industry, which respectively use servers and vehicle-mounted terminals as the executing entities, yields a preferred embodiment of the present invention. In this case, the data transmission relationship between the server and the terminal is as follows: Figure 2 As shown, the method includes the following steps: S100: Configure the server's capture and passenger flow detection trigger process, including timed triggering, interval triggering after startup, etc., and enter S200; at the same time, the GPS module integrated on the vehicle collects the vehicle's real-time speed and real-time location. S200: Snapshot control: When the time reaches the trigger condition, the server determines whether the vehicle speed exceeds the speed threshold based on the vehicle's real-time speed (first less than the first speed threshold, then greater than the second speed threshold). If the condition is met, proceed to S300; otherwise, terminate the process. S300: Capture Control: The server determines whether the vehicle is equipped with a detection geofence based on its real-time location. If so, it proceeds to S400; otherwise, it proceeds to S500. S400: Capture Control: The server determines whether the current vehicle's GPS latitude and longitude coordinates are within the bound electronic fence. If so, it proceeds to S500; otherwise, the process terminates. S500: The server sends a photo control command to the vehicle terminal, which then enters S600; S600: The vehicle terminal receives the photo control command, triggers the photo process, and the vehicle camera collects panoramic images of the vehicle interior in real time, then enters S700. S700: Capture Control: The server receives panoramic images and stores them in the storage service, then enters S800; S800: Passenger Recognition: Use the target recognition model to perform target recognition on the panoramic image, then proceed to S900; S900: Passenger Flow Statistics: Performs statistics on the targets identified by S800, compares them with the accessed ticket data, compares and statistically analyzes the number of tickets sold and the actual passenger flow data, and then proceeds to S1000; S1000: End.

[0054] This invention also includes a passenger flow calculation system for the passenger transport industry, such as... Figure 3 As shown, the system includes an in-vehicle terminal and a server (data center server), which are connected via a mobile communication module such as a 4G network. The in-vehicle terminal responds to and executes photo capture control commands from the data center server, and includes an in-vehicle camera and a GPS module. Preferably, the control module of the in-vehicle terminal is wirelessly connected to the in-vehicle camera and GPS module. The GPS module collects the vehicle's real-time speed and uploads it to the server via the in-vehicle terminal, or collects the vehicle's real-time speed and location and uploads it to the server via the in-vehicle terminal, providing vehicle status data support for the capture control service. Optionally, the GPS module can also directly transmit the collected real-time speed and location to the server. The in-vehicle terminal receives capture control commands from the server, triggering the in-vehicle camera to collect panoramic images covering the entire area inside the vehicle and upload them back to the server via the in-vehicle terminal. Optionally, the in-vehicle camera can also directly transmit the collected panoramic images to the server. The in-vehicle camera is installed in different locations depending on the vehicle model to capture real-time images of the vehicle.

[0055] The server comprises a visual management and configuration unit, a GPS data integration unit, a data storage unit, an image capture control unit, a passenger flow recognition and statistics unit, a dataset management unit, and a model training unit. The visual management and configuration unit integrates a visual web service, enabling system configuration and real-time passenger flow data visualization. The GPS data integration unit stores real-time speed and / or location information of vehicles received by the server. The data storage unit stores panoramic images uploaded by the vehicle-mounted terminal. The server determines whether the first or second capture trigger condition is met; if so, the image capture control unit, based on a 4G network module and a custom TCP control protocol, transmits capture control commands from the server to the vehicle-mounted terminal. The passenger flow recognition and statistics unit, based on a target recognition model trained using a deep neural network, performs passenger target recognition processing on the panoramic images and counts the actual number of passengers. The passenger flow recognition and statistics unit also compares the ticket sales data with the counted actual passenger numbers to determine the accuracy of real-time passenger flow; if they are inconsistent, it generates a ticketing anomaly alert. Furthermore, the dataset management unit, based on data storage services, provides data annotation services to establish a vehicle passenger target dataset. The passenger target model training unit uses a passenger flow dataset and is based on a deep convolutional neural network to train passenger targets. By continuously iterating and training on new datasets, a more accurate target recognition model is formed.

[0056] The system's in-vehicle terminal features flexible camera installation and deployment, supporting various vehicle types. By adjusting the camera angle and position, it achieves effective vehicle coverage and passenger flow identification. A target recognition model performs target identification on panoramic images, achieving a high passenger recognition rate. The model is scalable and can be iterated upon based on actual panoramic image datasets. By identifying areas of interest, specific locations within the vehicle can be masked, eliminating the impact of targets in those areas on passenger flow statistics, such as the driver's position and areas outside the windows. GPS speed limits are used for image capture and passenger flow identification. Since passengers and drivers are typically seated while the vehicle is running, this avoids the impact of vehicle stops for passengers getting on and off on passenger flow statistics, and also prevents passenger transport management personnel from getting on and off the vehicle, thus avoiding interference with identification. GPS latitude and longitude coordinates and electronic fences further enhance passenger flow detection at specific vehicle locations, increasing the dimensions and scope of operational supervision.

[0057] The above detailed embodiments are a description of the present invention. It should not be considered that the specific embodiments of the present invention are limited to these descriptions. For those skilled in the art, several simple deductions and substitutions can be made without departing from the concept of the present invention, and all of these should be considered to fall within the protection scope of the present invention.

Claims

1. A method for calculating passenger flow in the passenger transport industry, characterized in that, The execution entity is the server, and it includes the following steps: Obtain the vehicle's real-time speed; Determine whether the first capture trigger condition is met. If it is, send a capture control command to the vehicle terminal to enable the vehicle terminal to collect a panoramic image covering the entire area inside the vehicle and upload it back to the server. The first capture trigger condition is: within a preset time period, the real-time speed is first less than a first speed threshold and then greater than a second speed threshold. The first speed threshold corresponds to the maximum speed at which passengers are allowed to get on and off the vehicle, and the second speed threshold corresponds to the minimum speed at which passengers are not allowed to get on or off the vehicle. The preset time period is determined based on the time interval between one passenger pick-up and drop-off. Receive panoramic images, perform target recognition processing on the panoramic images, and count the actual number of passengers.

2. The passenger flow calculation method for the passenger transport industry according to claim 1, characterized in that, The first capture trigger condition is replaced with the second capture trigger condition, which is: within a preset time period, the real-time speed is first less than the first speed threshold and then greater than the second speed threshold, and the real-time position of the vehicle enters the electronic fence; where the electronic fence is the area outside the station.

3. The method for calculating passenger flow in the passenger transport industry according to claim 1 or 2, characterized in that, Before performing target recognition processing on the panoramic image, the method further includes: Based on the annotation information of passengers and seats in the current historical panoramic image of the vehicle, calculate the probability density distribution map of passengers and seats in the image; The current panoramic image is gridded based on the probability density distribution map, and the probability density value of each grid region is calculated. Grid regions where the probability density of passengers and seats is higher than the preset probability density threshold are identified as regions of interest.

4. The passenger flow calculation method for the passenger transport industry according to claim 3, characterized in that, After the region of interest is determined, the following steps are also included: The region of interest is expanded by boundary extension to obtain the final region of interest.

5. The passenger flow calculation method for the passenger transport industry according to claim 4, characterized in that, After the region of interest is determined, the following steps are also included: Calculate the first brightness difference between the average brightness of the passenger area and the average brightness of the background area in the current historical panoramic image of the vehicle. Calculate the second brightness difference between the average brightness of the region of interest and the average brightness of the non-region of interest in the current panoramic image; The first brightness difference and the second brightness difference are compared and processed, and the overall brightness of the area of ​​interest is compensated and adjusted based on the comparison result.

6. The passenger flow calculation method for the passenger transport industry according to claim 5, characterized in that, The step of compensating and adjusting the overall brightness of the area of ​​interest based on the comparison results includes: When the second brightness difference falls within the preset brightness threshold range centered on the first brightness difference, no brightness compensation processing is required. When the second brightness difference is greater than the upper limit of the preset brightness threshold range, the brightness of the area of ​​interest is increased; When the second brightness difference is less than the lower limit of the preset brightness threshold range, the brightness of the area of ​​interest is reduced.

7. The method for calculating passenger flow in the passenger transport industry according to claim 1, characterized in that, After calculating the actual number of passengers, the following is also included: The system compares the actual number of passengers with the ticket sales data for the corresponding vehicle trip. If the actual number of passengers does not match the ticket sales data, a ticketing error message is generated.

8. A method for calculating passenger flow in the passenger transport industry, characterized in that, The execution entity is the vehicle-mounted terminal, and the process includes the following steps: Collect the real-time speed of the vehicle and upload it to the server; The receiving server sends a capture control command when the first capture trigger condition is met based on the real-time speed. The first capture trigger condition is: within a preset time period, the real-time speed is first less than a first speed threshold and then greater than a second speed threshold. The first speed threshold corresponds to the maximum speed at which passengers are allowed to get on and off the vehicle, and the second speed threshold corresponds to the minimum speed at which passengers are not allowed to get on or off the vehicle. The preset time period is determined based on the time interval between when a vehicle completes one passenger pick-up and drop-off. The system collects panoramic images covering the entire interior of the vehicle and uploads them to a server. The server then performs target recognition processing on the panoramic images and counts the actual number of passengers.

9. The passenger flow calculation method for the passenger transport industry according to claim 8, characterized in that, The method further includes: Collect the real-time location of the vehicle and upload it to the server; The receiving server receives a capture control command when the second capture trigger condition is met based on the vehicle's real-time speed and real-time location. It then collects panoramic images covering the entire area inside the vehicle and uploads them to the server. The second capture trigger condition is: within a preset time period, the real-time speed is first less than the first speed threshold and then greater than the second speed threshold, and the real-time position of the vehicle enters the electronic fence; whereby the electronic fence is the area outside the station.

10. A passenger flow calculation system for the passenger transport industry, characterized in that, Including vehicle-mounted terminals and servers; The vehicle-mounted terminal is used to collect the real-time speed of the vehicle and upload it to the server, or to collect the real-time speed and real-time location of the vehicle and upload it to the server. The server is used to receive the vehicle's real-time speed, or to receive the vehicle's real-time speed and real-time location. The server is also used to determine whether the first or second capture trigger condition is met. If met, a capture control command is sent to the vehicle terminal. The first capture trigger condition is: within a preset time period, the real-time speed is first less than a first speed threshold and then greater than a second speed threshold. The first speed threshold corresponds to the maximum speed at which passengers are allowed to get on and off the vehicle, and the second speed threshold corresponds to the minimum speed at which passengers are not allowed to get on or off the vehicle. The preset time period is determined based on the time interval between a vehicle completing one passenger pick-up and drop-off. The second capture trigger condition is: within a preset time period, the real-time speed is first less than the first speed threshold and then greater than the second speed threshold, and the real-time location of the vehicle enters the electronic fence. The electronic fence is the area outside the station. The vehicle-mounted terminal is also used to receive capture control commands sent by the server, trigger the acquisition of panoramic images covering the entire area inside the vehicle, and upload them back to the server. The server is also used to receive panoramic images, perform target recognition processing on the panoramic images, and count the actual number of passengers.

Images