Rail Transit Fare Clearing Method Based on Deep Learning

Abstract

The present invention provides a rail transit fare clearing method based on deep learning. The method includes: first, obtaining the video information of the inbound passengers, binding the optimal face frame and the optimal body frame of the passenger after extracting them from the video information, then obtaining the face and body features of the passenger, narrowing down the search database to improve the recognition accuracy, and finally comprehensively evaluating by combining the rail transit train operation time data to accurately obtain the travel trajectory of the passenger. Through the method of the present invention, the accuracy of the passenger travel trajectory can be effectively improved, and at the same time, a query request can be initiated at any time, with stronger timeliness.

Description

Technical Field

[0001] The present disclosure relates to the technical field of rail transit monitoring, and particularly to a rail transit fare clearing method based on deep learning. Background Art

[0002] In current urban rail transit, the requirements for networked operation of urban rail transit, the need for convenient passenger travel transfers, and the need for revenue distribution among different operation entities determine that under the mode of barrier-free transfer, the urban rail transit fare clearing center (ACC) should provide accurate and timely ticket clearing services for all rail transit lines and operators in the city.

[0003] Passenger OD path: In the rail transit industry, the OD path refers to all transfer paths of a passenger starting from entry station A, passing through N (N = 0, 1, 2,...) transfers, passing through N transfer stations, and finally exiting from exit station B.

[0004] Existing fare clearing systems: Through the arrival schedules of trains on all lines, the entry time and entry location of each passenger, the exit time and exit location, and at the same time estimating the walking speed of passengers during transfers, predicting among several possible paths that a passenger may reach, and predicting which train on which line a passenger is most likely to catch based on the train arrival time, so as to obtain the passenger's OD path.

[0005] The existing methods for predicting passenger OD paths mainly have the following problems: 1. Strong correlation with the arrival schedule of passengers. When the schedule deviates, the prediction result changes accordingly; 2. Lack of real-time performance. It is impossible to know in real time the OD path of a certain passenger and which station the passenger is currently at; 3. Strong correlation with the walking speed of passengers in transfer stations. When there are behaviors such as passengers lingering or going to the toilet, there will be errors when predicting which train a passenger is likely to catch, thereby causing a decrease in the prediction accuracy of the passenger OD path. Summary of the Invention

[0006] In view of this, the embodiments of the present disclosure provide a rail transit fare clearing method based on deep learning. This method is based on computer vision, uses the monitoring cameras already deployed in rail transit to capture and identify the faces and bodies of passengers, compare features, and match personnel, and then obtain the OD path of each passenger.

[0007] To achieve the above object, the present invention provides the following technical solutions:

[0008] A rail transit fare clearing method based on deep learning, comprising the following steps:

[0009] S1. Obtain data sources: Collect video information of inbound passengers;

[0010] S2. Optimal face frame and optimal body frame extraction: Filter out the face frame that best matches the face features of the passenger and the body frame that best matches the body features from the passenger's video information; then, according to the matching algorithm, bind the optimal face frame and the optimal body frame of the same passenger.

[0011] S3. Face and body feature extraction: Perform affine transformation on the optimal face frame and the optimal body frame with the standard face and the standard body respectively to extract face and body features.

[0012] S4. Narrow down the database: Query and match the face and body features of the passenger from the transfer channels at the transfer stations on the line where the passenger gets off.

[0013] If the face and body features of the passenger are not found in the transfer channels at the transfer stations on this line, it means the passenger did not transfer.

[0014] If the face and body features of the passenger are found in the transfer channels at a transfer station on this line, match to the current line where the transfer station is located, and continue to query the face and body features of the passenger in the transfer channels at the transfer stations on the current line. Repeat the query and matching until the passenger did not transfer, match to the boarding station of the passenger, and obtain N kinds of riding path results of the passenger.

[0015] S5. Obtain the OD path of the passenger: Combine the N kinds of riding path results of the passenger with the train operation time, and based on time, exclude the paths that the passenger could not have taken, and comprehensively evaluate to obtain the OD path of the passenger.

[0016] In a preferred embodiment, the video information of the passenger is obtained by video acquisition devices arranged along the track.

[0017] In a preferred embodiment, the video acquisition device includes a network camera or an intelligent camera.

[0018] In a preferred embodiment, from the passenger's video information, the face frame that best matches the face features of the passenger is filtered out from multiple dimensions including face angle, face blur, face brightness, and whether the face is blocked.

[0019] In a preferred embodiment, from the passenger's video information, the body frame that best matches the body features of the passenger is filtered out from multiple dimensions including body integrity, body angle, and whether the body is blocked.

[0020] In a preferred embodiment, the matching algorithm includes IOU matching logic.

[0021] In a preferred embodiment, it further includes selecting the N ride path results with the highest similarity from the multiple ride path results of the passenger, and performing a comprehensive evaluation after combining them with the train operation schedule.

[0022] The rail transit fare clearing method of the present invention has the beneficial effects that: based on deep learning, the computer vision method is applied to the rail transit fare clearing system. First, the monitoring cameras already deployed in the rail transit are used to capture and identify the faces and bodies of passengers, compare features, and match personnel to obtain a database, and then the search database is narrowed down to improve the recognition accuracy. The present invention combines the rail transit train operation schedule data and the computer vision solution, makes a comprehensive judgment, accurately obtains the travel trajectory of the passenger, and can display the captured passenger pictures in real time, so that the passenger trajectory can be clearly traced back, and thus an intuitive judgment can be made on the accuracy of the prediction result. At the same time, a query request can be initiated at any time, which is more timely than the traditional fare clearing system in terms of timeliness. Brief Description of the Drawings

[0023] In order to more clearly illustrate the technical solutions of the embodiments of the present disclosure, the following will briefly introduce the drawings required in the embodiments. Obviously, the drawings in the following description are only some embodiments of the present disclosure. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0024] Figure 1 It is the flowchart of the rail transit fare clearing method of the present invention;

[0025] Figure 2 It is the schematic diagram of the calculation method of IOU in the present invention. Detailed Embodiments

[0026] The following will describe the embodiments of the present disclosure in detail with reference to the drawings.

[0027] The following uses specific specific examples to illustrate the embodiments of the present disclosure. Those skilled in the art can easily understand other advantages and effects of the present disclosure from the content disclosed in this specification. Obviously, the described embodiments are only a part of the embodiments of the present disclosure, rather than all the embodiments. The present disclosure can also be implemented or applied through other different specific embodiments, and various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present disclosure. It should be noted that, without conflict, the following embodiments and the features in the embodiments can be combined with each other. All other embodiments obtained by those of ordinary skill in the art based on the embodiments in the present disclosure without creative efforts belong to the scope of protection of the present disclosure.

[0028] It should be noted that the following description relates to various aspects of embodiments within the scope of the appended claims. It will be apparent that the aspects described herein can be embodied in a wide variety of forms, and any specific structure and / or function described herein is for illustrative purposes only. Based on this disclosure, those skilled in the art should understand that one aspect described herein can be implemented independently of any other aspect, and two or more of these aspects can be combined in various ways. For example, any number of aspects described herein can be used to implement a device and / or practice a method. Additionally, this device can be implemented and this method can be practiced using other structures and / or functionality in addition to one or more of the aspects described herein. In the technical solutions of this disclosure, the processing of the collection, storage, use, processing, transmission, provision, and disclosure of passenger personal information complies with the provisions of relevant laws and regulations and does not violate public order and good customs.

[0029] It should also be noted that the illustrations provided in the following embodiments only schematically illustrate the basic concept of this disclosure. The figures only show the components related to this disclosure and are not drawn according to the number, shape, and size of the components in actual implementation. The type, quantity, and proportion of each component in actual implementation can be arbitrarily changed, and the component layout type may also be more complex.

[0030] In addition, in the following description, specific details are provided to facilitate a thorough understanding of the examples. However, those skilled in the art will understand that the aspects can be practiced without these specific details.

[0031] As Figure 1 shown, an embodiment of this disclosure provides a rail transit clearing method based on deep learning, including the following steps:

[0032] S1. Obtain data sources: Collect video information of inbound passengers;

[0033] Obtain the video information of passengers through video acquisition devices installed along the track. Specifically, the above video acquisition devices include network cameras or intelligent cameras.

[0034] S2. Extract the optimal face frame and the optimal body frame: From the video information of passengers, select the face frame that best matches the face feature extraction of the passenger from multiple dimensions such as face angle, face blur, face brightness, and whether the face is blocked; select the body frame that best matches the body feature extraction of the passenger from multiple dimensions such as body integrity, body angle, and whether the body is blocked, and then bind the optimal face frame and the optimal body frame of the same passenger according to matching logics such as IOU;

[0035] Among them, the calculation method of the above IOU is as Figure 2As shown in the figure, IOU represents the ratio of the intersection area to the union area between two rectangular frames in the figure.

[0036] S3. Facial and human body feature extraction: Perform affine transformation on the optimal face frame and the optimal human body frame respectively with the standard face and the standard human body, and perform facial and human body feature extraction;

[0037] Among them, the above-mentioned affine transformation has become a commonly used algorithm in the field of face recognition, and the specific introduction is as follows: Affine transformation: Taking an example of a face affine transformation, the standard face is an upright face image facing forward. Using the face key point algorithm, the key points of the standard face (such as the commonly used 5 points, 68 points, 136 points, etc.) can be obtained, and the key points of the captured optimal face can also be obtained. Through the two groups of key points of the captured optimal face and the standard face, the homography matrix between the two can be obtained. After affine transformation, the captured optimal face is translated, scaled, and rotated into the standard face space, and the captured face can be turned into an upright face looking straight ahead, which can be used to improve the accuracy of face recognition.

[0038] S4. Narrow down the database: Query the facial and human body features of the passenger that match in the transfer channels of the transfer stations on the line where the passenger gets off the station;

[0039] If the facial and human body features of the passenger are not queried in the transfer channels of the transfer stations on this line, it means that the passenger has not transferred;

[0040] If the facial and human body features of the passenger are queried in the transfer channel of a transfer station on this line, it is matched to the current line where the transfer station is located, and continue to query the facial and human body features of the passenger in the transfer channels of the transfer stations on the current line. Repeat the query and matching until the passenger has not transferred, and match to the inbound station of the passenger to obtain N kinds of ride path results of the passenger;

[0041] S5. Obtain the OD path of the passenger: Combine the N kinds of ride path results of the passenger with the train operation time. Based on time, exclude the paths that the passenger could not have passed through, and comprehensively evaluate to obtain the OD path of the passenger;

[0042] Due to relying only on face or human body recognition, in the case of a large passenger flow (tens of millions), one face may be successfully matched by multiple faces, so there may be multiple possibilities for the matching routes of a certain outbound passenger, such as A, B, C, etc. From the subway ticketing system, the starting station and the final station of each passenger, as well as the card-in time and card-out time can be obtained. Among the matched routes, select the matched routes that appear within the time interval from the start to the end of the passenger.

[0043] Among them, the N kinds of ride path results of the passenger are the N results with the highest similarity among the multiple ride paths of the passenger that are matched.

[0044] Embodiment 1

[0045] A certain passenger swipes the card to enter the station at Station A, passes through three transfer stations B, C, and D, and finally leaves the subway at Station E. Through the subway ticketing system, it can be known that the passenger enters and exits the station. That is, in the OD path of this passenger, A and E are known. However, due to the criss-crossing road network, from A to E, the transfer stations that can be selected are (B1, B2...), (C1, C2...), (D1, D2...). Since there are face capture cameras evenly distributed at the entrance, transfer channels, and exit of this system, the passenger is captured by these capture cameras along the way, and the OD path of this passenger will be formed. However, due to the large passenger flow in the entire road network system, there will be multiple OD paths exceeding the threshold, such as Path 1: A - B - C - D - E (this is the accurate path), Path 2: A - B1 - C2 - D1 - E, Path 3: A - B2 - C - D2 - E, etc. Then, according to the train arrival schedule, it is judged that the passenger cannot take the inaccurate paths such as Path 2 and Path 3, and finally the only reliable riding path 1 is selected.

[0046] As described above, it is only the specific implementation manner of the present disclosure, but the protection scope of the present disclosure is not limited thereto. Any changes or substitutions that can be easily thought of by those skilled in the art within the technical scope disclosed by the present disclosure should be covered by the protection scope of the present disclosure. Therefore, the protection scope of the present disclosure should be subject to the protection scope of the claims.

Claims

1. A rail transit clearing method based on deep learning, characterized in that, It includes the following steps: S1. Obtain the data source: Collect the video information of inbound passengers; S2. Extract the optimal face frame and optimal body frame: Screen out the face frame that best matches the face feature extraction and the body frame that best matches the body feature extraction from the video information of the passengers; then, according to the matching algorithm, bind the optimal face frame and optimal body frame of the same passenger; S3. Extract face and body features: Perform affine transformation on the optimal face frame and optimal body frame with the standard face and standard body respectively to extract face and body features; S4. Narrow down the database: Query and match the face and body features of the passenger in the transfer channels of the transfer stations on the line where the passenger gets off; If the face and body features of the passenger are not found in the transfer channels of the transfer stations on this line, it means that the passenger has not transferred; If the face and body features of the passenger are found in the transfer channel of a transfer station on this line, match to the current line where the transfer station is located, and continue to query the face and body features of the passenger in the transfer channels of the transfer stations on the current line. Repeat the query and matching until the passenger has not transferred and the inbound station of the passenger is matched to obtain multiple riding path results of the passenger; S5. Obtain the OD path of the passenger: Combine the multiple riding path results of the passenger with the train operation time, and based on time, exclude the paths that the passenger could not have passed through, and comprehensively evaluate to obtain the OD path of the passenger.

2. The rail transit fare clearing method according to claim 1, wherein Obtain the video information of the passenger through the video acquisition devices set along the track.

3. The rail transit fare clearing method according to claim 2, wherein The video acquisition device includes a network camera or an intelligent camera.

4. The rail transit fare clearing method according to claim 1, wherein From the video information of the passengers, screen out the face frame that best matches the face feature extraction of the passenger from multiple dimensions including face angle, face blur degree, face brightness, and whether the face is blocked.

5. The rail transit fare clearing method according to claim 1, wherein From the video information of the passengers, screen out the body frame that best matches the body feature extraction of the passenger from multiple dimensions including body integrity, body angle, and whether the body is blocked.

6. The rail transit fare clearing method according to claim 1, characterized in that The matching algorithm includes the IOU matching logic.

7. The rail transit fare clearing method according to claim 1, wherein It also includes taking the N riding path results with the highest similarity in the multiple riding path results of the passenger, and combining them with the train operation time for comprehensive evaluation.

Images