A motor vehicle overweight detection method, device, equipment, system and storage medium

Abstract

The application discloses a kind of motor vehicle overweight detection method, device, equipment, system and storage medium, it is related to vehicle detection technical field.The method is first according to radar data and video data, using radar and video fusion tracking mode tracks each motor vehicle present in weighing road section area, obtains the road section passing track type and road section passing speed of each motor vehicle, and extracts the measured vehicle weight of each motor vehicle from weighing data, then according to the multiple known actual vehicle weight of experimental motor vehicle and when recording in multiple times through weighing road section area, multiple measured vehicle weight, multiple road section passing track type and multiple road section passing speed, mapping function of actual vehicle weight and measured vehicle weight, road section passing track type and road section passing speed is fitted, finally the measured vehicle weight, road section passing track type and road section passing speed of non-experimental motor vehicle are imported into mapping function, obtain the actual vehicle weight of non-experimental motor vehicle, and compare whether it is overweight.

Description

Technical Field

[0001] This invention belongs to the field of vehicle inspection technology, specifically relating to a method, device, equipment, system, and storage medium for detecting overweight motor vehicles. Background Technology

[0002] With the rapid development of the economy and society, the types and modes of transportation are becoming increasingly diversified. Modern transportation modes mainly include railway transportation, road transportation, waterway transportation and air transportation. Among them, road transportation has always been an important component of transportation due to its strong adaptability.

[0003] In actual road transport, the large number of vehicles traveling on the road every day leads to poor road continuity and requires frequent maintenance. In particular, overweight vehicles can cause significant damage to the road surface, especially when they cross bridges. This can result in minor internal damage to the bridge structure or even bridge collapse, leading to substantial economic losses and casualties. Therefore, it is necessary to conduct fixed-point weighing inspections of vehicles on the road to promptly detect overweight vehicles.

[0004] Current methods for detecting overweight vehicles rely on fixed-point ground weighing equipment to collect vehicle weight. This allows some vehicles to evade penalties by driving in an S-shaped route, thus causing the weighing equipment to misread the weight. Therefore, how to provide a new method for detecting overweight vehicles to prevent vehicles from cheating by driving in an S-shaped route is a topic that urgently needs to be studied by those skilled in the art. Summary of the Invention

[0005] The purpose of this invention is to provide a method, device, system, computer equipment, and computer-readable storage medium for detecting overweight motor vehicles, in order to solve the problem that existing overweight motor vehicle detection schemes cannot prevent motor vehicles from cheating by driving in an S-shaped route.

[0006] To achieve the above objectives, the present invention adopts the following technical solution:

[0007] Firstly, a method for detecting overweight motor vehicles is provided, including:

[0008] The system acquires radar data collected in real time from the weighing section area by radar, video data collected in real time from the weighing section area by a camera, and weighing data collected in real time by ground weighing equipment, wherein the ground weighing equipment is arranged in the weighing section area.

[0009] Based on the radar data and the video data, a radar-visual fusion tracking method is used to track each motor vehicle appearing in the weighing section area to obtain vehicle tracking results. The vehicle tracking results include the vehicle body marking frame and the speed of each motor vehicle passing through the section. The motor vehicles include experimental motor vehicles and non-experimental motor vehicles. The experimental motor vehicle refers to a motor vehicle that passes through the weighing section area multiple times before the non-experimental motor vehicle with different known actual vehicle weights, trajectory types and / or vehicle speeds.

[0010] For each motor vehicle, the corresponding road segment passage trajectory type is determined based on the vehicle body marker box in the vehicle tracking results;

[0011] The vehicle weights of each motor vehicle collected when passing through the weighing section area are extracted from the weighing data.

[0012] Based on the experimental motor vehicle's multiple known actual vehicle weights, multiple measured vehicle weights, multiple road segment trajectory types, and multiple road segment speeds recorded when it passed through the weighing section area multiple times, a mapping function between the actual vehicle weight and the measured vehicle weight, road segment trajectory type, and road segment speed is fitted, wherein the multiple known actual vehicle weights, the multiple measured vehicle weights, the multiple road segment trajectory types, and the multiple road segment speeds correspond one-to-one with the multiple times.

[0013] The measured vehicle weight, road segment trajectory type, and road segment speed of the non-experimental motor vehicle are imported into the mapping function to obtain the actual vehicle weight of the non-experimental motor vehicle.

[0014] Determine whether the actual weight of the non-experimental motor vehicle exceeds a preset weight threshold;

[0015] If so, the non-experimental motor vehicle is determined to be overweight, and an alarm is triggered.

[0016] Based on the above-mentioned invention, a novel scheme for detecting overweight vehicles by combining radar-visual fusion technology is provided. First, based on radar and video data, radar-visual fusion tracking is used to track each vehicle appearing in the weighing section area, obtaining the road segment passage trajectory type and speed of each vehicle, and extracting the measured weight of each vehicle from the weighing data. Then, based on multiple known actual vehicle weights, multiple measured vehicle weights, multiple road segment passage trajectory types, and multiple road segment passage speeds recorded by the experimental vehicle during multiple passages through the weighing section area, a mapping function is fitted to the actual vehicle weight and the measured vehicle weight, road segment passage trajectory type, and road segment passage speed. Finally, the measured vehicle weight, road segment passage trajectory type, and road segment passage speed of the non-experimental vehicle are imported into the mapping function to obtain the actual vehicle weight of the non-experimental vehicle. A threshold comparison method is then used to determine whether the vehicle is overweight. By combining radar-visual fusion technology, this scheme can effectively prevent vehicles from cheating by taking S-shaped routes during weighing, facilitating practical application and promotion.

[0017] In one possible design, based on the radar data and the video data, a radar-visual fusion tracking method is used to track each motor vehicle appearing in the weighing section area to obtain vehicle tracking results, including: radar target detection, radar target tracking, image target detection, image target tracking, and target recognition and tracking result fusion.

[0018] The radar target detection stage is used to detect the position, speed and azimuth of each obstacle appearing in the weighing section area based on the radar data;

[0019] The radar target tracking step is used to track each obstacle object using an HM object tracker based on the radar data, and obtain the radar target tracking results of each obstacle object.

[0020] The image target detection step is used to identify each target object appearing in the weighing section area based on the video data using the YOLO v4 target detection model, and to obtain the object bounding box of each target object, wherein the target object includes motor vehicles;

[0021] The image target tracking step is used to track each target object according to the video data using the deepsort multi-target tracking algorithm, and obtain the video target tracking results of each target object;

[0022] The target recognition and tracking result fusion step is used to determine at least one pair of obstacle objects and target objects that are the same object based on the position and azimuth of each obstacle object and the position and orientation of each target object relative to the camera determined by the camera calibration method. For each pair of obstacle objects and target objects, the speed and radar target tracking results of the corresponding obstacle object and the object marker box and video target tracking results of the corresponding target object are fused to obtain the vehicle tracking result. The vehicle tracking result includes the vehicle body marker box and the vehicle speed of each motor vehicle appearing in the weighing section area. The motor vehicles include experimental motor vehicles and non-experimental motor vehicles. The experimental motor vehicle refers to a motor vehicle that has passed through the weighing section area multiple times before the non-experimental motor vehicle with different known actual vehicle weights, trajectory types and / or vehicle speeds.

[0023] In one possible design, for each motor vehicle, the corresponding road segment trajectory type is determined based on the vehicle body marker frame in the vehicle tracking results, including:

[0024] For each motor vehicle, the center point of the corresponding vehicle body marker frame in the vehicle tracking result is taken as the corresponding vehicle body center point, and the corresponding motion trajectory is formed based on the three-dimensional coordinates of the vehicle body center point in time.

[0025] For each motor vehicle, the corresponding motion trajectory is imported into a pre-trained trajectory type recognition model based on a neural network, and the corresponding road segment passage trajectory type is output.

[0026] In one possible design, the road segment passes through trajectory types including straight trajectory type and S-curve trajectory type.

[0027] In one possible design, the mapping function is expressed by the following formula:

[0028]

[0029] In the formula, w represents the actual vehicle weight, and f() represents the mapping function. denoted by , where t represents the road segment trajectory type, v represents the road segment speed, a1 represents the coefficient of influence of the measured vehicle weight on the actual vehicle weight obtained through fitting when the road segment trajectory type is a straight line, b1 represents the coefficient of influence of the road segment speed on the actual vehicle weight obtained through fitting when the road segment trajectory type is a straight line, c1 represents the bias coefficient obtained through fitting that affects the actual vehicle weight when the road segment trajectory type is a straight line, a2 represents the coefficient of influence of the measured vehicle weight on the actual vehicle weight obtained through fitting when the road segment trajectory type is an S-curve, b2 represents the coefficient of influence of the road segment speed on the actual vehicle weight obtained through fitting when the road segment trajectory type is an S-curve, and c2 represents the bias coefficient obtained through fitting that affects the actual vehicle weight when the road segment trajectory type is an S-curve.

[0030] In one possible design, triggering alarm actions include:

[0031] Extract video footage containing vehicle images and vehicle body marking frames of the non-experimental motor vehicle from the vehicle tracking results;

[0032] The vehicle image of the non-experimental motor vehicle is extracted from the video footage, and the license plate rectangle outline is detected and processed to obtain the license plate marking frame.

[0033] Based on the license plate marking frame, extract the license plate image from the vehicle image;

[0034] The license plate image is processed by the character recognition package pytesseract to obtain a string, which is then used as the license plate number of the non-experimental motor vehicle.

[0035] An overweight alarm message is generated based on the license plate number and actual weight of the non-experimental motor vehicle, and the overweight alarm message is uploaded to the motor vehicle overweight monitoring platform.

[0036] Secondly, a motor vehicle overweight detection device is provided, including a data acquisition module, a vehicle tracking module, a trajectory type determination module, a vehicle weight extraction module, a mapping function fitting module, an actual vehicle weight calculation module, an overweight judgment module, and an overweight alarm module.

[0037] The data acquisition module is used to acquire radar data collected in real time by radar on the weighing section area, video data collected in real time by camera on the weighing section area, and weighing data collected in real time by ground weighing equipment, wherein the ground weighing equipment is arranged in the weighing section area.

[0038] The vehicle tracking module is communicatively connected to the data acquisition module and is used to track each motor vehicle appearing in the weighing section area using a radar-visual fusion tracking method based on the radar data and the video data, and obtain vehicle tracking results. The vehicle tracking results include the vehicle body marking frame of each motor vehicle and the vehicle speed passing through the section. The motor vehicles include experimental motor vehicles and non-experimental motor vehicles. The experimental motor vehicle refers to a motor vehicle that passes through the weighing section area multiple times before the non-experimental motor vehicle with different known actual vehicle weights, trajectory types and / or vehicle speeds.

[0039] The trajectory type determination module is communicatively connected to the vehicle tracking module and is used to determine the corresponding road segment passage trajectory type for each motor vehicle based on the corresponding vehicle body marker box in the vehicle tracking result.

[0040] The vehicle weight extraction module is communicatively connected to the data acquisition module and is used to extract the measured vehicle weights of each motor vehicle when passing through the weighing section area from the weighing data.

[0041] The mapping function fitting module is communicatively connected to the vehicle tracking module, the trajectory type determination module, and the measured vehicle weight extraction module. It is used to fit a mapping function between the actual vehicle weight and the measured vehicle weight, the trajectory type, and the speed based on multiple known actual vehicle weights, multiple measured vehicle weights, multiple road segment passing trajectories, and multiple road segment passing speeds recorded by the experimental motor vehicle when it passes through the weighing section area multiple times. The multiple known actual vehicle weights, the multiple measured vehicle weights, the multiple road segment passing trajectories, and the multiple road segment passing speeds correspond one-to-one with the multiple times recorded.

[0042] The actual vehicle weight calculation module is communicatively connected to the vehicle tracking module, the trajectory type determination module, the measured vehicle weight extraction module, and the mapping function fitting module, respectively. It is used to import the measured vehicle weight, road segment trajectory type, and road segment speed of the non-experimental motor vehicle into the mapping function to obtain the actual vehicle weight of the non-experimental motor vehicle.

[0043] The overweight judgment module is communicatively connected to the actual vehicle weight calculation module and is used to determine whether the actual vehicle weight of the non-experimental motor vehicle exceeds a preset vehicle weight threshold.

[0044] The overweight alarm module is communicatively connected to the overweight judgment module. It is used to determine that the non-experimental motor vehicle is overweight when the actual weight of the non-experimental motor vehicle exceeds the preset weight threshold, and to trigger an alarm action.

[0045] In one possible design, the vehicle tracking module includes a radar target detection unit, a radar target tracking unit, an image target detection unit, an image target tracking unit, and a target recognition and tracking result fusion unit;

[0046] The radar target detection unit is used to detect the position, speed and azimuth of each obstacle appearing in the weighing section area based on the radar data;

[0047] The radar target tracking unit is used to track each obstacle object using an HM object tracker based on the radar data, and obtain the radar target tracking results of each obstacle object.

[0048] The image target detection unit is used to identify each target object appearing in the weighing section area based on the video data using the YOLO v4 target detection model, and to obtain the object bounding box of each target object, wherein the target object includes motor vehicles.

[0049] The image target tracking unit is used to track each target object according to the video data using the deepsort multi-target tracking algorithm, and obtain the video target tracking results of each target object;

[0050] The target recognition and tracking result fusion unit is communicatively connected to the radar target detection unit, the radar target tracking unit, the image target detection unit, and the image target tracking unit. It is used to determine at least one pair of obstacle objects and target objects that are considered the same object based on the position and azimuth of each obstacle object and the position and azimuth of each target object relative to the camera, determined by camera calibration. For each pair of obstacle objects and target objects, it fuses the speed of the corresponding obstacle object and the radar target tracking result, as well as the object marker frame and video target tracking result of the corresponding target object, to obtain the vehicle tracking result. The vehicle tracking result includes the vehicle body marker frame and the vehicle speed of each motor vehicle appearing in the weighing section area. The motor vehicles include experimental motor vehicles and non-experimental motor vehicles. The experimental motor vehicle refers to a motor vehicle that has passed through the weighing section area multiple times before the non-experimental motor vehicle with different known actual vehicle weights, trajectory types, and / or speeds.

[0051] Thirdly, the present invention provides a vehicle overweight detection system, including a radar, a camera, a ground weighing device, and an intelligent roadside device, wherein the intelligent roadside device is communicatively connected to the radar, the camera, and the ground weighing device respectively;

[0052] The radar is used to collect radar data of the weighing section area in real time and transmit the radar data to the intelligent roadside device in real time.

[0053] The camera is used to collect video data of the weighing section area in real time and transmit the video data to the intelligent roadside device in real time.

[0054] The ground weighing equipment is arranged in the weighing section area to collect weighing data in real time and transmit the weighing data to the intelligent roadside equipment in real time.

[0055] The intelligent roadside equipment is used to perform the motor vehicle overweight detection method as described in the first aspect or any possible design in the first aspect.

[0056] Fourthly, the present invention provides a computer device comprising a memory, a processor, and a transceiver connected in sequence for communication, wherein the memory is used to store a computer program, the transceiver is used to send and receive messages, and the processor is used to read the computer program and execute the motor vehicle overweight detection method as described in the first aspect or any possible design in the first aspect.

[0057] Fifthly, the present invention provides a computer-readable storage medium storing instructions that, when executed on a computer, perform the motor vehicle overweight detection method as described in the first aspect or any possible design of the first aspect.

[0058] In a sixth aspect, the present invention provides a computer program product containing instructions that, when executed on a computer, cause the computer to perform the motor vehicle overweight detection method as described in the first aspect or any possible design in the first aspect.

[0059] The beneficial effects of the above scheme are:

[0060] (1) This invention creatively provides a new scheme for detecting overweight motor vehicles by combining radar and video fusion technology. First, based on radar data and video data, radar and video fusion tracking is used to track each motor vehicle appearing in the weighing section area to obtain the road passage trajectory type and road passage speed of each motor vehicle, and the measured weight of each motor vehicle is extracted from the weighing data. Then, based on the experimental motor vehicle's multiple known actual weights, multiple measured weights, multiple road passage trajectory types, and multiple road passage speeds recorded when passing through the weighing section area multiple times, a mapping function is fitted to obtain the actual weight and the measured weight, road passage trajectory type, and road passage speed. Finally, the measured weight, road passage trajectory type, and road passage speed of the non-experimental motor vehicle are imported into the mapping function to obtain the actual weight of the non-experimental motor vehicle, and whether it is overweight is determined by threshold comparison. In this way, by combining radar and video fusion technology, it can effectively prevent motor vehicles from cheating by taking S-shaped routes, which is convenient for practical application and promotion. Attached Figure Description

[0061] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0062] Figure 1 This is a flowchart illustrating the method for detecting overweight motor vehicles provided in an embodiment of this application.

[0063] Figure 2 This is a schematic diagram illustrating the principle of obstacle azimuth detection based on radar data, provided in an embodiment of this application.

[0064] Figure 3 An example diagram showing the vehicle tracking results obtained based on the radar-visual fusion tracking method provided in this application embodiment.

[0065] Figure 4 This is a schematic diagram of the structure of the motor vehicle overweight detection device provided in the embodiments of this application.

[0066] Figure 5 This is a schematic diagram of the structure of the motor vehicle overweight detection system provided in an embodiment of this application.

[0067] Figure 6 A schematic diagram of the structure of a computer device provided in an embodiment of this application. Detailed Implementation

[0068] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the present invention will be briefly introduced below in conjunction with the accompanying drawings and descriptions of the embodiments or the prior art. Obviously, the following description of the structure of the accompanying drawings is only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. It should be noted that the description of these embodiments is for the purpose of helping to understand the present invention, but does not constitute a limitation of the present invention.

[0069] It should be understood that although the terms "first" and "second", etc., may be used herein to describe various objects, these objects should not be limited by these terms. These terms are only used to distinguish one object from another. For example, the first object may be referred to as the second object, and similarly, the second object may be referred to as the first object, without departing from the scope of the exemplary embodiments of the invention.

[0070] It should be understood that the term "and / or" that may appear in this document is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can mean: A exists alone, B exists alone, or A and B exist simultaneously. Another example is A, B and / or C, which can mean that any one of A, B, and C or any combination thereof exists. The term " / and" that may appear in this document describes another relationship between related objects, indicating that two relationships can exist. For example, A / and B can mean: A exists alone or A and B exist simultaneously. In addition, the character " / " that may appear in this document generally indicates that the related objects before and after it are in an "or" relationship.

[0071] Example:

[0072] like Figure 1 As shown, the method for detecting overweight motor vehicles provided in the first aspect of this embodiment can be executed, but is not limited to, by a computer device with certain computing resources that is communicatively connected to radar, cameras, and ground weighing equipment. For example, it can be executed by electronic devices such as intelligent roadside units (RSUs, which are key infrastructure devices for the networking and intelligentization of road infrastructure in vehicle-road cooperative systems), platform servers, personal computers (PCs, referring to a type of multi-purpose computer suitable for personal use in terms of size, price, and performance; desktop computers, laptops, mini-laptops, tablets, and ultrabooks are all considered personal computers), smartphones, personal digital assistants (PDAs), or wearable devices. Figure 1 As shown, the method for detecting overweight motor vehicles may include, but is not limited to, the following steps S1 to S8.

[0073] S1. Acquire radar data collected in real time from the radar on the weighing section area, video data collected in real time from the camera on the weighing section area, and weighing data collected in real time from the ground weighing equipment, wherein the ground weighing equipment is arranged in the weighing section area.

[0074] In step S1, the weighing section area is the fixed-point weighing area in the existing vehicle overweight detection scheme. The radar's detection range covers the weighing section area, and it is used to collect radar data in real time as traffic monitoring data through electromagnetic wave transmission and reception; the radar may be, but is not limited to, millimeter-wave radar. The camera's field of view also covers the weighing section area, and it is used to collect video frame images of the weighing section area in real time, obtaining video data containing several consecutive video frame images as another traffic monitoring data; the camera may be, but is not limited to, a binocular camera. The ground weighing equipment is an essential device in the existing vehicle overweight detection scheme. In addition, the radar, the camera, and the ground weighing equipment can be arranged in a conventional manner, and the collected data can be transmitted to local equipment.

[0075] S2. Based on the radar data and the video data, a radar-visual fusion tracking method is used to track each motor vehicle appearing in the weighing section area to obtain vehicle tracking results. The vehicle tracking results include the vehicle body marking frame and the speed of each motor vehicle passing through the section. The motor vehicles include experimental motor vehicles and non-experimental motor vehicles. The experimental motor vehicle refers to a motor vehicle that passes through the weighing section area multiple times before the non-experimental motor vehicle with different known actual vehicle weights, trajectory types and / or vehicle speeds.

[0076] In step S2, radar-visual fusion involves multi-dimensionally fusing video data with millimeter-wave radar data and combining it with relevant algorithmic capabilities to achieve data acquisition and feature analysis of road surface information through multi-band perception, and to make decisions. Specifically, based on the radar data and the video data, a radar-visual fusion tracking method is used to track each motor vehicle appearing in the weighing section area, obtaining vehicle tracking results, including but not limited to: radar target detection, radar target tracking, image target detection, image target tracking, and target recognition and tracking result fusion.

[0077] The radar target detection stage is used to detect the position, speed and azimuth of each obstacle appearing in the weighing section area based on the radar data. The obstacle may be, but is not limited to, objects such as motor vehicles. The specific methods for obtaining its position, speed and azimuth may include, but are not limited to, the following: (1) Regarding the position, since the millimeter-wave radar emits directional millimeter waves of the corresponding band through the transmitting antenna, when the millimeter wave encounters an obstacle, it will be reflected back. The reflected millimeter wave is then received by the receiving antenna. Thus, based on the band of the millimeter wave, the flight time of the millimeter wave can be calculated by the formula × speed of light ÷ 2. Combined with the stationary characteristics of the radar and the driving speed of the vehicle, the relative distance between the millimeter-wave radar and the obstacle can be known, and the position of the obstacle can also be known. (2) Regarding speed, according to the Doppler effect, the frequency change of the millimeter wave reflected echo is closely related to the vehicle's speed relative to the radar. Therefore, based on the frequency change of the millimeter wave reflected echo, the relative speed of the obstacle being tracked in real time in front of the radar compared to the stationary radar can be determined. That is, when the obstacle decelerates, the frequency of the millimeter wave reflected echo will increase, and vice versa. (3) Regarding azimuth, the detection principle of millimeter radar is as follows: after the millimeter wave is emitted by the transmitting antenna of the millimeter radar, if the millimeter wave encounters an obstacle, it will be reflected back. Through the parallel receiving antennas of the millimeter radar, the phase difference of the millimeter waves reflected back from the same obstacle can be received, and the azimuth of the obstacle can be calculated. For example, such as Figure 2 As shown, the azimuth angle α can be calculated using trigonometric functions based on the geometric distance d between the millimeter-wave radar receiving antennas RX1 and RX2, and the phase difference b between the reflected echoes received by these two millimeter-wave radar antennas. AZ The value of can be used to determine the azimuth angle of the obstacle object Target.

[0078] The radar target tracking stage is used to track each obstacle object using the HM object tracker based on the radar data, and obtain the radar target tracking results of each obstacle object. The HM object tracker is an existing perception analysis tool for tracking segmented detected obstacles; it usually forms and updates the tracking list by associating the current detection with the existing tracking list. If it no longer exists, the old tracking list is deleted, and a new tracking list is generated when a new detection is identified; the motion state of the updated tracking list will be estimated after association. In the HM object tracker, the Hungarian algorithm is used to detect the tracking association, and a robust Kalman filter is used for motion estimation. The main tracking process of the HM object tracker is as follows: (1) Preprocessing, including object coordinate system transformation; creating a tracking object for each object and adding it to the tracking list; recording the currently tracked object; (2) Kalman filter filtering, predicting the current position and velocity of the object (Kalman filter stage 1: prediction stage); (3) Hungarian algorithm matching, associating the detected object and the tracked object; (4) Kalman filter, updating the position and velocity information of the tracked object (Kalman filter stage 2: update stage).

[0079] The image target detection step is used to identify various target objects appearing in the weighing section area based on the video data using a YOLOv4 target detection model, and to obtain object bounding boxes for each target object, wherein the target objects include motor vehicles. The YOLOv4 target detection model consists of three parts: a backbone network, a neck network, and a head network. The backbone network can use a CSPDarknet53 network (CSP stands for Cross Stage Partial) for feature extraction. The neck network consists of an SPP (Spatial Pyramid Pooling block) and a PANet (Path Aggregation Network). The former is used to increase the receptive field and separate the most important features, while the latter is used to ensure that semantic features are received from higher-level layers and fine-grained features are received from lower-level layers of the cross backbone network. The head network described above performs detection based on anchor boxes and uses three different feature maps of varying sizes: 13×13, 26×26, and 52×52. These feature maps are used to detect targets from largest to smallest (here, larger feature maps contain more information; therefore, the 52×52 feature map is used to detect small targets, and vice versa). Furthermore, the specific usage of the YOLO v4 object detection model for target object detection is existing technology and will not be elaborated upon here.

[0080] The image target tracking step is used to track each target object according to the video data using the deepsort multi-target tracking algorithm, and obtain the video target tracking result of each target object. The specific process of the deepsort multi-target tracking algorithm is as follows: (1) When the first frame of video image comes in, the tracker is initialized and a new tracker is created with the detected target and labeled with ID; (2) When the subsequent video image frame comes in, the state prediction and covariance prediction generated by the box in the previous frame are obtained in the Kalman filter, the IOU (Intersection over Union) of all target state predictions of the tracker and the box detected in this frame is calculated, the unique match with the largest IOU (data association part) is obtained through the Hungarian assignment algorithm, and then the matching pairs with matching values ​​less than iou_threshold are removed; (3) The Kalman tracker is updated with the target detection box matched in this frame, the Kalman gain, state update and covariance update are calculated, and the state update value is output as the tracking box of this frame. At the same time, the tracker is re-initialized for the target that is not matched in this frame. Considering that unmatched tracking targets may include occluded targets that are not detected in the current frame, the Deepsort multi-target tracking algorithm preferably includes the following steps: For unmatched tracking targets, determine whether they have left the weighing section area based on their current predicted position (i.e., determine whether they still appear in the frame image based on the positional relationship between the current predicted position and the frame image boundary; if they do, it is determined that they have not left the weighing section area; otherwise, it is determined that they have left the weighing section area). If so, the tracking target is discarded in the next position prediction; otherwise, the Kalman filter method is still used to predict the corresponding position in the next position prediction. Thus, when an occluded tracking target is detected (i.e., the tracking target is not matched with a detected target and its current predicted position still appears in the current frame image), the tracking target can be marked as occluded (masked), and the Kalman filter method is still used to predict its position in the next frame image until a detected target is matched or the tracking target is found to be outside the image range. Therefore, compared to general discarding methods, this avoids premature termination of video tracking due to occlusion, ensuring long-term tracking continuity.

[0081] The target recognition and tracking result fusion step is used to determine at least one pair of obstacle objects and target objects that are the same object, based on the position and azimuth of each obstacle object and the position and azimuth of each target object relative to the camera determined by the camera calibration method. For each pair of obstacle objects and target objects, the speed and radar target tracking results of the corresponding obstacle object, as well as the object bounding box and video target tracking results of the corresponding target object, are fused to obtain vehicle tracking results. The vehicle tracking results include the vehicle body bounding boxes and the speed of each motor vehicle appearing in the weighing section area. The motor vehicles include experimental motor vehicles and non-experimental motor vehicles. The experimental motor vehicle refers to a motor vehicle that has passed through the weighing section area multiple times before the non-experimental motor vehicle with different known actual vehicle weights, trajectory types, and / or speeds. The aforementioned camera calibration calculates camera parameters by inputting a calibration board with a calibration pattern, so that a simple mathematical model (solving this mathematical model, i.e., solving for the camera parameters, including the camera's intrinsic parameters, extrinsic parameters, and distortion parameters, etc.) can be used to express the complex imaging process. This embodiment employs a traditional camera calibration method. This method requires a calibration object of known size. By establishing a correspondence between points with known coordinates on the calibration object and their corresponding image points, a specific algorithm is used to obtain the intrinsic and extrinsic parameters of the camera model. Depending on the type of calibration object, it can be divided into three-dimensional calibration objects and planar calibration objects. Three-dimensional calibration objects can be calibrated using a single image, offering high calibration accuracy, but the processing and maintenance of high-precision three-dimensional calibration objects are difficult. Planar calibration objects are simpler to manufacture than three-dimensional calibration objects, and their accuracy is easier to guarantee, but two or more images must be used for calibration. The traditional camera calibration method always requires a calibration object during the calibration process, and the manufacturing accuracy of the calibration object affects the calibration results. The specific details of the aforementioned fusion are prior art details, and the final result is as follows: Figure 3 The radar-visual fusion effect shown, for example, in the display label "car(79:45)", "car" represents the target detection category—motor vehicle, "79" represents the vehicle ID, and "45" represents the vehicle speed (in kilometers per hour). The trajectory type can include, but is not limited to, straight trajectory types and S-shaped trajectory types; the vehicle speed can include, but is not limited to, 1 km / h, 2 km / h, 5 km / h, 10 km / h, and 20 km / h. The experimental motor vehicle's state each time it passes through the weighing section area is any combination of multiple known actual vehicle weights, at least two trajectory types, and multiple vehicle speeds. Furthermore, the non-experimental motor vehicle is the target vehicle to be detected for overweight.

[0082] S3. For each motor vehicle, determine the corresponding road segment passage trajectory type based on the corresponding vehicle body marker box in the vehicle tracking result.

[0083] In step S3, the road segment trajectory types include, but are not limited to, straight-line trajectory types and S-curve trajectory types. Therefore, for a specific motor vehicle, the corresponding road segment trajectory type will be either a straight-line trajectory type or an S-curve trajectory type. Specifically, for each motor vehicle, the corresponding road segment trajectory type is determined based on the vehicle body marker box in the vehicle tracking result. This includes, but is not limited to: first, for each motor vehicle, taking the center point of the corresponding vehicle body marker box in the vehicle tracking result as the corresponding vehicle body center point, and forming the corresponding motion trajectory based on the three-dimensional coordinates of the vehicle body center point in time; then, for each motor vehicle, importing the corresponding motion trajectory into a pre-trained trajectory type recognition model based on a neural network, and outputting the corresponding road segment trajectory type. The aforementioned three-dimensional coordinates can be determined based on traditional camera calibration methods. Furthermore, the trajectory type recognition model can be conventionally trained based on a large number of trajectory samples labeled with type information (i.e., straight-line trajectory type and S-curve trajectory type).

[0084] S4. Extract the measured vehicle weight of each motor vehicle when it passes through the weighing section area from the weighing data.

[0085] In step S4, since the weighing data is also collected in real time, the measured vehicle weight can be associated with the vehicle based on the time relationship. For example, for a vehicle that passes through the weighing section area at 11:00, the weighing data collected at 11:00 can be used as the corresponding measured vehicle weight.

[0086] S5. Based on the multiple known actual vehicle weights, multiple measured vehicle weights, multiple road segment passage trajectory types, and multiple road segment passage speeds recorded by the experimental motor vehicle when it passed through the weighing section area multiple times, a mapping function between the actual vehicle weight and the measured vehicle weight, the road segment passage trajectory type, and the road segment passage speed is fitted, wherein the multiple known actual vehicle weights, the multiple measured vehicle weights, the multiple road segment passage trajectory types, and the multiple road segment passage speeds correspond one-to-one with the multiple times.

[0087] In step S5, the multiple known actual vehicle weights are mainly obtained by the experimenters through input and recording during the experiment, while other information can be automatically recorded based on the aforementioned steps S2 to S4. The specific fitting method described above is existing technology, such as fitting based on the least squares method. The final fitted mapping function can be, but is not limited to, expressed by the following formula:

[0088]

[0089] In the formula, w represents the actual vehicle weight, and f() represents the mapping function. denoted by , where t represents the road segment trajectory type, v represents the road segment speed, a1 represents the coefficient of influence of the measured vehicle weight on the actual vehicle weight obtained through fitting when the road segment trajectory type is a straight line, b1 represents the coefficient of influence of the road segment speed on the actual vehicle weight obtained through fitting when the road segment trajectory type is a straight line, c1 represents the bias coefficient obtained through fitting that affects the actual vehicle weight when the road segment trajectory type is a straight line, a2 represents the coefficient of influence of the measured vehicle weight on the actual vehicle weight obtained through fitting when the road segment trajectory type is an S-curve, b2 represents the coefficient of influence of the road segment speed on the actual vehicle weight obtained through fitting when the road segment trajectory type is an S-curve, and c2 represents the bias coefficient obtained through fitting that affects the actual vehicle weight when the road segment trajectory type is an S-curve. Furthermore, if the probability η of an S-shaped trajectory is obtained by importing the motion trajectory of each experimental vehicle into the trajectory type recognition model, then another mapping function can be finally fitted as follows:

[0090]

[0091] In the formula, w represents the actual vehicle weight, and f′() represents the other mapping function. Let represent the measured vehicle weight, t represent the road segment trajectory type, v represent the road segment speed, a3 represent the coefficient of influence of the measured vehicle weight on the actual vehicle weight (obtained through fitting), d3 represent the coefficient of influence of probability η on the actual vehicle weight (obtained through fitting), b3 represent the coefficient of influence of road segment speed on the actual vehicle weight (obtained through fitting), and c3 represent the bias coefficient that affects the actual vehicle weight (obtained through fitting).

[0092] S6. Import the measured vehicle weight, road segment trajectory type, and road segment speed of the non-experimental motor vehicle into the mapping function to obtain the actual vehicle weight of the non-experimental motor vehicle.

[0093] In step S6, if the probability η1 of a straight trajectory and the probability η2 of an S-curve trajectory are output after the motion trajectory of the non-experimental vehicle is imported into the trajectory type recognition model, then the actual weight of the non-experimental vehicle can be calculated according to the measured vehicle weight, road segment trajectory type, road segment speed, and the mapping function, using the following formula: Where η1+η2=1; or the actual weight of the non-experimental motor vehicle is calculated according to the following formula based on the measured vehicle weight, road segment trajectory type, road segment speed, and the other mapping function:

[0094] S7. Determine whether the actual weight of the non-experimental motor vehicle exceeds a preset weight threshold.

[0095] In step S7, the preset vehicle weight threshold can be specifically set according to the load-bearing design of the road surface ahead, for example, 30 tons.

[0096] S8. If so, the non-experimental motor vehicle is determined to be overweight, and an alarm is triggered.

[0097] In step S8, specifically, triggering the alarm action includes, but is not limited to, the following steps S71 to S75: S71. Extracting a video frame containing the vehicle image and vehicle body marking frame of the non-experimental motor vehicle from the vehicle tracking results; S72. Extracting the vehicle image of the non-experimental motor vehicle from the video frame and performing license plate rectangle contour detection processing on the vehicle image to obtain a license plate marking frame; S73. Extracting the license plate image from the vehicle image based on the license plate marking frame; S74. Performing character recognition processing on the license plate image using the character recognition package pytesseract to obtain a string, and using the string as the license plate number of the non-experimental motor vehicle; S75. Generating an overweight alarm message based on the license plate number and actual vehicle weight of the non-experimental motor vehicle, and uploading the overweight alarm message to the motor vehicle overweight monitoring platform. In step S72, specifically, but not limited to, the rectangular contour detection function in the cross-platform computer vision library OpenCV can be used to find the license plate marking box. This can be combined with known information such as the exact size, color, and approximate location of the license plate to improve detection accuracy. The character recognition package pytesseract is an existing OCR (Optical Character Recognition) tool. Because the overweight alarm message carries the license plate number and actual weight of the overweight vehicle, it facilitates the deployment of manpower and resources by the vehicle overweight monitoring platform for verification and punishment, effectively preventing vehicles from cheating by driving in S-shaped routes. Furthermore, after obtaining the license plate number of the non-experimental vehicle, the annual inspection weight and rated load of the non-experimental vehicle can be queried online based on the license plate number. Then, combined with the actual vehicle weight, it can be determined whether the current load of the non-experimental vehicle (i.e., the actual vehicle weight minus the annual inspection weight) exceeds the rated load. If so, it can be determined that the non-experimental vehicle has an overload problem, and an overload alarm message can be generated based on the license plate number, actual vehicle weight, and current load of the non-experimental vehicle. The overload alarm message is then uploaded to the motor vehicle overload supervision platform (which can be the same platform as the motor vehicle overweight supervision platform).

[0098] Therefore, based on the motor vehicle overweight detection method described in steps S1 to S8 above, a new scheme for motor vehicle overweight detection combining radar-visual fusion technology is provided. First, based on radar and video data, radar-visual fusion tracking is used to track each motor vehicle appearing in the weighing section area, obtaining the road segment passage trajectory type and speed of each vehicle, and extracting the measured weight of each vehicle from the weighing data. Then, based on multiple known actual vehicle weights, multiple measured vehicle weights, multiple road segment passage trajectory types, and multiple road segment passage speeds recorded by the experimental vehicle during multiple passages through the weighing section area, a mapping function is fitted to obtain the mapping function between the actual vehicle weight and the measured vehicle weight, road segment passage trajectory type, and road segment passage speed. Finally, the measured vehicle weight, road segment passage trajectory type, and road segment passage speed of the non-experimental vehicle are imported into the mapping function to obtain the actual vehicle weight of the non-experimental vehicle, and a threshold comparison method is used to determine whether it is overweight. By combining radar-visual fusion technology, this effectively prevents motor vehicles from cheating by taking S-shaped routes, facilitating practical application and promotion.

[0099] like Figure 4 As shown, the second aspect of this embodiment provides a virtual device for implementing the motor vehicle overweight detection method described in the first aspect, including a data acquisition module, a vehicle tracking module, a trajectory type determination module, a vehicle weight extraction module, a mapping function fitting module, an actual vehicle weight calculation module, an overweight judgment module, and an overweight alarm module;

[0100] The data acquisition module is used to acquire radar data collected in real time by radar on the weighing section area, video data collected in real time by camera on the weighing section area, and weighing data collected in real time by ground weighing equipment, wherein the ground weighing equipment is arranged in the weighing section area.

[0101] The vehicle tracking module is communicatively connected to the data acquisition module and is used to track each motor vehicle appearing in the weighing section area using a radar-visual fusion tracking method based on the radar data and the video data, and obtain vehicle tracking results. The vehicle tracking results include the vehicle body marking frame of each motor vehicle and the vehicle speed passing through the section. The motor vehicles include experimental motor vehicles and non-experimental motor vehicles. The experimental motor vehicle refers to a motor vehicle that passes through the weighing section area multiple times before the non-experimental motor vehicle with different known actual vehicle weights, trajectory types and / or vehicle speeds.

[0102] The trajectory type determination module is communicatively connected to the vehicle tracking module and is used to determine the corresponding road segment passage trajectory type for each motor vehicle based on the corresponding vehicle body marker box in the vehicle tracking result.

[0103] The vehicle weight extraction module is communicatively connected to the data acquisition module and is used to extract the measured vehicle weights of each motor vehicle when passing through the weighing section area from the weighing data.

[0104] The mapping function fitting module is communicatively connected to the vehicle tracking module, the trajectory type determination module, and the measured vehicle weight extraction module. It is used to fit a mapping function between the actual vehicle weight and the measured vehicle weight, the trajectory type, and the speed based on multiple known actual vehicle weights, multiple measured vehicle weights, multiple road segment passing trajectories, and multiple road segment passing speeds recorded by the experimental motor vehicle when it passes through the weighing section area multiple times. The multiple known actual vehicle weights, the multiple measured vehicle weights, the multiple road segment passing trajectories, and the multiple road segment passing speeds correspond one-to-one with the multiple times recorded.

[0105] The actual vehicle weight calculation module is communicatively connected to the vehicle tracking module, the trajectory type determination module, the measured vehicle weight extraction module, and the mapping function fitting module, respectively. It is used to import the measured vehicle weight, road segment trajectory type, and road segment speed of the non-experimental motor vehicle into the mapping function to obtain the actual vehicle weight of the non-experimental motor vehicle.

[0106] The overweight judgment module is communicatively connected to the actual vehicle weight calculation module and is used to determine whether the actual vehicle weight of the non-experimental motor vehicle exceeds a preset vehicle weight threshold.

[0107] The overweight alarm module is communicatively connected to the overweight judgment module. It is used to determine that the non-experimental motor vehicle is overweight when the actual weight of the non-experimental motor vehicle exceeds the preset weight threshold, and to trigger an alarm action.

[0108] In one possible design, the vehicle tracking module includes a radar target detection unit, a radar target tracking unit, an image target detection unit, an image target tracking unit, and a target recognition and tracking result fusion unit;

[0109] The radar target detection unit is used to detect the position, speed and azimuth of each obstacle appearing in the weighing section area based on the radar data;

[0110] The radar target tracking unit is used to track each obstacle object using an HM object tracker based on the radar data, and obtain the radar target tracking results of each obstacle object.

[0111] The image target detection unit is used to identify each target object appearing in the weighing section area based on the video data using the YOLO v4 target detection model, and to obtain the object bounding box of each target object, wherein the target object includes motor vehicles.

[0112] The image target tracking unit is used to track each target object according to the video data using the deepsort multi-target tracking algorithm, and obtain the video target tracking results of each target object;

[0113] The target recognition and tracking result fusion unit is communicatively connected to the radar target detection unit, the radar target tracking unit, the image target detection unit, and the image target tracking unit. It is used to determine at least one pair of obstacle objects and target objects that are considered the same object based on the position and azimuth of each obstacle object and the position and azimuth of each target object relative to the camera, determined by camera calibration. For each pair of obstacle objects and target objects, it fuses the speed of the corresponding obstacle object and the radar target tracking result, as well as the object marker frame and video target tracking result of the corresponding target object, to obtain the vehicle tracking result. The vehicle tracking result includes the vehicle body marker frame and the vehicle speed of each motor vehicle appearing in the weighing section area. The motor vehicles include experimental motor vehicles and non-experimental motor vehicles. The experimental motor vehicle refers to a motor vehicle that has passed through the weighing section area multiple times before the non-experimental motor vehicle with different known actual vehicle weights, trajectory types, and / or speeds.

[0114] The working process, working details and technical effects of the aforementioned device provided in the second aspect of this embodiment can be found in the motor vehicle overweight detection method described in the first aspect, and will not be repeated here.

[0115] like Figure 5 As shown, the third aspect of this embodiment provides a physical system applying the vehicle overload detection method as described in the first aspect, including a radar, a camera, a ground weighing device, and a smart roadside device. The smart roadside device is communicatively connected to the radar, the camera, and the ground weighing device. The radar is used to collect radar data of the weighing section area in real time and transmit the radar data to the smart roadside device in real time. The camera is used to collect video data of the weighing section area in real time and transmit the video data to the smart roadside device in real time. The ground weighing device is arranged in the weighing section area and is used to collect weighing data in real time and transmit the weighing data to the smart roadside device in real time. The smart roadside device is used to execute the vehicle overload detection method as described in the first aspect.

[0116] The working process, working details and technical effects of the aforementioned system provided in the third aspect of this embodiment can be found in the motor vehicle overweight detection method described in the first aspect, and will not be repeated here.

[0117] like Figure 6 As shown, the fourth aspect of this embodiment provides a computer device for executing the vehicle overweight detection method as described in the first aspect. The device includes a memory, a processor, and a transceiver connected in sequence. The memory stores a computer program, the transceiver sends and receives messages, and the processor reads the computer program and executes the vehicle overweight detection method as described in the first aspect. Specifically, the memory may include, but is not limited to, random-access memory (RAM), read-only memory (ROM), flash memory, first-in-first-out (FIFO) memory, and / or first-in-last-out (FILO) memory, etc.; the processor may include, but is not limited to, a microprocessor of the STM32F105 series. Furthermore, the computer device may also include, but is not limited to, a power module, a display screen, and other necessary components.

[0118] The working process, working details and technical effects of the aforementioned computer equipment provided in the fourth aspect of this embodiment can be found in the motor vehicle overweight detection method described in the first aspect, and will not be repeated here.

[0119] This fifth aspect of the embodiment provides a computer-readable storage medium storing instructions comprising the method for detecting overweight motor vehicles as described in the first aspect. Specifically, the computer-readable storage medium stores instructions that, when executed on a computer, perform the method for detecting overweight motor vehicles as described in the first aspect. The computer-readable storage medium refers to a data storage medium, and may include, but is not limited to, floppy disks, optical disks, hard disks, flash memory, USB flash drives, and / or Memory Sticks. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices.

[0120] The working process, working details and technical effects of the aforementioned computer-readable storage medium provided in the fifth aspect of this embodiment can be found in the motor vehicle overweight detection method described in the first aspect, and will not be repeated here.

[0121] The sixth aspect of this embodiment provides a computer program product containing instructions that, when executed on a computer, cause the computer to perform the motor vehicle overweight detection method as described in the first aspect. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device.

[0122] Finally, it should be noted that the above description is merely a preferred embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method of detecting overloading of a motor vehicle, characterized in that, include: The system acquires radar data collected in real time from the weighing section area by radar, video data collected in real time from the weighing section area by a camera, and weighing data collected in real time by ground weighing equipment, wherein the ground weighing equipment is arranged in the weighing section area. Based on the radar data and the video data, a radar-visual fusion tracking method is used to track each motor vehicle appearing in the weighing section area to obtain vehicle tracking results. The vehicle tracking results include the vehicle body marking frame and the speed of each motor vehicle passing through the section. The motor vehicles include experimental motor vehicles and non-experimental motor vehicles. The experimental motor vehicle refers to a motor vehicle that passes through the weighing section area multiple times before the non-experimental motor vehicle with different known actual vehicle weights, trajectory types and / or vehicle speeds. For each motor vehicle, the corresponding road segment trajectory type is determined based on the vehicle body marker box in the vehicle tracking result, wherein the road segment trajectory type includes a straight line trajectory type and an S-curve trajectory type; The vehicle weights of each motor vehicle collected when passing through the weighing section area are extracted from the weighing data. Based on multiple known actual vehicle weights, multiple measured vehicle weights, multiple road segment trajectory types, and multiple road segment speeds recorded by the experimental motor vehicle during multiple passages through the weighing section area, a mapping function is fitted to the actual vehicle weight and the measured vehicle weight, road segment trajectory type, and road segment speed. The multiple known actual vehicle weights, the multiple measured vehicle weights, and the multiple road segment trajectory types each correspond one-to-one with the multiple road segment speeds. The mapping function uses the formula... To indicate, Indicates the actual vehicle weight. Indicates the measurement of vehicle weight. Indicates the speed of vehicles passing through the road segment. This represents the influence coefficient of the measured vehicle weight on the actual vehicle weight, obtained through fitting, when the trajectory type of the road segment is a straight line. This represents the coefficient of influence of vehicle speed on actual vehicle weight, obtained through fitting, when the trajectory type of the road segment is a straight line. This represents the bias coefficient, obtained through fitting, that affects the actual vehicle weight when traversing a road segment with a straight-line trajectory. This represents the influence coefficient of the measured vehicle weight on the actual vehicle weight, obtained through fitting, when the trajectory type on the road segment is an S-curve. This represents the coefficient of influence of vehicle speed on actual vehicle weight, obtained through fitting, when the trajectory type of the road segment is S-curve. This represents the bias coefficient, obtained through fitting, that affects the actual vehicle weight when traversing a road segment with an S-shaped trajectory. This represents the probability that a road segment traverses a straight-line trajectory. This represents the probability that the road segment passes through an S-shaped trajectory. ; The measured vehicle weight, the probability of the road segment trajectory type, and the road segment speed of the non-experimental motor vehicle are imported into the mapping function to obtain the actual vehicle weight of the non-experimental motor vehicle. Determine whether the actual weight of the non-experimental motor vehicle exceeds a preset weight threshold; If so, the non-experimental motor vehicle is determined to be overweight, and an alarm is triggered.

2. The motor vehicle overloading detection method of claim 1, wherein, Based on the radar data and the video data, a radar-visual fusion tracking method is used to track each motor vehicle appearing in the weighing section area to obtain vehicle tracking results, including: radar target detection, radar target tracking, image target detection, image target tracking and target recognition and tracking result fusion. The radar target detection stage is used to detect the position, speed and azimuth of each obstacle appearing in the weighing section area based on the radar data; The radar target tracking step is used to track each obstacle object using an HM object tracker based on the radar data, and obtain the radar target tracking results of each obstacle object. The image target detection step is used to identify each target object appearing in the weighing section area based on the video data using the YOLO v4 target detection model, and to obtain the object bounding box of each target object, wherein the target object includes motor vehicles; The image target tracking step is used to track each target object according to the video data using the deepsort multi-target tracking algorithm, and obtain the video target tracking results of each target object; The target recognition and tracking result fusion step is used to determine at least one pair of obstacle objects and target objects that are the same object based on the position and azimuth of each obstacle object and the position and orientation of each target object relative to the camera determined by the camera calibration method. For each pair of obstacle objects and target objects, the speed and radar target tracking results of the corresponding obstacle object and the object marker box and video target tracking results of the corresponding target object are fused to obtain the vehicle tracking result. The vehicle tracking result includes the vehicle body marker box and the vehicle speed of each motor vehicle appearing in the weighing section area. The motor vehicles include experimental motor vehicles and non-experimental motor vehicles. The experimental motor vehicle refers to a motor vehicle that has passed through the weighing section area multiple times before the non-experimental motor vehicle with different known actual vehicle weights, trajectory types and / or vehicle speeds.

3. The method of claim 1, wherein, For each motor vehicle, based on the corresponding vehicle body marker frame in the vehicle tracking results, the corresponding road segment passage trajectory type is determined, including: For each motor vehicle, the center point of the corresponding vehicle body marker frame in the vehicle tracking result is taken as the corresponding vehicle body center point, and the corresponding motion trajectory is formed based on the three-dimensional coordinates of the vehicle body center point in time. For each motor vehicle, the corresponding motion trajectory is imported into a pre-trained trajectory type recognition model based on a neural network, and the corresponding road segment passage trajectory type is output.

4. The method of claim 1, wherein, The actions that trigger the alarm include: Extract video footage containing vehicle images and vehicle body marking frames of the non-experimental motor vehicle from the vehicle tracking results; The vehicle image of the non-experimental motor vehicle is extracted from the video footage, and the license plate rectangle outline is detected and processed to obtain the license plate marking frame. Based on the license plate marking frame, extract the license plate image from the vehicle image; The license plate image is processed by the character recognition package pytesseract to obtain a string, which is then used as the license plate number of the non-experimental motor vehicle. An overweight alarm message is generated based on the license plate number and actual weight of the non-experimental motor vehicle, and the overweight alarm message is uploaded to the motor vehicle overweight monitoring platform.

5. An apparatus for detecting overloading of a motor vehicle, characterized in that It includes a data acquisition module, a vehicle tracking module, a trajectory type determination module, a vehicle weight extraction module, a mapping function fitting module, an actual vehicle weight calculation module, an overweight judgment module, and an overweight alarm module; The data acquisition module is used to acquire radar data collected in real time by radar on the weighing section area, video data collected in real time by camera on the weighing section area, and weighing data collected in real time by ground weighing equipment, wherein the ground weighing equipment is arranged in the weighing section area. The vehicle tracking module is communicatively connected to the data acquisition module and is used to track each motor vehicle appearing in the weighing section area using a radar-visual fusion tracking method based on the radar data and the video data, and obtain vehicle tracking results. The vehicle tracking results include the vehicle body marking frame of each motor vehicle and the vehicle speed passing through the section. The motor vehicles include experimental motor vehicles and non-experimental motor vehicles. The experimental motor vehicle refers to a motor vehicle that passes through the weighing section area multiple times before the non-experimental motor vehicle with different known actual vehicle weights, trajectory types and / or vehicle speeds. The trajectory type determination module is communicatively connected to the vehicle tracking module and is used to determine the corresponding road segment passing trajectory type for each motor vehicle based on the corresponding vehicle body marking box in the vehicle tracking result. The road segment passing trajectory type includes straight trajectory type and S-curve trajectory type. The vehicle weight extraction module is communicatively connected to the data acquisition module and is used to extract the measured vehicle weights of each motor vehicle when passing through the weighing section area from the weighing data. The mapping function fitting module is communicatively connected to the vehicle tracking module, the trajectory type determination module, and the measured vehicle weight extraction module. It is used to fit a mapping function between the actual vehicle weight and the measured vehicle weight, the trajectory type, and the speed based on multiple known actual vehicle weights, multiple measured vehicle weights, multiple road segment passage trajectories, and multiple road segment speeds recorded by the experimental motor vehicle during multiple passages through the weighing section area. The multiple known actual vehicle weights, the multiple measured vehicle weights, and the multiple road segment passage trajectories correspond one-to-one with the multiple road segment speeds. The mapping function uses a formula... To indicate, Indicates the actual vehicle weight. Indicates the measurement of vehicle weight. Indicates the speed of vehicles passing through the road segment. This represents the influence coefficient of the measured vehicle weight on the actual vehicle weight, obtained through fitting, when the trajectory type of the road segment is a straight line. This represents the coefficient of influence of vehicle speed on actual vehicle weight, obtained through fitting, when the trajectory type of the road segment is a straight line. This represents the bias coefficient, obtained through fitting, that affects the actual vehicle weight when traversing a road segment with a straight-line trajectory. This represents the influence coefficient of the measured vehicle weight on the actual vehicle weight, obtained through fitting, when the trajectory type on the road segment is an S-curve. This represents the coefficient of influence of vehicle speed on actual vehicle weight, obtained through fitting, when the trajectory type of the road segment is S-curve. This represents the bias coefficient, obtained through fitting, that affects the actual vehicle weight when traversing a road segment with an S-shaped trajectory. This represents the probability that a road segment traverses a straight-line trajectory. This represents the probability that the road segment passes through an S-shaped trajectory. ; The actual vehicle weight calculation module is communicatively connected to the vehicle tracking module, the trajectory type determination module, the measured vehicle weight extraction module, and the mapping function fitting module, respectively. It is used to import the measured vehicle weight of the non-experimental motor vehicle, the probability of passing the trajectory type of the road segment, and the passing speed of the road segment into the mapping function to obtain the actual vehicle weight of the non-experimental motor vehicle. The overweight judgment module is communicatively connected to the actual vehicle weight calculation module and is used to determine whether the actual vehicle weight of the non-experimental motor vehicle exceeds a preset vehicle weight threshold. The overweight alarm module is communicatively connected to the overweight judgment module. It is used to determine that the non-experimental motor vehicle is overweight when the actual weight of the non-experimental motor vehicle exceeds the preset weight threshold, and to trigger an alarm action.

6. A motor vehicle overloading detection system characterised in that, It includes radar, camera, ground weighing equipment and intelligent roadside equipment, wherein the intelligent roadside equipment is communicatively connected to the radar, the camera and the ground weighing equipment respectively; The radar is used to collect radar data of the weighing section area in real time and transmit the radar data to the intelligent roadside device in real time. The camera is used to collect video data of the weighing section area in real time and transmit the video data to the intelligent roadside device in real time. The ground weighing equipment is arranged in the weighing section area to collect weighing data in real time and transmit the weighing data to the intelligent roadside equipment in real time. The intelligent roadside equipment is used to perform the motor vehicle overweight detection method as described in any one of claims 1 to 4.

7. A computer device, characterized by The device includes a memory, a processor, and a transceiver connected in sequence for communication. The memory is used to store a computer program, the transceiver is used to send and receive messages, and the processor is used to read the computer program and execute the motor vehicle overweight detection method as described in any one of claims 1 to 4.

8. A computer-readable storage medium, characterized in that The computer-readable storage medium stores instructions that, when executed on a computer, perform the motor vehicle overweight detection method as described in any one of claims 1 to 4.

Images