A method and system for reconstructing bridge vehicle load distribution based on machine vision
By combining machine vision and bridge deck targets, the problem of difficult identification of bridge vehicle load distribution was solved, and high-precision identification and assessment of bridge vehicle load spatial distribution was achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- 中电建路桥集团有限公司
- Filing Date
- 2026-04-23
- Publication Date
- 2026-06-02
Smart Images
Figure CN122135318A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of machine vision measurement technology, and in particular to a method and system for reconstructing bridge vehicle load distribution based on machine vision. Background Technology
[0002] In bridge engineering, vehicle load is one of the key loads affecting the load-bearing capacity and reliability of bridges. Measuring vehicle load is crucial for the actual design, evaluation, maintenance, and life-cycle analysis of bridge structures. However, vehicle load is a complex random variable, and its spatiotemporal distribution fluctuates. Therefore, obtaining an accurate spatial probability distribution of vehicle load is essential for achieving precise spatiotemporal modeling of vehicle load and calculating its effects on bridge loads.
[0003] Currently, Pavement Weigh-In-Motion (PWIM) technology is widely used in transportation and civil engineering, providing reliable data (such as total vehicle weight, axle load, number of axles, and travel time) unaffected by traffic disturbances. However, most existing bridges are not equipped with Pavement-In-Motion (WIM) systems. Furthermore, the data acquired by WIM systems only reflects the load characteristics of specific cross-sections and cannot reveal the precise locational distribution of vehicle loads across the entire bridge surface.
[0004] In recent years, driven by the rapid development of computer vision and machine learning, numerous technical solutions have begun to utilize traffic monitoring camera systems to acquire vehicle location information across the entire bridge deck, and combine this with PWIM systems to obtain the traffic load distribution across the entire bridge deck. However, although the aforementioned PWIM-based methods have demonstrated reliable and accurate performance in many solutions, they still have significant limitations. Specifically, the PWIM must be embedded within the bridge's cross-sectional structure, thus requiring the integration of PWIM data from bridge toll stations with machine vision technology to obtain the spatiotemporal distribution of vehicle loads. However, such solutions primarily focus on acquiring the spatiotemporal distribution of vehicle loads on the bridge deck, making it difficult to identify the long-term vehicle load distribution of the bridge.
[0005] Machine vision technology can be used not only to identify vehicles but also to perform traffic flow analysis and vehicle counting. Although some technologies currently integrate machine vision and PWIM data to perform statistical analysis of the spatial distribution of vehicle loads, they still lack analysis of the distribution of vehicle position and weight within lanes and the estimation of vehicle loads to obtain the spatial distribution of vehicle loads across the entire bridge deck. Summary of the Invention
[0006] The purpose of this invention is to provide a method and system for reconstructing bridge vehicle load distribution based on machine vision, so as to achieve accurate modeling of traffic flow load and subsequent detailed bridge condition assessment.
[0007] To achieve the above objectives, this invention provides a method for reconstructing bridge vehicle load distribution based on machine vision, comprising the following steps: Using a video monitoring system deployed at the cross section of the road dynamic weighing system, video image data of traffic flow load covering the current cross section is collected, and the weight information and passage time of each vehicle obtained by the road dynamic weighing system are also collected. A machine vision-based dual-target detection model is used to detect vehicle targets in traffic flow load video image data, obtain the detection box of each vehicle's front and the overall detection box, and record the video timestamp of each vehicle passing through the current section; The horizontal coordinate of the bottom of the detection frame of each vehicle's front end and the vertical coordinate of the overall detection frame are selected as the precise position of each vehicle in the traffic flow load video image data. Using video timestamps and passage times as synchronization benchmarks, establish the correspondence between each vehicle and its weight information, and assign weight information to each vehicle; By using bridge targets that are evenly spaced and arranged on the bridge surface in advance, the projection transformation relationship between the image coordinate system and the bridge surface coordinate system is determined, and the precise position of each vehicle in the traffic flow load video image data is transformed to the bridge surface coordinate system to obtain the precise position of each vehicle on the bridge surface. Based on the precise location and corresponding weight information of each vehicle on the bridge deck, a statistical analysis is performed on the vehicle load distribution at the cross-section where the dynamic weighing system is located to obtain the statistical distribution of vehicle load on the bridge cross-section.
[0008] Preferably, the dual-target detection model is a dual-target detection model based on YOLOv11.
[0009] Preferably, the video surveillance system is arranged as follows: an L-shaped pole symmetrically positioned along the longitudinal centerline of the bridge deck is erected on the roadside at the location of the dynamic weighing system section, and the camera is positioned on the L-shaped pole directly above the road surface.
[0010] Preferably, the camera in the video surveillance system has a resolution of 1920×1080 and a sampling frequency of 25Hz.
[0011] Preferably, the road surface dynamic weighing system adopts an existing road surface dynamic weighing system on the same route as the bridge, used to obtain the total weight of passing vehicles, the number of lanes, and the passage time.
[0012] Preferably, when determining the projection transformation relationship using the bridge surface target, perspective projection transformation or homography matrix is used for coordinate transformation.
[0013] Preferably, the statistical distribution of vehicle loads on the bridge cross-section includes: the statistical distribution of the frequency of vehicle occurrence at each location on the bridge cross-section, and the statistical distribution of the average weight of vehicle loads at each location.
[0014] This invention provides a machine vision-based bridge vehicle load distribution reconstruction system for implementing the aforementioned machine vision-based bridge vehicle load distribution reconstruction method, comprising: The video surveillance system is used to collect video image data of traffic flow load at the cross-section where the dynamic weighing system for the road surface is located; The road dynamic weighing system is used to obtain the weight information and passage time of vehicles passing through the current section; Bridge surface targets are evenly spaced on the bridge surface to achieve coordinate transformation between the image coordinate system and the bridge surface coordinate system; The computing platform is equipped with a machine vision-based dual-target detection model and a cross-sectional traffic flow load statistical analysis model, which are used to identify and locate vehicle targets in video images, assign weight information to corresponding vehicles through spatiotemporal registration, and perform statistical analysis on the vehicle load distribution at the cross-section.
[0015] Preferably, the machine vision model is a dual-target vehicle detection model based on YOLOv11, used to simultaneously identify the detection boxes at the front of each vehicle and the overall detection box.
[0016] The present invention provides an electronic device, including a memory and a processor. The memory stores a computer program, and when the processor calls the computer program in the memory, it implements the steps of the above-described machine vision-based bridge vehicle load distribution reconstruction method.
[0017] In summary, the machine vision-based bridge vehicle load distribution reconstruction method and system provided by this invention offers the following advantages compared to traditional technologies: Utilizing data from video surveillance systems and dynamic weighing systems on the road surface along the same traffic route of the bridge, the spatial distribution of bridge vehicle loads can be identified with high precision and lightweight design, eliminating the need for embedded weighing systems on the bridge deck; statistical analysis of the spatial distribution of vehicle loads at the cross-sections of the dynamic weighing systems along the same route is sufficient, without the need to deploy multiple cameras along the bridge, to statistically represent the vehicle load distribution at each cross-section of the bridge; the dual-target vehicle detection model can simultaneously detect the entire vehicle and its front end, and coordinate transformation is performed using bridge deck targets to obtain the precise position of the vehicle on the bridge deck; this lays the foundation for developing accurate and reliable traffic load models that can be used for subsequent detailed bridge condition assessments.
[0018] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description
[0019] Figure 1 This is a flowchart illustrating the bridge vehicle load distribution reconstruction method based on machine vision in this invention. Figure 2 This is a schematic diagram showing the arrangement of various components in the machine vision-based bridge vehicle load distribution reconstruction system in an embodiment of the present invention. Figure 3 This is a network diagram of a YOLOv11-based dual-target vehicle detection model in an embodiment of the present invention; Figure 4 These are key points representing the precise location of the vehicle in the embodiments of the present invention; Figure 5 This is a schematic diagram of the detection results and detection area of two targets on a vehicle in an embodiment of the present invention; Figure 6 This is a schematic diagram illustrating the relationship between the image coordinate system, camera coordinate system, and world coordinate system in an embodiment of the present invention; Figure 7 This is a schematic diagram showing the statistical distribution of vehicle frequency at various locations on a bridge cross-section according to an embodiment of the present invention. Figure 8 This is a schematic diagram showing the statistical distribution of the average weight of vehicle loads at various locations along the bridge cross-section according to an embodiment of the present invention.
[0020] Figure label: 1. Video surveillance system; 101. Camera; 2. Road surface dynamic weighing system; 3. Bridge deck target; 4. Calculation platform; 5. L-shaped pole. Detailed Implementation
[0021] The technical method of the present invention will be further described below with reference to the accompanying drawings and embodiments. It should be noted that, unless otherwise specifically stated, the relative arrangement, numerical expressions, and values of the components and steps described in these embodiments do not limit the scope of this application.
[0022] The following description of at least one exemplary embodiment is merely illustrative and is in no way intended to limit the scope of this application and its application or use.
[0023] Techniques, systems, and equipment known to those skilled in the art may not be discussed in detail, but where appropriate, they should be considered part of the instruction manual.
[0024] In all the examples shown and discussed herein, any specific values should be interpreted as merely exemplary and not as limitations. Therefore, other examples of exemplary embodiments may have different values.
[0025] Unless otherwise defined, the technical or scientific terms used in this invention shall have the ordinary meaning as understood by one of ordinary skill in the art to which this invention pertains.
[0026] This invention provides a machine vision-based method for reconstructing bridge vehicle load distribution, such as... Figure 1 As shown, it includes the following steps: S1. Utilize a video surveillance system deployed at the cross-section of the road surface dynamic weighing system to collect video image data of traffic flow load covering the current cross-section, and collect vehicle weight information and passage time information obtained by the road surface dynamic weighing system. The video surveillance system is deployed as follows: L-shaped poles symmetrically positioned along the longitudinal centerline of the bridge deck are erected on the roadside at the cross-section of the road surface dynamic weighing system. Cameras are positioned on the L-shaped poles directly above the road surface, with a resolution of 1920×1080 and a sampling frequency of 25Hz. The road surface dynamic weighing system adopts the existing road surface dynamic weighing system along the same route as the bridge, used to obtain the total weight of passing vehicles, the number of lanes, and passage time.
[0027] S2. A machine vision-based dual-object detection model is used to detect vehicle targets in traffic flow load video image data, obtaining the detection bounding boxes for each vehicle's front and the overall detection bounding box, and recording the video timestamp of each vehicle passing through the current section. The dual-object detection model is a YOLOv11-based model.
[0028] S3. Select the horizontal coordinate of the bottom of the detection frame of each vehicle's front and the vertical coordinate of the overall detection frame as the precise position of each vehicle in the traffic flow load video image data.
[0029] S4. Using video timestamps and passage times as synchronization benchmarks, establish the correspondence between each vehicle and its weight information, and assign weight information to each vehicle.
[0030] S5. Using bridge surface targets that are evenly spaced and pre-arranged on the bridge surface, determine the projection transformation relationship between the image coordinate system and the bridge surface coordinate system. Transform the precise position of each vehicle in the traffic flow load video image data to the bridge surface coordinate system to obtain the precise position of each vehicle on the bridge surface. When determining the projection transformation relationship using the bridge surface targets, perspective projection transformation or homography matrix is used for coordinate transformation.
[0031] S6. Based on the precise location and corresponding weight information of each vehicle on the bridge deck, perform statistical analysis on the vehicle load distribution at the cross-section where the dynamic weighing system is located to obtain the statistical distribution of vehicle load on the bridge cross-section. The statistical distribution of vehicle load on the bridge cross-section includes: the statistical distribution of the frequency of vehicle occurrence at each location on the bridge cross-section, and the statistical distribution of the average weight of vehicle load at each location.
[0032] This invention provides a machine vision-based bridge vehicle load distribution reconstruction system for implementing the aforementioned machine vision-based bridge vehicle load distribution reconstruction method, comprising: The video surveillance system is used to collect video image data of traffic flow load at the cross-section where the dynamic weighing system for the road surface is located.
[0033] The road dynamic weighing system is used to obtain vehicle weight information and passage time information for vehicles passing through the current section.
[0034] Bridge surface targets are evenly spaced on the bridge surface to achieve coordinate transformation between the image coordinate system and the bridge surface coordinate system.
[0035] The computing platform is equipped with a machine vision-based dual-target detection model and a cross-sectional traffic flow load statistical analysis model, which are used to identify and locate vehicle targets in video images, assign weight information to corresponding vehicles through spatiotemporal registration, and perform statistical analysis on the vehicle load distribution at the cross-section.
[0036] In an exemplary embodiment of the present invention, a bridge vehicle load distribution reconstruction system based on machine vision is provided, such as... Figure 2 As shown, it includes: a video surveillance system 1, a road surface dynamic weighing system 2, a bridge deck target 3, and a computing platform 4 equipped with a machine vision-based dual-target detection model and a cross-sectional traffic flow load statistical analysis model.
[0037] The video monitoring system 1 consists of a camera at the dynamic weighing system 2 on the same route as the target bridge. This camera acquires video image data of traffic flow loads at the cross-section covering the WIM system. The camera has a resolution of 1920×1080 and a sampling frequency of 25Hz. The dynamic weighing system 2 utilizes the PWIM system on the same route as the target bridge to obtain vehicle weight information on the bridge. Bridge surface targets 3 are evenly distributed at equal intervals on the bridge surface to transform the vehicle positions in the traffic flow load video image data from the image coordinate system to the bridge surface coordinate system. The computing platform 4 can be deployed near the power distribution box of the target bridge. The dual-target detection model based on YOLOv11 and the cross-sectional traffic flow load statistical analysis model deployed on the computing platform are used to accurately identify and locate vehicle targets in traffic flow load video image data based on machine vision algorithms. Through precise spatiotemporal registration technology, the vehicle load data of each vehicle measured by PWIM are assigned to the vehicle targets at the corresponding spatiotemporal locations. Then, the cross-sectional traffic flow load statistical analysis model is used to perform statistical analysis on the vehicle load distribution of the cross section where the road dynamic weighing system is located, and obtain the statistical distribution of vehicle load on the bridge cross section.
[0038] The camera 101 in the video surveillance system is positioned on an L-shaped pole 5 erected symmetrically along the longitudinal centerline of the bridge deck at the PWIM system cross-section location. The camera 101 is positioned directly above the road surface on the L-shaped pole 5. It provides video image data of traffic flow load at the PWIM system cross-section location. The road dynamic weighing system 2 can use the PWIM system of the same route as the target bridge to obtain vehicle weight information and corresponding passage times for vehicles crossing the bridge. The bridge deck targets 3 are evenly spaced on the bridge deck... Figure 2 The target shown is used to transform the vehicle position in the traffic flow load video image data from the image coordinate system to the bridge surface coordinate system. Data from the video surveillance system 1 and the road dynamic weighing system 2 are connected to the data acquisition system (industrial control computer) via a communication cable through a specific interface, and then wirelessly transmitted to the computing platform 4, which can be deployed on both the edge and cloud, via a 4G network communication system. The data acquisition system (industrial control computer) provides stable data acquisition for a micro-host equipped with a CPU processor, while the computing platform 4 is a computer host equipped with a CPU and a GPU graphics card for training a YOLOv11-based dual-target vehicle detection model.
[0039] A machine vision-based method for reconstructing bridge vehicle load distribution employs methods such as... Figure 2 The machine vision-based bridge vehicle load distribution reconstruction system shown includes the following steps: S1. Using the video monitoring system 1 located at the cross section of the road dynamic weighing system 2, collect video image data of traffic flow load covering the current cross section, and collect the total weight of each vehicle, the number of lanes occupied, and the passage time obtained by the road dynamic weighing system.
[0040] S2. A YOLOv11-based dual-target vehicle detection model is used to detect vehicle targets in traffic flow load video image data. The model acquires the front bounding boxes and overall bounding boxes for each vehicle and records the video timestamp of each vehicle passing through the current cross-section. A network diagram of the dual-target vehicle detection model is shown below. Figure 3As shown, the YOLOv11 vehicle dual-target detection network adopts a three-level architecture: Backbone-Neck-Head. After inputting a 416×416 vehicle image, the backbone network extracts multi-scale features through the CBS, C3k2, SPPF, and C2PSA modules. The Neck part uses FPN+PAN bidirectional feature pyramids to fuse shallow details and deep semantic information. Finally, three detection heads (13×13, 26×26, and 52×52) output the bounding box positions and category predictions for the vehicle's front and overall dimensions in parallel, achieving accurate detection of multi-scale vehicle dual targets. Specifically, the CBS module is a convolutional block; the C3k2 module is a cross-stage partial bottleneck module (with 3 convolution layers and a kernel size of 2); the SPPF module is a fast spatial pyramid pooling module; and the C2PSA module is a cross-stage partial self-attention module. All of the above modules are implemented using the standard YOLOv11 network architecture.
[0041] S3. Select the horizontal coordinate of the bottom of the detection frame of each vehicle's front and the vertical coordinate of the overall vehicle detection frame as the precise position of each vehicle in the traffic flow load video image data, such as... Figure 4 coordinates in x fc , y bc As shown in the figure, where x fc The horizontal coordinate of the bottom of the vehicle front detection frame is the overall detection frame. y bc The vertical axis is denoted by . Figure 4 In the text, Car 0.95 indicates that the overall vehicle monitoring accuracy of the YOLO algorithm (YOLOv11) is 0.95, and Car Front 0.82 indicates that the vehicle front monitoring accuracy of the YOLO algorithm is 0.82.
[0042] S4. Using video timestamps and passage times as synchronization benchmarks, establish the correspondence between each vehicle and its total weight, assign weight information to each vehicle, and record it simultaneously. Figure 5 This is a diagram showing the dual-target detection results for vehicles on the bridge and the detection area. Figure 5In the diagram, Car 0.95 indicates that the overall accuracy of the YOLO algorithm in vehicle monitoring is 0.95, Car Front 0.82 indicates that the accuracy of the YOLO algorithm in vehicle front monitoring is 0.82, and Car Front 0.81 indicates that the accuracy of the YOLO algorithm in vehicle front monitoring is 0.81.
[0043] S5. Using bridge surface targets 3 that are pre-arranged at equal intervals on the bridge surface, combined with... Figure 6 The image coordinate system shown xo´y (in, o´ The origin of the image. x The horizontal axis of the image coordinate system is... y (Image coordinate system ordinate) and camera coordinate system O c - X c Y c Z c (in, O c For the camera optical center, X c The horizontal axis of the camera coordinate system is... Y c The vertical axis of the camera coordinate system Z c (The vertical axis of the camera coordinate system) and the world coordinate system (bridge surface coordinate system) with the bridge surface as the reference. O w - X w Y w Z w ,in, O w The origin of the bridge deck, X w The horizontal axis of the bridge deck coordinate system is... Y w The vertical axis of the bridge deck coordinate system is denoted as . Z w The relationship between the vertical axis of the bridge deck coordinate system and the projection transformation relationship between the image coordinate system and the bridge deck coordinate system is determined by the relationship between them. Figure 6 In P I 、P c Each is a point in the real world P w The corresponding point is the same in the image coordinate system and the camera coordinate system. The precise position of each vehicle in the traffic flow load video image data is transformed to the bridge deck coordinate system to obtain the precise position of each vehicle on the bridge deck.
[0044] S6. Based on the precise location and corresponding weight information of each vehicle on the bridge deck, perform statistical analysis on the vehicle load distribution at the cross section where the dynamic weighing system 2 is located to obtain the statistical distribution of vehicle load on the bridge cross section. Figure 7 and Figure 8 These are the statistical distribution results of vehicle frequency at different locations along the bridge cross-section and the statistical distribution results of average vehicle load weight, respectively. From... Figure 8 It can be seen that vehicles in lanes 1 and 4, closer to the edge of the bridge, generally travel near the center line of their lanes. In lanes 2 and 3, closer to the bridge towers, vehicles tend to travel closer to the edge of the lanes. This is likely related to drivers' habitual tendency to maintain a slight distance from the bridge towers. Figure 8 The average weight of vehicles traveling in each lane on the bridge cross section is basically the same, and is around 3000 kg. This indicates that in reality, the vehicles traveling on this bridge are mainly cars.
[0045] The present invention provides an electronic device, including a memory and a processor. The memory stores a computer program, and when the processor calls the computer program in the memory, it implements the steps of the above-mentioned machine vision-based bridge vehicle load distribution reconstruction method.
[0046] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the technical solutions of the present invention, and these modifications or equivalent substitutions cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.
Claims
1. A method for reconstructing bridge vehicle load distribution based on machine vision, characterized in that, Includes the following steps: Using a video monitoring system deployed at the cross section of the road dynamic weighing system, video image data of traffic flow load covering the current cross section is collected, and the weight information and passage time of each vehicle obtained by the road dynamic weighing system are also collected. A machine vision-based dual-target detection model is used to detect vehicle targets in traffic flow load video image data, obtain the detection box of each vehicle's front and the overall detection box, and record the video timestamp of each vehicle passing through the current section; The horizontal coordinate of the bottom of the detection frame of each vehicle's front end and the vertical coordinate of the overall detection frame are selected as the precise position of each vehicle in the traffic flow load video image data. Using video timestamps and passage times as synchronization benchmarks, establish the correspondence between each vehicle and its weight information, and assign weight information to each vehicle; By using bridge targets that are evenly spaced and arranged on the bridge surface in advance, the projection transformation relationship between the image coordinate system and the bridge surface coordinate system is determined, and the precise position of each vehicle in the traffic flow load video image data is transformed to the bridge surface coordinate system to obtain the precise position of each vehicle on the bridge surface. Based on the precise location and corresponding weight information of each vehicle on the bridge deck, a statistical analysis is performed on the vehicle load distribution at the cross-section where the dynamic weighing system is located to obtain the statistical distribution of vehicle load on the bridge cross-section.
2. The bridge vehicle load distribution reconstruction method based on machine vision according to claim 1, characterized in that, The dual-target detection model is a dual-target detection model based on YOLOv11.
3. The bridge vehicle load distribution reconstruction method based on machine vision according to claim 1, characterized in that, The video surveillance system is arranged as follows: L-shaped poles symmetrical along the longitudinal centerline of the bridge deck are erected on the roadside at the cross-section of the dynamic weighing system, and the camera is placed on the L-shaped pole directly above the road surface.
4. The bridge vehicle load distribution reconstruction method based on machine vision according to claim 3, characterized in that, The camera in the video surveillance system has a resolution of 1920×1080 and a sampling frequency of 25Hz.
5. The bridge vehicle load distribution reconstruction method based on machine vision according to claim 1, characterized in that, The road surface dynamic weighing system adopts the existing road surface dynamic weighing system along the same route as the bridge, and is used to obtain the total weight of passing vehicles, the number of lanes, and the passage time.
6. The bridge vehicle load distribution reconstruction method based on machine vision according to claim 1, characterized in that, When using bridge deck targets to determine projection transformation relationships, perspective projection transformation or homography matrix is used for coordinate transformation.
7. The bridge vehicle load distribution reconstruction method based on machine vision according to claim 1, characterized in that, The statistical distribution of vehicle loads on the bridge cross-section includes: the statistical distribution of the frequency of vehicle occurrence at each location on the bridge cross-section, and the statistical distribution of the average weight of vehicle loads at each location.
8. A bridge vehicle load distribution reconstruction system based on machine vision, characterized in that, The method for reconstructing bridge vehicle load distribution based on machine vision as described in any one of claims 1-7 includes: The video surveillance system is used to collect video image data of traffic flow load at the cross-section where the dynamic weighing system for the road surface is located; The road dynamic weighing system is used to obtain the weight information and passage time of vehicles passing through the current section; Bridge surface targets are evenly spaced on the bridge surface to achieve coordinate transformation between the image coordinate system and the bridge surface coordinate system; The computing platform is equipped with a machine vision-based dual-target detection model and a cross-sectional traffic flow load statistical analysis model, which are used to identify and locate vehicle targets in video images, assign weight information to corresponding vehicles through spatiotemporal registration, and perform statistical analysis on the vehicle load distribution at the cross-section.
9. The bridge vehicle load distribution reconstruction system based on machine vision according to claim 8, characterized in that, The machine vision model is a dual-target vehicle detection model based on YOLOv11, used to simultaneously identify the detection boxes at the front of each vehicle and the overall detection box.
10. An electronic device comprising a memory and a processor, wherein the memory stores a computer program, and the processor, when invoking the computer program in the memory, implements the steps of the machine vision-based bridge vehicle load distribution reconstruction method according to any one of claims 1-7.