Vehicle overload identification method and device and computer equipment

Abstract

The invention discloses a vehicle overload identification method and device and computer equipment. The method comprises the following steps: acquiring a single-frame image of a to-be-identified vehicle in video stream data, and performing target detection on the single-frame image of the to-be-identified vehicle to obtain a target detection result: pairing and binding a human body part and a human head in the to-be-identified vehicle with the to-be-identified vehicle to obtain a first identification list of passengers in the to-be-identified vehicle, determining an overload mass score of the to-be-identified vehicle according to a first identifier list, wherein the first identifier list comprises a human body identifier bound with the to-be-identified vehicle; acquiring time information of a to-be-identified vehicle, a positive binding feature of the to-be-identified vehicle and a manned feature of the to-be-identified vehicle; and determining an identification result of the to-be-identified vehicle based on the time information of the to-be-identified vehicle, the overload mass score of the to-be-identified vehicle, the positive binding feature of the to-be-identified vehicle and the manned feature of the to-be-identified vehicle, wherein the identification result is used for representing whether the to-be-identified vehicle is overloaded or not.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and more specifically, to a vehicle overload identification method, device and computer equipment. Background Art

[0002] Overloading of electric vehicles and motorcycles poses a major threat to road traffic safety and personal safety. At present, one method of automatic identification of overloading and overcrowding monitoring is to install various sensors on the vehicle, and perform comprehensive calculations on the detection data and compare them with the approved load to determine whether it is overloaded. This method cannot determine whether the specific number of people is overloaded, and it requires the installation of equipment and devices, which is costly. One method is to use image recognition methods, target detection and tracking based on the data captured by the surveillance camera to determine whether the number of detected personnel exceeds the specified number, and then determine whether it is overloaded or overcrowded. For example: a load weighing module is installed to weigh the vehicle to obtain overload label information, and a radar vision integrated machine is installed to perform data analysis to obtain vehicle driving behavior. Finally, a data set is constructed based on the basic information of the vehicle, overload information, and driving behavior information to train an overload recognition model. This recognition model is used to determine whether the vehicle is overloaded based on the basic information of the vehicle and driving behavior information. This method has no clear output for whether the personnel are overloaded, and the equipment installation and maintenance costs are high, and the data collection costs are high. In another method, the vehicle and human body are detected by the target, and then the number of passengers is determined based on the bbox (bounding box) position information of the two, through the overlapping area ratio threshold and the coordinate inclusion relationship, so as to determine whether the passengers are overloaded. This method is relatively simple to calculate the binding relationship between the human body and the vehicle. For complex backgrounds in actual scenes, such as passers-by near the vehicle, it is easy to misjudge them as passengers; for human body occlusion caused by angles, it is easy to cause the missed judgment of passengers, thereby causing missed reports and false reports of overloading. Similarly, by demarcating the area to detect the head information of the person, and determining the number of passengers based on the number of heads, it is also easy to have missed judgments and false judgments. In another way, it is proposed to use pedestrian re-identification to effectively count the pedestrians appearing in multiple frames on the basis of target detection and tracking, eliminate the interference of irrelevant passers-by, and use the similarity of the bbox position movement of pedestrians in multiple frames to make a secondary judgment on overloading. This method has extremely high requirements for the detection and tracking of electric vehicles and passengers. When electric vehicles or passengers are lost due to different degrees of occlusion in continuous frames or the size of the detection frame changes greatly, it will have a great interference on the re-identification of pedestrians and the similarity of position movement, which is easy to cause misjudgment and missed judgment. Therefore, the existing technology has the following problems: 1. When electric vehicles and motorcycles are carrying passengers, there is a high probability that the passengers are blocked or even lost to different degrees. For example, when an adult holds a child on an electric vehicle or motorcycle, the child's body is easy to be missed, and the human body detection frame changes greatly in continuous frames, the similarity of the re-identification features is low, and the similarity of position movement is greatly disturbed, resulting in discontinuous human body tracking, difficulty in determining the accompanying relationship between people and vehicles, and inability to determine the effective number of passengers. 2. Only using the detection frame and re-identification features cannot effectively eliminate the interference of irrelevant passers-by, resulting in misjudgment of overcrowding and overloading.3. In video stream data, there is a problem of jumps in the number of passengers in a single frame due to vehicle movement, pedestrian obstruction, etc. Most methods use the fusion statistics of the number of passengers in multiple frames to make the final overload judgment. During the fusion, many empirical parameters need to be debugged and optimized, and they cannot be universally adapted. Summary of the invention

[0003] The embodiments of the present application provide a vehicle overload identification method, device and computer equipment to at least solve the technical problem of low accuracy in identifying the number of passengers in a vehicle in the related art.

[0004] According to one aspect of an embodiment of the present application, a method for identifying vehicle overload is provided, comprising: obtaining a single-frame image of a vehicle to be identified in video stream data, and performing target detection on the single-frame image of the vehicle to be identified to obtain a target detection result, wherein the target detection result includes at least: a body part of a human being in the vehicle to be identified, a human head, and the vehicle to be identified; pairing and binding the body part and the human head in the vehicle to be identified with the vehicle to be identified, and identifying the human body bound to the vehicle to be identified to obtain a first identification list of people in the vehicle to be identified, and determining an overload mass score of the vehicle to be identified based on the first identification list, wherein the overload mass score is used to represent the The probability that the vehicle to be identified is overloaded, the first identification list contains human identifications bound to the vehicle to be identified; obtaining the time information of the vehicle to be identified in each frame of the image, the positive binding features of the vehicle to be identified and the passenger features of the vehicle to be identified, wherein the positive binding features are used to indicate the effective number of passengers in the vehicle to be identified, and the passenger features are used to indicate the passenger status of the vehicle to be identified; determining the identification result of the vehicle to be identified based on the time information of the vehicle to be identified in each frame of the image, the overload mass score of the vehicle to be identified, the positive binding features of the vehicle to be identified and the passenger features of the vehicle to be identified, and the identification result is used to indicate whether the vehicle to be identified is overloaded.

[0005] Optionally, obtaining the passenger feature of the vehicle to be identified includes: tracking the vehicle to be identified and the human body identifier that has been bound to the vehicle to be identified in the first identifier list, respectively, to obtain the identifier trajectory of the vehicle to be identified and the human body identifier trajectory bound to the vehicle to be identified, the identifier trajectory being used to represent the movement trajectory of the target in the video; correcting the human body identifier trajectory bound to the vehicle to be identified to obtain a second identifier list, the second identifier list containing the corrected human body identifier trajectory; determining the intersection of the identifier trajectory in the second identifier list and the first identifier list as a third identifier list, the third identifier list containing the human body identifier trajectory in the intersection of the identifier trajectory in the second identifier list and the first identifier list; determining the passenger feature of the vehicle to be identified and the positive binding feature of the vehicle to be identified based on the human body identifier trajectory in the third identifier list.

[0006] Optionally, the human body trajectory information bound to the vehicle to be identified is corrected to obtain a second identification list, including: obtaining the accompanying relationship between the human body and the vehicle to be identified; eliminating the human body identification trajectories that have no accompanying relationship with the vehicle to be identified from the human body identification trajectories bound to the vehicle to be identified, and obtaining the identification trajectories in the second identification list.

[0007] Optionally, obtaining the accompanying relationship between a human body and the vehicle to be identified includes: obtaining a frame number list contained in the identification trajectory of the vehicle to be identified and a frame number list contained in the human body identification trajectory, the frame number list being used to represent image frames in which the vehicle identification or human body identification to be identified appears; obtaining an image frame number list in which a binding relationship is successfully established between the passenger and the vehicle to be identified; determining the frame number list contained in the identification trajectory of the identified vehicle as a first frame number list, determining the frame number list contained in the human body identification trajectory as a second frame number list, and determining the image frame number list in which a binding relationship is successfully established between the passenger and the vehicle to be identified as a third frame number list; determining the intersection of the first frame number list and the second frame number list as a fourth frame number list; performing a difference between the third frame number list and the fourth frame number list to obtain the accompanying relationship between the human body and the vehicle to be identified.

[0008] Optionally, the human body identifier bound to the vehicle to be identified is tracked to obtain the human body identifier trajectory bound to the vehicle to be identified, including: matching the high-confidence detection frame and the low-confidence detection frame in the current image frame with the prediction frame in sequence, and when the match is successful, determining that the human body identifier corresponding to the detection frame belongs to the human body identifier trajectory to be matched; in the case of an unsuccessful matching detection frame, obtaining the human body identifier bound to the unsuccessful matching detection frame, and when the vehicle identifier bound to the unsuccessful matching detection frame is the same as the identifier of the vehicle to be identified and the human body identifier in the vehicle to be identified is unique, determining that the human body identifier corresponding to the unsuccessful matching detection frame belongs to the human body identifier trajectory to be matched; when the vehicle identifier bound to the unsuccessful matching detection frame is the same as the identifier of the vehicle to be identified and the human body identifier in the vehicle to be identified is not unique, matching the human body identifier corresponding to the successfully matched detection frame to the human body identifier trajectory with the most similar pedestrian re-identification feature.

[0009] Optionally, the method further includes: when the vehicle identification bound to the unsuccessfully matched detection frame is different from the identification of the vehicle to be identified, determining that the human body identification corresponding to the unsuccessfully matched detection frame has failed to match.

[0010] Optionally, the human body identification trajectory bound to the vehicle to be identified is corrected to obtain a second identification list, including: obtaining pedestrian re-identification features corresponding to the human body identification trajectory bound to the vehicle to be identified; merging the human body identification trajectory bound to the vehicle to be identified according to the similarity of the pedestrian re-identification features corresponding to the human body identification trajectory bound to the vehicle to be identified to obtain the second identification list.

[0011] Optionally, the passenger characteristics of the vehicle to be identified and the positive binding characteristics of the vehicle to be identified are determined based on the human identification trajectory in the third identification list, including: weighted averaging the human characteristics corresponding to the human identification trajectory in the third identification list and the vehicle characteristics of the vehicle to be identified to obtain the passenger characteristics; and determining the length of the third identification list as the positive binding characteristics.

[0012] According to another aspect of an embodiment of the present application, a vehicle overload identification device is also provided, including: a detection module, used to obtain a single-frame image of a vehicle to be identified, and perform target detection on the single-frame image of the vehicle to be identified to obtain a target detection result, wherein the target detection result at least includes: a body part of a human being in the vehicle to be identified, a human head, and the vehicle to be identified; a pairing module, used to pair and bind the body part and the human head in the vehicle to be identified with the vehicle to be identified to obtain a first identification list of people in the vehicle to be identified, and determine the overload mass score of the vehicle to be identified based on the first identification list, wherein the overload mass score is used to indicate the probability that the vehicle to be identified is overloaded, and the The first identification list includes human identifications bound to the vehicle to be identified; an acquisition module is used to acquire the time information of the vehicle to be identified in each frame image, the positive binding feature of the vehicle to be identified and the passenger feature of the vehicle to be identified, wherein the positive binding feature is used to indicate the effective number of passengers in the vehicle to be identified, and the passenger feature is used to indicate the passenger status of the vehicle to be identified; an identification module is used to determine the identification result of the vehicle to be identified based on the time information of the vehicle to be identified in each frame image, the overload mass score of the vehicle to be identified, the positive binding feature of the vehicle to be identified and the passenger feature of the vehicle to be identified, and the identification result is used to indicate whether the vehicle to be identified is overloaded.

[0013] According to another aspect of the embodiment of the present application, a computer device is also provided, including: a memory and a processor, wherein the memory is used to store program instructions; the processor is connected to the memory and is used to execute the above-mentioned vehicle overload identification method.

[0014] According to another aspect of the embodiment of the present application, a non-volatile storage medium is further provided, which includes a stored computer program, wherein the device where the non-volatile storage medium is located executes the above-mentioned vehicle overload identification method by running the computer program.

[0015] According to another aspect of the embodiment of the present application, a computer program product is also provided, including computer instructions, which implement the above-mentioned vehicle overload identification method when executed by a processor.

[0016] In an embodiment of the present application, a single-frame image of a vehicle to be identified in video stream data is obtained, and target detection is performed on the single-frame image of the vehicle to be identified to obtain a target detection result, wherein the target detection result at least includes: a body part of a human being, a human head, and the vehicle to be identified in the vehicle to be identified; the body part of the human being and the human head in the vehicle to be identified are paired and bound with the vehicle to be identified, and the human body bound to the vehicle to be identified is identified to obtain a first identification list of people in the vehicle to be identified, and an overload mass score of the vehicle to be identified is determined based on the first identification list, the overload mass score is used to indicate the probability of overloading of the vehicle to be identified, and the first identification list contains human body identifications bound to the vehicle to be identified; the time information of the vehicle to be identified in each frame of the image, the positive binding features of the vehicle to be identified, and the positive binding features of the vehicle to be identified are obtained. The passenger characteristics of the vehicle to be identified, wherein the positive binding characteristics are used to indicate the effective number of passengers in the vehicle to be identified, and the passenger characteristics are used to indicate the passenger status of the vehicle to be identified; based on the time information of the vehicle to be identified in each frame of the image, the overload mass score of the vehicle to be identified, the positive binding characteristics of the vehicle to be identified, and the passenger characteristics of the vehicle to be identified, the identification result of the vehicle to be identified is determined, and the identification result is used to indicate whether the vehicle to be identified is overloaded, thereby achieving the purpose of determining the identification result of the vehicle to be identified based on the time information of the vehicle to be identified in each frame of the image, the overload mass score of the vehicle to be identified, the positive binding characteristics of the vehicle to be identified, and the passenger characteristics of the vehicle to be identified, thereby achieving the technical effect of improving the accuracy of vehicle overload identification, and further solving the technical problem of low accuracy in identifying the number of passengers in a vehicle in the related technology. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0018] Figure 1 is a hardware structure block diagram of a computer terminal for implementing a vehicle overload identification method according to an embodiment of the present application;

[0019] Figure 2 is a flow chart of a vehicle overload identification method according to an embodiment of the present application;

[0020] Figure 3 is a flow chart of a method for determining a companion relationship according to an embodiment of the present application;

[0021] Figure 4 is a flow chart of a tracking algorithm according to an embodiment of the present application;

[0022] Figure 5 is a flow chart of a binding feature matching method according to an embodiment of the present application;

[0023] Figure 6 is a recognition flow chart of an overload recognition classifier according to an embodiment of the present application;

[0024] Figure 7 is a structural diagram of a vehicle overload identification system according to an embodiment of the present application;

[0025] Figure 8 It is a structural diagram of a vehicle overload identification device according to an embodiment of the present application. DETAILED DESCRIPTION

[0026] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present application.

[0027] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0028] The information collected in the embodiments of the present application is information and data authorized by the user or fully authorized by all parties, and the collection, storage, use, processing, transmission, provision, disclosure and application of the relevant data comply with the relevant laws, regulations and standards of the relevant regions, take necessary confidentiality measures, do not violate public order and good morals, and provide corresponding operation entrances for users to choose to authorize or reject automated decision-making results; if the user chooses to reject, the expert decision-making process will be entered.

[0029] In order to solve the problems existing in the related art, the embodiment of the present application provides a vehicle overload identification method, which can be run on Figure 1In the computer terminal shown, the computer terminal is explained below.

[0030] The vehicle overload identification method embodiment provided in the embodiment of the present application can be executed in a mobile terminal, a computer terminal or a similar computing device. Figure 1 FIG. 1 shows a hardware structure block diagram of a computer terminal for implementing a vehicle overload identification method. Figure 1 As shown, the computer terminal 10 may include one or more (102a, 102b, ..., 102n are used to illustrate) processors (the processor may include but is not limited to a processing device such as a microprocessor MCU or a programmable logic device FPGA), a memory 104 for storing data, and a transmission module 106 for communication functions connected via a wired and / or wireless network. In addition, it may also include: a display, a keyboard, a cursor control device, an input / output interface (I / O interface), a universal serial bus (USB) port (which may be included as one of the ports of the I / O interface), a network interface, and a BUS bus. Those skilled in the art can understand that Figure 1 The structure shown is only for illustration and does not limit the structure of the above electronic device. Figure 1 More or fewer components as shown, or with Figure 1 Different configurations are shown.

[0031] It should be noted that the one or more processors and / or other data processing circuits described above may generally be referred to herein as "data processing circuits". The data processing circuits may be embodied in whole or in part as software, hardware, firmware, or any other combination thereof. In addition, the data processing circuit may be a single independent processing module, or may be incorporated in whole or in part into any of the other components in the computer terminal 10. As described in the embodiments of the present application, the data processing circuit acts as a processor control (e.g., selection of a variable resistor terminal path connected to an interface).

[0032] The memory 104 can be used to store software programs and modules of application software, such as the program instructions / data storage device corresponding to the vehicle overload identification method in the embodiment of the present application. The processor executes various functional applications and data processing by running the software programs and modules stored in the memory 104, that is, realizing the above-mentioned vehicle overload identification method. The memory 104 may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some examples, the memory 104 may further include a memory remotely arranged relative to the processor, and these remote memories may be connected to the computer terminal 10 via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0033] The transmission module 106 is used to receive or send data via a network. The specific example of the above network may include a wireless network provided by a communication provider of the computer terminal 10. In one example, the transmission module 106 includes a network adapter (Network Interface Controller, NIC), which can be connected to other network devices through a base station so as to communicate with the Internet. In one example, the transmission module 106 can be a radio frequency (RF) module, which is used to communicate with the Internet wirelessly.

[0034] The display may be, for example, a touch screen liquid crystal display (LCD) that enables a user to interact with a user interface of the computer terminal 10 .

[0035] It should be noted that, in some optional embodiments, the above Figure 1 The computer terminal shown may include hardware elements (including circuits), software elements (including computer code stored on a computer-readable medium), or a combination of hardware elements and software elements. It should be noted that Figure 1 This is merely one example of a particular embodiment and is intended to illustrate the types of components that may be present in the computer terminal described above.

[0036] In the above operating environment, an embodiment of the present application provides an embodiment of a vehicle overload identification method. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although the logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0037] Figure 2 is a flow chart of a vehicle overload identification method according to an embodiment of the present application, such as Figure 2As shown, the method comprises the following steps:

[0038] Step S202, obtaining a single frame image of a vehicle to be identified in the video stream data, and performing target detection on the single frame image of the vehicle to be identified to obtain a target detection result, wherein the target detection result at least includes: a body part of a human body in the vehicle to be identified, a human head, and the vehicle to be identified;

[0039] In step S202, the vehicle to be identified can be a motorcycle or an electric vehicle. The detection model for target detection is a detection model that is fine-tuned using the labeled data to adjust the general target detection model, which improves the model's detection effect on truncated or occluded human body parts, human heads, and vehicles when overloaded. Labeling the training data at least includes: labeling the detection frames and categories of human body parts, human heads, and vehicles. The training data includes passenger pictures and videos of electric vehicles and motorcycles.

[0040] Step S204, pairing and binding the human body part and the human head in the vehicle to be identified with the vehicle to be identified, and identifying the human body bound to the vehicle to be identified, obtaining a first identification list of the people in the vehicle to be identified, and determining the overload mass score of the vehicle to be identified according to the first identification list, the overload mass score is used to indicate the probability that the vehicle to be identified is overloaded, and the first identification list contains the human body identification bound to the vehicle to be identified;

[0041] In step S204, the single-frame image is detected using the fine-tuned detection model to obtain the detection results, which include: the detected feature embedding, detection frame, and confidence of the human body, human head, and vehicle, and assign a target ID (human body identifier) ​​to each target (human body). Specifically, through the comprehensive judgment of position, size, and features, the effective human body part-human head-vehicle pairing relationship is screened out, and invalid pairings that may be caused by misdetection or occlusion are eliminated, for example: the human body part and the human head are parts of the same person, for example: check whether the detection frame of the human body part or the human head intersects with the detection frame of the vehicle. If the intersection area exceeds the preset threshold, it is considered that the human body or the human head may be located on the vehicle, thereby establishing a preliminary pairing relationship. For each detected electric vehicle or motorcycle, find the human body part and human head that best match it. Based on the preliminary pairing, further verify the rationality of the binding relationship. For example, check whether the position of the human head is consistent with that of the human body, whether the detection frame of the human head is located above the human body detection frame, and whether the position of the human body is consistent with the expected position of the vehicle seat. Considering that occlusion and truncation may cause incomplete detection of human bodies or heads, the system needs to be able to handle such complex situations. For example, when a detected human body is partially occluded, the system can try to restore the complete pairing relationship by analyzing its relative position to the vehicle and the detected feature parts.

[0042] For each successfully matched person in each vehicle, its ID is collected into list T1 (first identification list). The number of elements in T1 represents the number of valid passengers detected in the vehicle.

[0043] In an optional manner, the specific process of determining the overload mass score of the vehicle to be identified based on the first identification list is as follows: the number of passengers is determined based on the number in the first identification list, and the vehicle's passenger situation is evaluated based on the number of passengers, the vehicle's passenger limit, the confidence weight of each human body, the location of each human body's detection frame and the size of the detection frame of the vehicle to be identified.

[0044] Step S206, obtaining the time information of the vehicle to be identified in each frame of the image, the positive binding feature of the vehicle to be identified, and the passenger feature of the vehicle to be identified, wherein the positive binding feature is used to indicate the number of valid passengers in the vehicle to be identified, and the passenger feature is used to indicate the passenger status of the vehicle to be identified;

[0045] In step S206, the time information of the vehicle to be identified in each frame of the image is used to indicate the image frame where the vehicle to be identified is located in the video stream.

[0046] Step S208, determining the identification result of the vehicle to be identified based on the time information of the vehicle to be identified in each frame image, the overload mass score of the vehicle to be identified, the positive binding feature of the vehicle to be identified, and the passenger feature of the vehicle to be identified. The identification result is used to indicate whether the vehicle to be identified is overloaded.

[0047] Through the above steps S202 to S208, a single-frame image of the vehicle to be identified in the video stream data is obtained, and target detection is performed on the single-frame image of the vehicle to be identified to obtain a target detection result, which at least includes: the body part of the human body, the human head, and the vehicle to be identified in the vehicle to be identified; the body part and the human head in the vehicle to be identified are paired and bound with the vehicle to be identified, and the human body bound to the vehicle to be identified is identified to obtain a first identification list of people in the vehicle to be identified, and the overload mass score of the vehicle to be identified is determined according to the first identification list, and the overload mass score is used to indicate the probability of overloading of the vehicle to be identified, and the first identification list contains the human body identification bound to the vehicle to be identified; the time information of the vehicle to be identified in each frame of the image, the positive binding features of the vehicle to be identified, and the The passenger characteristics of the vehicle to be identified, wherein the positive binding characteristics are used to indicate the effective number of passengers in the vehicle to be identified, and the passenger characteristics are used to indicate the passenger status in the vehicle to be identified; the identification result of the vehicle to be identified is determined based on the time information of the vehicle to be identified in each frame of the image, the overload mass score of the vehicle to be identified, the positive binding characteristics of the vehicle to be identified, and the passenger characteristics of the vehicle to be identified, and the identification result is used to indicate whether the vehicle to be identified is overloaded, thereby achieving the purpose of determining the identification result of the vehicle to be identified based on the time information of the vehicle to be identified in each frame of the image, the overload mass score of the vehicle to be identified, the positive binding characteristics of the vehicle to be identified, and the passenger characteristics of the vehicle to be identified, thereby achieving the technical effect of improving the accuracy of vehicle overload identification, and further solving the technical problem of low accuracy in identifying the number of passengers in a vehicle in the related technology. The following is a detailed description.

[0048] In some embodiments of the present application, obtaining the passenger feature of the vehicle to be identified and the positive binding feature of the vehicle to be identified includes: tracking the vehicle to be identified and the human body identifier bound to the vehicle to be identified in the first identifier list (list T1) respectively to obtain the identifier trajectory of the vehicle to be identified and the human body identifier trajectory bound to the vehicle to be identified, wherein the identifier trajectory is used to represent the movement trajectory of the target in the video; correcting the human body identifier trajectory bound to the vehicle to be identified to obtain a second identifier list (list S2), wherein the second identifier list contains the corrected human body identifier trajectory; determining the intersection of the identifier trajectory in the second identifier list and the first identifier list as a third identifier list (T2), wherein the third identifier list contains the human body identifier trajectory in the intersection of the identifier trajectory in the second identifier list and the first identifier list; determining the passenger feature of the vehicle to be identified and the positive binding feature of the vehicle to be identified based on the human body identifier trajectory in the third identifier list.

[0049] It is understandable that, in the case of obtaining the human body identification trajectory, the vehicle to be identified is first tracked using a general target detection algorithm to obtain the identification trajectory of the vehicle to be identified; the human body on the vehicle to be identified is tracked using a target detection algorithm that applies the binding relationship in the first identification list, and in the process of forming the human body identification trajectory, a tracking algorithm based on the binding relationship is applied to track the human body in each frame. The input of the tracking algorithm includes not only the detection result of the current frame, but also the historical trajectory information of the human body and its bound vehicle. Prediction: Use the trajectory information of the vehicle and human body in the previous frame to predict the possible position of the human body in the current frame. Match the detection result of the current frame with the predicted position, giving priority to the vehicle that has a binding relationship with the human body in the previous frame. The position relationship, appearance features (such as reID features) and motion characteristics of the human body and the vehicle are considered in the matching process. Once the human body and the vehicle are successfully matched in the current frame, the trajectory information of the human body is updated, including its position, appearance features and vehicle identification of the bound vehicle to be identified in the current frame, and the target is the detected target, such as: the vehicle to be identified and the human body.

[0050] Since a person may be temporarily lost in a video stream due to occlusion, angle change, or local movement, the tracking algorithm needs to be able to handle these complex situations. For example, if a person bound to a vehicle is not detected in the current frame, the tracking algorithm that applies the binding relationship can predict the possible position of the person based on the vehicle's motion trajectory and maintain the continuity of its trajectory ID until the person is detected again in a subsequent frame.

[0051] In some embodiments of the present application, the human body trajectory information bound to the vehicle to be identified is corrected to obtain a second identification list, including: obtaining the accompanying relationship between the human body and the vehicle to be identified; eliminating the human body identification trajectories that do not have an accompanying relationship with the vehicle to be identified from the human body identification trajectories bound to the vehicle to be identified, and obtaining the identification trajectories in the second identification list.

[0052] After correcting the human body identification trajectory bound to the vehicle to be identified, the method also includes: obtaining pedestrian re-identification features corresponding to the human body identification trajectory bound to the vehicle to be identified; merging the human body identification trajectory bound to the vehicle to be identified according to the similarity of the pedestrian re-identification features corresponding to the human body identification trajectory bound to the vehicle to be identified, and obtaining the second identification list.

[0053] In actual application scenarios, after eliminating human bodies that do not have a companion relationship, list S1 is obtained. In order to avoid repeated counting of the number of passengers due to changes in identification caused by discontinuous tracking, different human bodies in S1 are merged using the similarity of pedestrian re-identification features, and the list of human identification trajectories matching each car is simplified again S2.

[0054] In some embodiments of the present application, the specific steps of obtaining the accompanying relationship between a human body and the vehicle to be identified include: obtaining a frame number list contained in the identification trajectory of the vehicle to be identified and a frame number list contained in the human body identification trajectory, the frame number list being used to indicate the image frame in which the vehicle identification or human body identification to be identified appears; obtaining an image frame number list in which a binding relationship is successfully established between the passenger and the vehicle to be identified; determining the frame number list contained in the identification trajectory of the identified vehicle as a first frame number list (Fv), determining the frame number list contained in the human body identification trajectory as a second frame number list (Fp), and determining the image frame number list in which a binding relationship is successfully established between the passenger and the vehicle to be identified as a third frame number list (Bp); determining the intersection of the first frame number list and the second frame number list as a fourth frame number list (Fc); performing a difference set between the third frame number list and the fourth frame number list to obtain the accompanying relationship between the human body and the vehicle to be identified, specifically, as Figure 3 As shown, the trajectory frame number list Fv of the vehicle to be identified, the trajectory frame number list Fp of the human body, and the frame number list Bp of the successful binding of the human and vehicle in a single frame are recorded, and the intersection calculation of the vehicle trajectory frame number Fv and the human body trajectory frame number Fp is performed to obtain the frame number Fc where the two appear together, and then the difference calculation of Fc and Bp is performed, that is, to determine whether there is a binding failure phenomenon in the frame where the two appear together. If so, the accompanying relationship of the two trajectories fails, and if not, the accompanying relationship of the two trajectories is established. By converting the complex position calculation and similarity calculation into the intersection and difference calculation of the time series, the algorithm is not limited by the inability to detect and track people and vehicles at the same time, and can efficiently obtain an accurate accompanying relationship.

[0055] It should be noted that the frame number of the successful binding of a person and a vehicle indicates the image frame that can determine the binding of the person and the vehicle to be identified. That is, during the binding, the binding of the person and the vehicle to be identified can be determined through a single frame image, and the person is determined to be in the vehicle to be identified.

[0056] In some embodiments of the present application, a human body identifier bound to a vehicle to be identified is tracked to obtain a human body identifier trajectory bound to the vehicle to be identified, including: matching a high-confidence detection frame and a low-confidence detection frame in a current image frame with a prediction frame in sequence, and when the match is successful, determining that the human body identifier corresponding to the detection frame belongs to the human body identifier trajectory to be matched; in the case of an unsuccessful matching detection frame, obtaining the human body identifier bound to the unsuccessful matching detection frame, and when the vehicle identifier bound to the unsuccessful matching detection frame is the same as the identifier of the vehicle to be identified and the human body identifier in the vehicle to be identified is unique, determining that the human body identifier corresponding to the unsuccessful matching detection frame belongs to the human body identifier trajectory to be matched; when the vehicle identifier bound to the unsuccessful matching detection frame is the same as the identifier of the vehicle to be identified and the human body identifier in the vehicle to be identified is not unique, matching the human body identifier corresponding to the successfully matched detection frame to the human body identifier trajectory with the most similar pedestrian re-identification feature.

[0057] Because the human body in the vehicle is very easy to be blocked and truncated, the size of the detection frame varies greatly, and the similarity of the pedestrian re-identification features is low. Using conventional tracking algorithms to track the movement and feature matching of the detection frame position and pedestrian re-identification features will result in multiple trajectories for one person. In this regard, the embodiment of the present application proposes a method for determining the trajectory of a human body identifier bound to the vehicle to be identified, such as Figure 4 As shown in , after high confidence matching and low confidence matching, the unmatched detection boxes and trajectories are supplemented with matching according to the binding relationship features. Specifically, Figure 5 As shown, analysis is performed based on the trajectory identifier of the vehicle bound to the single frame of the human body. If the vehicle identifier bound to the current detection frame is consistent with the vehicle identifier to be identified bound to the trajectory of the human body identifier to be matched, and the bound human body on the vehicle is unique, the two are directly bound successfully, the trajectory is updated, and the human body identifier corresponding to the current detection frame is added to the trajectory of the human body identifier to be matched; if the human body on the vehicle is not unique, the current detection frame is matched to the trajectory of the human body identifier to be matched with the most similar pedestrian re-identification feature, and the trajectory is updated; if the bound vehicle identifiers of the two are inconsistent, the matching is determined to have failed.

[0058] It can be understood that the vehicle bound to the human body identification trajectory to be matched is the vehicle to be identified.

[0059] For example, when the vehicle identification bound to the unsuccessfully matched detection frame is different from the identification of the vehicle to be identified, it is determined that the matching of the human body identification corresponding to the unsuccessfully matched detection frame fails.

[0060] In some embodiments of the present application, the specific steps of determining the passenger characteristics of the vehicle to be identified and the positive binding characteristics of the vehicle to be identified based on the human identification trajectory in the third identification list include: performing weighted averaging processing on the human characteristics corresponding to the human identification trajectory in the third identification list and the vehicle characteristics of the vehicle to be identified to obtain the passenger characteristics; and determining the length of the third identification list as the positive binding characteristics.

[0061] The specific method of determining the identification result of the vehicle to be identified based on the time information of the vehicle to be identified in each frame of the image, the overload mass score of the vehicle to be identified, the positive binding feature of the vehicle to be identified, and the passenger feature of the vehicle to be identified is as follows: Figure 6 As shown, four kinds of multivariate information, namely single-frame time features (time information of the vehicle to be identified in each frame of the image, frame number token), positive binding features (positive binding token), overloaded mass features (mass token), and passenger features (passenger embedding), are used as the input of the classifier, rather than a single modal information such as a picture. Because in most overloaded pictures, the visible area of ​​the overloaded human body and head is very small, directly using pictures for feature extraction and classification cannot achieve the effective fusion of high-level semantics and large-scale features, resulting in low classification accuracy. In the embodiment of the present application, four kinds of multivariate features are used as the input of the classifier, and through self-attention interactive learning, the multi-granularity information of single-frame, multi-frame, high-level semantic features, and underlying visual features can be used simultaneously to effectively improve the accuracy of overload identification. And compared with the overload identification method of multi-frame fusion counting statistics, this classifier does not require any hyper-parameter settings and can be flexibly deployed and applied.

[0062] The vehicle overload identification method provided by the embodiment of the present application combines the detection of human heads, human bodies and vehicles, and performs mutual binding between the three targets in a single frame. In the tracking stage, the target tracking order of following the vehicle first and then the person is used, as well as the judgment of the accompanying relationship between people and vehicles. This can effectively solve the problem of missing the detection of human bodies with large occlusions, and the loss of detection and tracking of people and vehicles caused by occlusion, the failure of the accompanying relationship judgment, and the inability to obtain the effective number of people. And by integrating fine-grained effective features, overloading behavior can be automatically identified without manual intervention for super-parameter setting. The input of the classifier includes multivariate information such as time features, positive binding features, overload mass features, and passenger features. Through interactive learning of feature self-attention, automatic identification of overloading behavior is achieved. No super-parameter setting is required, and no secondary extraction of image features is required. The overload binary classification result can be accurately output. By comparing the current technology of using pedestrian re-identification and position information to calculate the similarity of position movement, the accompanying relationship between people and vehicles can be efficiently and accurately calculated under the current situation where people and vehicles cannot be detected and tracked at the same time, ensuring the correctness of the number of people. By binding the tracking algorithm of people and vehicles, human body trajectories can be formed, which can effectively solve the problem of frequent interruptions and discontinuities in human body tracking caused by tracking only using position and appearance.

[0063] The present application also provides a vehicle overload identification system, such as Figure 7 As shown, it includes six modules: a frame extraction processing module, a detection module, a single-frame passenger determination module, a tracking module, a companion correction module and an overload recognition classifier, wherein the frame extraction processing module is used to extract a single-frame image from a continuous video stream; the detection module is used to apply a target detection algorithm to identify electric vehicles, motorcycles, human body parts and human heads in the extracted single-frame images; the single-frame passenger determination module is used to determine the binding relationship between the human body and the vehicle based on the output of the detection module; the tracking module is used to output the motion trajectory of each target human body, add binding features to the tracking module, and comprehensively utilize multiple information such as motion, position, appearance, and binding features to solve the problem of frequent interruption and discontinuity of tracking caused by occlusion and truncation of passengers, thereby improving the tracking effect of the human body and facilitating the judgment of the effective number of passengers; the companion correction module is used to remove invalid or erroneous binding relationships (such as irrelevant passers-by) through an efficient companion relationship judgment algorithm, and at the same time merge multiple IDs that may be generated due to discontinuous tracking as the same human body; the overload recognition classifier is used to directly perform a binary classification judgment on whether it is overloaded based on multivariate information analysis.

[0064] Figure 8 A vehicle overload identification device according to an embodiment of the present application includes:

[0065] The detection module 80 is used to obtain a single-frame image of the vehicle to be identified, and perform target detection on the single-frame image of the vehicle to be identified to obtain a target detection result, wherein the target detection result at least includes: a body part of a human body in the vehicle to be identified, a human head, and the vehicle to be identified;

[0066] a pairing module 82, configured to pair and bind the human body parts and human head in the vehicle to be identified with the vehicle to be identified, obtain a first identification list of persons in the vehicle to be identified, and determine an overload mass score of the vehicle to be identified based on the first identification list, wherein the overload mass score is used to indicate a probability that the vehicle to be identified is overloaded, and the first identification list includes human body identifications bound to the vehicle to be identified;

[0067] An acquisition module 84 is used to acquire the time information of the vehicle to be identified in each frame of the image, the positive binding feature of the vehicle to be identified, and the passenger feature of the vehicle to be identified, wherein the positive binding feature is used to indicate the number of valid passengers in the vehicle to be identified, and the passenger feature is used to indicate the passenger status of the vehicle to be identified;

[0068] The identification module 86 is used to determine the identification result of the vehicle to be identified based on the time information of the vehicle to be identified in each frame image, the overload mass score of the vehicle to be identified, the positive binding feature of the vehicle to be identified, and the passenger feature of the vehicle to be identified, and the identification result is used to indicate whether the vehicle to be identified is overloaded.

[0069] Through the above-mentioned vehicle overload identification device, a single-frame image of the vehicle to be identified in the video stream data is obtained, and target detection is performed on the single-frame image of the vehicle to be identified to obtain a target detection result, and the target detection result at least includes: the body part of the human body, the human head, and the vehicle to be identified in the vehicle to be identified; the body part and the human head in the vehicle to be identified are paired and bound with the vehicle to be identified, and the human body bound to the vehicle to be identified is identified to obtain a first identification list of the people in the vehicle to be identified, and the overload mass score of the vehicle to be identified is determined according to the first identification list, and the overload mass score is used to indicate the probability of overloading of the vehicle to be identified, and the first identification list contains the human body identification bound to the vehicle to be identified; the time information of the vehicle to be identified in each frame of the image, the positive binding features of the vehicle to be identified, and the human body binding features of the vehicle to be identified are obtained. The passenger characteristics of the vehicle to be identified, wherein the positive binding characteristics are used to indicate the effective number of passengers in the vehicle to be identified, and the passenger characteristics are used to indicate the passenger status of the vehicle to be identified; the identification result of the vehicle to be identified is determined based on the time information of the vehicle to be identified in each frame image, the overload mass score of the vehicle to be identified, the positive binding characteristics of the vehicle to be identified, and the passenger characteristics of the vehicle to be identified, and the identification result is used to indicate whether the vehicle to be identified is overloaded, thereby achieving the purpose of determining the identification result of the vehicle to be identified based on the time information of the vehicle to be identified in each frame image, the overload mass score of the vehicle to be identified, the positive binding characteristics of the vehicle to be identified, and the passenger characteristics of the vehicle to be identified, thereby achieving the technical effect of improving the accuracy of vehicle overload identification, and further solving the technical problem of low accuracy in identifying the number of passengers in a vehicle in the related technology.

[0070] The acquisition module 84 includes: an acquisition submodule, which is used to obtain the passenger characteristics of the vehicle to be identified and the positive binding characteristics of the vehicle to be identified, including: tracking the vehicle to be identified and the human body identification that has been bound to the vehicle to be identified in the first identification list respectively, to obtain the identification trajectory of the vehicle to be identified and the human body identification trajectory bound to the vehicle to be identified, the identification trajectory is used to represent the movement trajectory of the target in the video; correcting the human body identification trajectory bound to the vehicle to be identified to obtain a second identification list, the second identification list contains the corrected human body identification trajectory; determining the intersection of the identification trajectory in the second identification list and the first identification list as a third identification list, the third identification list contains the human body identification trajectory in the intersection of the identification trajectory in the second identification list and the first identification list; determining the passenger characteristics of the vehicle to be identified and the positive binding characteristics of the vehicle to be identified based on the human body identification trajectory in the third identification list.

[0071] The acquisition submodule includes: a correction unit, which is used to correct the human body trajectory information bound to the vehicle to be identified to obtain a second identification list, including: obtaining the accompanying relationship between the human body and the vehicle to be identified; eliminating the human body identification trajectory that has no accompanying relationship with the vehicle to be identified from the human body identification trajectory bound to the vehicle to be identified, and obtaining the identification trajectory in the second identification list.

[0072] The correction unit includes: an accompanying subunit, which is used to obtain the accompanying relationship between the human body and the vehicle to be identified, including: obtaining a frame number list contained in the identification trajectory of the vehicle to be identified and a frame number list contained in the human body identification trajectory, the frame number list being used to represent the image frame in which the vehicle identification or human body identification to be identified appears; obtaining an image frame number list in which the passenger successfully establishes a binding relationship with the vehicle to be identified; determining the frame number list contained in the identification trajectory of the identified vehicle as the first frame number list, determining the frame number list contained in the human body identification trajectory as the second frame number list, and determining the image frame number list in which the passenger successfully establishes a binding relationship with the vehicle to be identified as the third frame number list; determining the intersection of the first frame number list and the second frame number list as the fourth frame number list; performing a difference set between the third frame number list and the fourth frame number list to obtain the accompanying relationship between the human body and the vehicle to be identified.

[0073] The acquisition submodule includes: a trajectory unit, which is used to track the human body identifier bound to the vehicle to be identified, and obtain the human body identifier trajectory bound to the vehicle to be identified, including: matching the high-confidence detection frame and the low-confidence detection frame in the current image frame with the prediction frame in sequence, and when the match is successful, determining that the human body identifier corresponding to the detection frame belongs to the human body identifier trajectory to be matched; in the case of an unsuccessful matching detection frame, obtaining the human body identifier bound to the unsuccessful matching detection frame, and when the vehicle identifier bound to the unsuccessful matching detection frame is the same as the identifier of the vehicle to be identified and the human body identifier in the vehicle to be identified is unique, determining that the human body identifier corresponding to the unsuccessful matching detection frame belongs to the human body identifier trajectory to be matched; when the vehicle identifier bound to the unsuccessful matching detection frame is the same as the identifier of the vehicle to be identified and the human body identifier in the vehicle to be identified is not unique, matching the human body identifier corresponding to the successfully matched detection frame to the human body identifier trajectory with the most similar pedestrian re-identification feature.

[0074] The trajectory unit also includes: a matching subunit, which is used to determine that the human body identification corresponding to the unsuccessfully matched detection frame fails to match when the vehicle identification bound to the unsuccessfully matched detection frame is different from the identification of the vehicle to be identified.

[0075] The acquisition submodule includes: a merging unit, which is used to, after correcting the human body identification trajectory bound to the vehicle to be identified, the method also includes: obtaining pedestrian re-identification features corresponding to the human body identification trajectory bound to the vehicle to be identified; merging the human body identification trajectory bound to the vehicle to be identified according to the similarity of the pedestrian re-identification features corresponding to the human body identification trajectory bound to the vehicle to be identified, to obtain the second identification list.

[0076] The acquisition submodule includes: a determination unit, which is used to determine the passenger feature of the vehicle to be identified and the positive binding feature of the vehicle to be identified based on the human body identification trajectory in the third identification list, including:

[0077] The human features corresponding to the human identification trajectory in the third identification list and the vehicle features of the vehicle to be identified are weighted averaged to obtain the passenger features; and the length of the third identification list is determined as the positive binding feature.

[0078] It should be noted that Figure 8 The vehicle overload identification device shown is used to perform Figure 2 The vehicle overload identification method shown in the figure, therefore the relevant explanations and instructions in the above-mentioned vehicle overload identification method are also applicable to the vehicle overload identification device, and will not be repeated here.

[0079] An embodiment of the present application also provides a computer device, including: a memory and a processor, wherein the memory is used to store program instructions; the processor is connected to the memory and is used to execute the above-mentioned vehicle overload identification method.

[0080] An embodiment of the present application also provides a non-volatile storage medium, which includes a stored computer program, wherein the device where the non-volatile storage medium is located executes the above-mentioned vehicle overload identification method by running the computer program.

[0081] An embodiment of the present application also provides a computer program product, including computer instructions, which, when executed by a processor, implement the steps of the vehicle overload identification method in the present application.

[0082] The serial numbers of the above-mentioned embodiments of the present application are for description only and do not represent the advantages or disadvantages of the embodiments.

[0083] In the above embodiments of the present application, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, please refer to the relevant description of other embodiments.

[0084] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only schematic. For example, the division of the units can be a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of units or modules, which can be electrical or other forms.

[0085] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple units. Some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0086] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0087] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application can be essentially or partly or all or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), mobile hard disk, disk or optical disk, etc., which can store program code.

[0088] The above is only a preferred implementation of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present application. These improvements and modifications should also be regarded as the scope of protection of the present application.

Claims

1. A vehicle overload identification method, characterized in that: include: Acquire a single-frame image of a vehicle to be identified in the video stream data, and perform target detection on the single-frame image of the vehicle to be identified to obtain a target detection result, wherein the target detection result at least includes: a body part of a human body in the vehicle to be identified, a human head, and the vehicle to be identified; Pairing and binding a human body part and a human head in the vehicle to be identified with the vehicle to be identified, and identifying the human body bound to the vehicle to be identified, obtaining a first identification list of people in the vehicle to be identified, and determining an overload mass score of the vehicle to be identified based on the first identification list, the overload mass score being used to indicate a probability that the vehicle to be identified is overloaded, wherein the first identification list includes human body identifications bound to the vehicle to be identified; Acquire the time information of the vehicle to be identified in each frame of the image, the positive binding feature of the vehicle to be identified, and the passenger feature of the vehicle to be identified, wherein the positive binding feature is used to indicate the effective number of passengers in the vehicle to be identified, and the passenger feature is used to indicate the passenger status of the vehicle to be identified; An identification result of the vehicle to be identified is determined based on the time information of the vehicle to be identified in each frame of the image, the overload mass score of the vehicle to be identified, the positive binding feature of the vehicle to be identified, and the passenger feature of the vehicle to be identified. The identification result is used to indicate whether the vehicle to be identified is overloaded.

2. The method according to claim 1, characterized in that Acquiring the passenger feature of the vehicle to be identified and the positive binding feature of the vehicle to be identified, including: Tracking the vehicle to be identified and the human body identifier bound to the vehicle to be identified in the first identifier list respectively, obtaining the identifier trajectory of the vehicle to be identified and obtaining the human body identifier trajectory bound to the vehicle to be identified, wherein the identifier trajectory is used to represent the moving trajectory of the target in the video; Correcting the human body identification track bound to the vehicle to be identified to obtain a second identification list, wherein the second identification list includes the corrected human body identification track; Determine the intersection of the identification tracks in the second identification list and the first identification list as a third identification list, wherein the third identification list includes the human body identification tracks in the intersection of the identification tracks in the second identification list and the first identification list; The passenger feature of the vehicle to be identified and the positive binding feature of the vehicle to be identified are determined based on the human body identification trajectory in the third identification list.

3. The method according to claim 2, characterized in that Correcting the human body trajectory information bound to the vehicle to be identified to obtain a second identification list, including: Acquiring the accompanying relationship between the human body and the vehicle to be identified; The human body identification tracks that have no accompanying relationship with the vehicle to be identified are eliminated from the human body identification tracks bound to the vehicle to be identified, so as to obtain the identification tracks in the second identification list.

4. The method according to claim 3, characterized in that: Acquiring the accompanying relationship between the human body and the vehicle to be identified, including: Obtaining a frame number list contained in the identification track of the vehicle to be identified and a frame number list contained in the human identification track, wherein the frame number list is used to indicate the image frame in which the vehicle identification or human identification to be identified appears; Obtain a list of image frame numbers in which a binding relationship is successfully established between the passenger and the vehicle to be identified; The frame number list included in the identification track of the identified vehicle is determined as the first frame number list, the frame number list included in the human body identification track is determined as the second frame number list, and the image frame number list in which the passenger successfully establishes a binding relationship with the vehicle to be identified is determined as the third frame number list; Determine an intersection of the first frame number list and the second frame number list as a fourth frame number list; The third frame number list and the fourth frame number list are subtracted to obtain the accompanying relationship between the human body and the vehicle to be identified.

5. The method according to claim 2, characterized in that: Tracking the human body identifier bound to the vehicle to be identified to obtain the trajectory of the human body identifier bound to the vehicle to be identified includes: Sequentially match the high-confidence detection frame and the low-confidence detection frame in the current image frame with the prediction frame. If the match is successful, determine that the human body marker corresponding to the detection frame belongs to the human body marker trajectory to be matched; In the case where there is an unsuccessfully matched detection frame, obtaining a human body identifier bound to the unsuccessfully matched detection frame, and in the case where the vehicle identifier bound to the unsuccessfully matched detection frame is the same as the identifier of the vehicle to be identified and the human body identifier in the vehicle to be identified is unique, determining that the human body identifier corresponding to the unsuccessfully matched detection frame belongs to the human body identifier trajectory to be matched; When the vehicle identifier bound to the unmatched detection frame is the same as the identifier of the vehicle to be identified and the human identifier in the vehicle to be identified is not unique, the human identifier corresponding to the successfully matched detection frame is matched to the human identifier trajectory with the most similar pedestrian re-identification feature.

6. The method according to claim 5, characterized in that The method further comprises: When the vehicle identification bound to the unsuccessfully matched detection frame is different from the identification of the vehicle to be identified, it is determined that the matching of the human body identification corresponding to the unsuccessfully matched detection frame fails.

7. The method according to claim 2, characterized in that After correcting the trajectory of the human body marker bound to the vehicle to be identified, the method further includes: Obtaining pedestrian re-identification features corresponding to the human body identification trajectory bound to the vehicle to be identified; The human body identification tracks bound to the vehicle to be identified are merged according to the similarity of the pedestrian re-identification features corresponding to the human body identification tracks bound to the vehicle to be identified to obtain the second identification list.

8. The method according to claim 2, characterized in that: Determining the passenger feature of the vehicle to be identified and the positive binding feature of the vehicle to be identified based on the human body identification trajectory in the third identification list includes: Performing weighted average processing on the human features corresponding to the human identification trajectory in the third identification list and the vehicle features of the vehicle to be identified to obtain the passenger features; The length of the third identification list is determined as the positive binding feature.

9. A vehicle overload identification device, characterized in that: include: A detection module, used to obtain a single-frame image of a vehicle to be identified, and perform target detection on the single-frame image of the vehicle to be identified to obtain a target detection result, wherein the target detection result at least includes: a body part of a human body in the vehicle to be identified, a human head, and the vehicle to be identified; a pairing module, used for pairing and binding a human body part and a human head in the vehicle to be identified with the vehicle to be identified, obtaining a first identification list of persons in the vehicle to be identified, and determining an overload mass score of the vehicle to be identified according to the first identification list, wherein the overload mass score is used to indicate a probability that the vehicle to be identified is overloaded, and the first identification list contains human body identifications bound to the vehicle to be identified; an acquisition module, used to acquire the time information of the vehicle to be identified in each frame of the image, the positive binding feature of the vehicle to be identified, and the passenger feature of the vehicle to be identified, wherein the positive binding feature is used to indicate the effective number of passengers in the vehicle to be identified, and the passenger feature is used to indicate the passenger status of the vehicle to be identified; An identification module is used to determine the identification result of the vehicle to be identified based on the time information of the vehicle to be identified in each frame of the image, the overload mass score of the vehicle to be identified, the positive binding feature of the vehicle to be identified, and the passenger feature of the vehicle to be identified, wherein the identification result is used to indicate whether the vehicle to be identified is overloaded.

10. A computer device, characterized in that: include: A memory and a processor, wherein the memory is used to store program instructions; The processor is connected to the memory and is used to execute the vehicle overload identification method described in any one of claims 1 to 8.

11. A computer program product comprising computer instructions, characterized in that: When the computer instructions are executed by a processor, the vehicle overload identification method described in any one of claims 1 to 8 is implemented.

Images