AI-based vehicle weighing detection method

Abstract

The present invention relates to the field of weighing technology for scales, and in particular to an AI-based method for detecting whether a vehicle has been fully weighed. In traditional vehicle full-weighing detection, infrared radiation or laser radar is mostly used to confirm whether the vehicle has been weighed, but the infrared radiation or laser radar has strict and narrow requirements on the detection range of the vehicle to be tested. Once the size of the vehicle is larger than the detection area of ​​the infrared radiation and laser radar, the signal will be blocked and unable to be detected. The present invention provides an AI-based method for detecting whether a vehicle has been fully weighed. The method calculates the distance between the vehicle and the camera by utilizing the characteristic that the tire size is proportional to the distance between the cameras at the same focal length in the image captured by the camera, and determines whether the vehicle has been fully weighed by comparing it with the preset distance between the camera and the scale. While simplifying the judgment conditions, the operation that the vehicle needs to perform in order to be fully weighed is directly determined by the distance between the vehicle and the camera.

Description

Technical Field

[0001] The present invention relates to the field of weighing technology for scales, and in particular to an AI-based method for detecting whether a vehicle is fully weighed. Background Art

[0002] In traditional vehicle weighing inspections, infrared radiation or laser radar is often used to confirm whether the vehicle has been weighed.

[0003] However, with the development of technology and the maturity of AI recognition technology, the effect of AI recognition has been greatly improved.

[0004] Infrared radar or laser radar has strict and narrow requirements for the detection range of the vehicle to be tested. Once the size of the vehicle is larger than the detection area of ​​infrared radar or laser radar, the infrared radar or laser radar will report an error and fail to detect.

[0005] On the other hand, when detecting whether the vehicle is fully weighed, it is only necessary to ensure that the vehicle is completely in the weighing area. There is no need to make high-precision measurements of the vehicle's position, which causes the infrared radiation and lidar to have overflowing functions in detecting whether the vehicle is fully weighed, resulting in a waste of resources.

[0006] AI recognition technology significantly improves the detection range while meeting the accuracy requirements for vehicle weighing, and the cost of the camera used in AI recognition technology is also lower than that of lidar.

[0007] Therefore, there is an urgent need to develop an AI-based vehicle full-scale detection method to increase the detection range and reduce the detection cost while meeting current needs. Summary of the Invention

[0008] In order to solve the problem that the laser radar in the prior art has strict and narrow requirements for the detection range of the vehicle to be tested, once the size of the vehicle is larger than the detection area of ​​the laser radar, the laser radar will report an error and fail to detect. Therefore, the present invention provides an AI-based vehicle full-scale detection method, which includes the following steps:

[0009] S11: Verify vehicle information and confirm through camera that the vehicle has arrived at the weighing location;

[0010] S12: After the vehicle arrives at the weighing location, images of the vehicle are captured using cameras in different directions. AI analyzes the images captured by the cameras and identifies the position of the tires and the contact points between the tires and the ground. The tire positions and heights relative to the weighing scale are then marked in the captured images.

[0011] S13: Use historical vehicle weighing images for recognition training to obtain the vehicle's tires

[0012] The height difference model in the image is used to calculate the relative height of the actual vehicle tires in the captured image. The position of the tires and the contact points between the tires and the ground captured by cameras in different directions are compared to determine whether the vehicle is fully weighed. If the vehicle is not fully weighed, the driver is prompted to make adjustments.

[0013] S14: After confirming that the vehicle is completely loaded onto the scale, the vehicle is weighed and the weighing result is recorded after the weighing result is stable;

[0014] S15: After weighing is completed, the gate is opened and the vehicle leaves the scale.

[0015] Specifically, the vehicle information comes from the vehicle information uploaded by the driver when registering. The vehicle information includes: the license plate number of the vehicle and the driver's information. After confirming that the image taken by the camera is consistent with the registered information, the vehicle information is displayed on the large screen for the driver of the vehicle to confirm.

[0016] Furthermore, the camera includes a group of main cameras and a group of auxiliary cameras. The main cameras are installed on both sides of the scale to capture the side images of the vehicle when it is weighed; the auxiliary cameras are installed at the exit and entrance of the scale to capture images of the vehicle entering and exiting the scale and assist in determining the position of the vehicle relative to the scale.

[0017] Furthermore, the steps to determine whether the vehicle is correctly weighed are as follows:

[0018] S21. After verifying the vehicle information, the auxiliary camera captures images of the entrance and exit of the scale, and after determining that there are no other vehicles on the scale, the gate is opened to allow the current vehicle to enter;

[0019] S22. After the vehicle enters the weighbridge area, the auxiliary camera determines that the vehicle has entered the weighbridge area and closes the gate, and after closing the gate, activates the main camera to capture and identify the vehicle;

[0020] S23. After closing the barrier, the driver of the vehicle adjusts the position of the vehicle until the vehicle is no longer recognizable in the image captured by the auxiliary camera. The main camera is used to identify the left and right images of the vehicle. The curvature angles of the left and right images of the vehicle are adjusted using a deep learning model to obtain left and right images of the vehicle at the same angle. The positions of the vehicle tires are identified from the left and right images and marked with boxes.

[0021] S24. Further analyze and calculate the height of the tires in the left and right images of the vehicle, and use AI to identify the contact points between the tires and the ground;

[0022] S25. Filter out all tires of the vehicle having contact points with the ground from the left image and the right image of the vehicle, and refer to them as valid tires;

[0023] When the vehicle cannot be identified in the image taken by the auxiliary camera, and the number of valid tires in the left image of the vehicle and the right image of the vehicle obtained after being taken and processed by the main camera is equal, it means that the vehicle has reached the correct weighing position.

[0024] Furthermore, after the vehicle is correctly weighed, the effective tire distribution pattern of the vehicle is obtained by using the height difference model, thereby determining the corresponding effective tires located at the same position in the left image and the right image of the vehicle, wherein the training data used by the height difference model includes the heights of the effective tires in the left image and the right image of different vehicles that have been completely weighed;

[0025] The formula for determining whether the valid tires in the left vehicle image and the right vehicle image correspond to each other is:

[0026] Q=α*(h1-h2)+β*(o1-o2)+ε*(d1-d2)

[0027] Wherein, Q is the comparison result. When Q is less than a specific threshold, it indicates that the group of valid tires participating in the comparison corresponds. α, β, and ε are weight parameters, and the sum of the three is 1. h1 and h2 are the heights of the group of valid tires participating in the comparison. o1-o2 are the heights of the tire axles of the group of valid tires participating in the comparison. d1-d2 are the heights of the contact points of the group of valid tires participating in the comparison with the ground.

[0028] When the comparison result is less than the threshold, it indicates that the valid tires involved in the comparison correspond to each other on the left and right and are not damaged.

[0029] Furthermore, the step of determining whether the vehicle is completely weighed based on the positional relationship between the vehicle's tires and the weighbridge is as follows:

[0030] Frame the weighbridge area in the vehicle left image and the vehicle right image;

[0031] Determine the position of the contact point between the effective tire and the ground in the left image and the right image of the vehicle according to the position of the effective tire; and confirm using AI recognition technology that the contact point between the effective tire and the ground in the left image and the right image of the vehicle is within the selected weighing scale area;

[0032] Determining the height of the effective tire of the vehicle through the left image and the right image of the vehicle;

[0033] By calculating the height difference between the effective tires corresponding to the left image and the right image of the vehicle;

[0034] The height differences of the effective tires are summed to calculate a gap value. If the gap value is greater than a preset threshold, it means that the vehicle is too close to a camera, causing one side of the vehicle's tire to be located at the edge of the weighbridge.

[0035] When the number of valid tires in the left image of the vehicle and the number of valid tires in the right image of the vehicle are equal and correspond one to one, and the difference value of the vehicle is less than a threshold, it is determined that the vehicle is fully weighed;

[0036] The calculation formula of the gap value is:

[0037]

[0038] Among them, C is the gap value, λ is the distance parameter, the size of λ is related to the distance between the scale and the main camera, h is the average height of the effective tires, x is the height of the effective tires in the left image of the vehicle, and y is the height of the effective tires in the right image of the vehicle; when the gap value is greater than the preset threshold, the left side or the right side of the vehicle is significantly closer to the main camera than the other side, and the approach distance is less than the maximum approach distance that the vehicle can reach when it is fully weighed.

[0039] Furthermore, the image captured by the main camera needs to be pre-processed before it can be used to determine whether the vehicle has been fully weighed. The image processing method is as follows:

[0040] After the vehicle is weighed and comes to rest, performing preliminary processing on the vehicle weighing image acquired by the auxiliary camera, including noise reduction and contrast optimization;

[0041] The Mask R-CNN model is used to perform AI recognition on the locations of the vehicle tires and the weighing scale, and the vehicle tires and the weighing scale are distinguished from other objects through feature extraction;

[0042] According to the outline of the vehicle, the center line of the vehicle is marked, and based on the tire edges of the vehicle, the geometric center of each tire is calculated using an ellipse fitting algorithm.

[0043] Furthermore, a method for extracting features of parameters for determining whether the vehicle is fully weighed from the processed image is as follows:

[0044] During the vehicle weighing process, an image of the vehicle generated during the weighing process is collected in real time, and the image of the vehicle weighing is corrected by perspective transformation;

[0045] Correcting the vehicle boundary lines and center lines in the image captured by the auxiliary camera using camera calibration, and establishing a rectangular coordinate system for the image;

[0046] Using the calibrated image captured by the main camera, marking the outline of the vehicle, the geometric center of the tires of the vehicle, determining the positions of the tires of the vehicle relative to the vehicle, and selecting tires located in the correct position;

[0047] The image captured by the main camera is input into the machine learning model for size transformation, the left image and the right image of the vehicle are adjusted to the left image and the right image of the vehicle with the same angle, the tires on the left and right sides of the vehicle are set in the same rectangular coordinate system, and the coordinates of the screened tires and the coordinates of the contact points between the tires and the ground are extracted; the coordinate position of the tire is used to determine the position of the tire relative to the vehicle and calculate the height of the tire relative to the ground, and the coordinates of the contact point between the tire and the ground are used to determine the position of the contact point relative to the ground.

[0048] Furthermore, the main camera is installed on both sides of the weighing scale, and the auxiliary camera is installed at the gate for entering and exiting the weighing scale area. The main camera is used to determine whether the vehicle has been weighed, and the auxiliary camera is used to assist in opening and closing the gate and ensure that only a single vehicle is weighed and to assist in determining that the vehicle is completely weighed. During the weighing process of the vehicle, the gate is controlled to be closed to prohibit other vehicles from entering the weighing scale area, ensuring that the vehicle body is in a stable state during the weighing process.

[0049] Furthermore, the method of prompting the driver of the vehicle to make adjustments includes: controlling the gate to remain closed, informing the driver of the vehicle of the analyzed reason why the vehicle is not currently fully weighed through voice, regularly checking whether the vehicle is fully weighed and updating the reason why the vehicle is not fully weighed until the vehicle is fully weighed.

[0050] Beneficial effects:

[0051] 1. The present invention eliminates interference factors such as spare tires by identifying the contact points between the vehicle's tires and the ground, ensuring that the identified tires are the tires actually used by the vehicle. At the same time, the tires on both sides of the vehicle are matched using a three-point positioning method, preventing matching errors caused by vehicles with tires of different sizes and improving recognition efficiency.

[0052] 2. The present invention calculates the distance between the vehicle and the camera by utilizing the characteristic that the tire size is proportional to the distance between the camera and the vehicle under the same focal length in the image captured by the camera. The distance between the vehicle and the camera is quickly determined by comparing the height difference between the matching tires on the left and right sides, thereby improving the recognition speed. Whether the vehicle is fully weighed can be determined by comparing it with the preset distance between the camera and the weighing scale. While simplifying the judgment conditions, the subsequent operations that the vehicle needs to perform in order to be fully weighed can be directly determined by the distance between the vehicle and the camera. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 A flow chart for detecting that a vehicle is fully weighed according to the present invention;

[0054] Figure 2 This is a step diagram of the method for determining whether a vehicle has been fully weighed according to the present invention. DETAILED DESCRIPTION

[0055] The present invention will be further described below with reference to the accompanying drawings and examples.

[0056] like Figure 1 As shown, the present invention provides an AI-based method for detecting whether a vehicle is fully weighed, comprising the following steps:

[0057] S11: Verify vehicle information and confirm through camera that the vehicle has arrived at the weighing location;

[0058] Specifically, the vehicle information comes from the vehicle information uploaded by the driver when registering. The vehicle information includes: the vehicle's license plate number and the driver's information. After confirming that the image taken by the camera is consistent with the registered information, the vehicle information is displayed on the large screen for the driver to confirm. The camera includes a set of main cameras and a set of auxiliary cameras. The main cameras are installed on both sides of the scale to capture the side image of the vehicle when it is weighed; the auxiliary cameras are installed at the exit and entrance of the scale to capture images of the vehicle entering and exiting the scale and assist in determining the vehicle's position relative to the scale.

[0059] The main camera is installed on both sides of the weighing scale, and the auxiliary camera is installed at the gate entering and exiting the weighing scale area. The main camera is used to confirm that the vehicle has been weighed, and the auxiliary camera is used to assist in opening and closing the gate and ensure that only a single vehicle is weighed and to assist in confirming that the vehicle is completely weighed. During the vehicle weighing process, the gate is controlled to close to prohibit other vehicles from entering the weighing scale area, ensuring that the vehicle body is in a stable state during the weighing process.

[0060] The steps to determine whether the vehicle is correctly weighed are:

[0061] S21. After verifying the vehicle information, the auxiliary camera captures the entrance and exit images of the scale, and after confirming that there are no other vehicles on the scale, the gate is opened to allow the current vehicle to enter;

[0062] S22. After the vehicle enters the weighbridge area, the auxiliary camera confirms that the vehicle has entered the weighbridge area and closes the gate. After closing the gate, the main camera is activated to capture and identify the vehicle.

[0063] S23. After closing the barrier, the driver adjusts the vehicle's position until the vehicle is no longer visible in the image captured by the auxiliary camera. The main camera then uses the main camera to identify the left and right sides of the vehicle. Using a deep learning model, the driver adjusts the curvature angles of the left and right sides of the vehicle to obtain left and right sides of the vehicle at the same angle. The tire positions of the vehicle are identified from the left and right sides of the vehicle and marked with boxes.

[0064] S24. Further analyze and calculate the height of the tires in the left and right images of the vehicle, and use AI to identify the contact points between the tires and the ground;

[0065] S25. Filtering all vehicle tires that have contact points with the ground from the left and right vehicle images as valid tires;

[0066] The main camera used to identify the vehicle is a pair of wide-angle cameras, which are used to shoot the left and right sides of the vehicle. For identification needs, the recognition range of the main camera includes the complete side of the vehicle to be identified, which requires the camera to have a sufficiently wide angle. However, the wide-angle camera will cause angular distortion at the edges of the captured image. In order to ensure identification accuracy, the main camera cannot be moved away from the weighing area to reduce distortion. On the other hand, if there are corners in the weighing area, the side of the vehicle in the image taken by the main camera will be bent at an angle. Therefore, the image taken by the main camera needs to be adjusted later so that the angular distortion of the left and right side images of the vehicle taken by the main camera is corrected or at the same angular distortion.

[0067] When the vehicle cannot be recognized in the image captured by the auxiliary camera and the number of valid tires in the left image of the vehicle and the right image of the vehicle captured and processed by the main camera are equal, it means that the vehicle has reached the correct weighing position.

[0068] S12: After the vehicle arrives at the weighing location, images of the vehicle are captured using cameras in different directions. AI analyzes the images captured by the cameras and identifies the position of the tires and the contact points between the tires and the ground. The tire positions and heights relative to the weighing scale are then marked in the captured images.

[0069] Some vehicles have spare tires or objects with similar appearances hanging outside the vehicle, which can be misleading when directly identifying the vehicle's tires based on their appearance. Therefore, a three-point positioning method is introduced. In addition to ensuring that the shape of the object to be identified matches that of a tire and has an axle, the object must also be in contact with the ground to be recognized as a tire. Considering that the tires of a vehicle correspond to the left and right sides, the identity of the object is further confirmed by looking for a tire in the same position in the image of the other side.

[0070] like Figure 2 As shown, after the vehicle is correctly weighed, the effective tire distribution pattern of the vehicle is obtained by using a height difference model, thereby determining the corresponding effective tires located at the same position in the left image and the right image of the vehicle. The training data used by the height difference model includes the heights of the effective tires in the left image and the right image of different vehicles that have been completely weighed;

[0071] The formula for determining whether the valid tires in the left and right vehicle images correspond to each other is:

[0072] Q=α*(h1-h2)+β*(o1-o2)+ε*(d1-d2)

[0073] Where Q is the comparison result. When Q is less than a specific threshold, it indicates that the two valid tires participating in the comparison correspond to each other. α, β, and ε are weight parameters, and the sum of the three is 1. h1 and h2 are the heights of the two valid tires participating in the comparison. o1-o2 are the heights of the tire axles of the two valid tires participating in the comparison. d1-d2 are the heights of the contact points of the two valid tires participating in the comparison with the ground.

[0074] The valid tires involved in the comparison are in a corresponding position. If the valid tires involved in the comparison are judged to be non-corresponding, it means that one of the tires is leaking or damaged relative to the other, which will cause unstable detection results on the vehicle scale and cause errors. At this time, the vehicle driver needs to be reminded to make adjustments to prevent affecting subsequent weighing. When the comparison result is less than the threshold, it means that the valid tires involved in the comparison correspond to each other and are not damaged. The specific threshold of the comparison result is determined by the requirements for recognition accuracy in the actual recognition process.

[0075] S13: Use past vehicle weighing images to train the AI ​​to identify and identify the vehicle. This allows the AI ​​to obtain a height difference model of the vehicle's tires in the image. The AI ​​then substitutes the relative height of the actual vehicle tires in the captured image into the height difference model for calculation. The AI ​​then compares the positions of the tires captured by cameras in different directions and the contact points between the tires and the ground to determine whether the vehicle is fully weighed. If the vehicle is not fully weighed, the driver is prompted to make adjustments.

[0076] Since the actual weighing environment is not stable, the height of the scale may be different at different locations. Directly calculating the height difference of the effective tires will lead to inaccurate judgment. Therefore, a height difference model is introduced. By using past weighing images, a functional relationship between the height, height difference and average height of all vehicle tires is established. The height difference model is corrected using the data generated during actual weighing, thereby offsetting the influence of the height difference of the scale on the judgment of whether the vehicle is fully weighed.

[0077] The images captured by the main camera need to be pre-processed before they can be used to determine whether the vehicle is fully weighed. The image processing method is as follows:

[0078] After the vehicle is weighed and stationary, the vehicle weighing image acquired by the auxiliary camera is preliminarily processed, including noise reduction and contrast optimization;

[0079] The Mask R-CNN model is used to perform AI recognition of the locations of vehicle tires and weighbridges, distinguishing them from other objects through feature extraction.

[0080] According to the vehicle's outline, the vehicle's centerline is marked, and based on the vehicle's tire edges, an ellipse fitting algorithm is used to calculate the geometric center of each tire.

[0081] The method for extracting features of parameters used to determine whether the vehicle is fully loaded from the processed image is:

[0082] During the vehicle weighing process, the images of the vehicle are collected in real time and corrected through perspective transformation.

[0083] Use camera calibration to correct the vehicle boundary lines and center lines in the image captured by the auxiliary camera, and establish a rectangular coordinate system for the image;

[0084] Using the image captured by the calibrated main camera, mark the approximate outline of the vehicle, the geometric center of the vehicle's tires, determine the position of the vehicle's tires relative to the vehicle, and select the tires located in the correct position as valid tires;

[0085] The image captured by the main camera is input into the machine learning model for size transformation, the left and right images of the vehicle are adjusted to vehicle left and right images with the same angle, the tires on the left and right sides of the vehicle are set in the same rectangular coordinate system, and the coordinates of the screened tires and the coordinates of the contact points between the tires and the ground are extracted; the coordinate position of the tire is used to determine the position of the tire relative to the vehicle and calculate the height of the tire relative to the ground, and the coordinates of the contact point between the tire and the ground are used to determine the position of the contact point relative to the ground.

[0086] The steps to determine whether the vehicle is fully weighed based on the positional relationship between the vehicle's tires and the scale are as follows:

[0087] Select the weighbridge area in the left and right vehicle images;

[0088] Determine the position of the contact point between the effective tire and the ground in the left and right images of the vehicle based on the position of the effective tire; use AI recognition technology to confirm that the contact point between the effective tire and the ground in the left and right images of the vehicle is within the selected weighing area;

[0089] Ensure that all the vehicle's effective tires are within the scale area by confirming that the contact point between the vehicle and the scale is within the scale area;

[0090] Determine the height of the effective tire of the vehicle through the left image and the right image of the vehicle;

[0091] By calculating the height difference between the effective tires corresponding to the left image of the vehicle and the right image of the vehicle;

[0092] The height difference of the effective tires is summed to calculate the gap value. If the gap value is greater than the preset threshold, it means that the vehicle is too close to a camera, causing one side of the vehicle's tire to be at the edge of the scale.

[0093] When the number of valid tires in the left image of the vehicle and the number of valid tires in the right image of the vehicle are equal and correspond one to one, and the difference value of the vehicle is less than a threshold, it is determined that the vehicle is fully weighed;

[0094] After ensuring that all the vehicle's valid tires are within the weighing area, it is determined that the distance between the vehicle's main cameras does not exceed the threshold, thereby ensuring that the contact point between the vehicle's valid tires and the weighing scale is not at the critical edge of the weighing scale, and further ensuring that the vehicle is completely within the weighing area, so that the weighing scale can accurately weigh the vehicle.

[0095] The calculation formula of the gap value is:

[0096]

[0097] Where C is the gap value, λ is the distance parameter, the size of λ is related to the distance between the scale and the main camera, h is the average height of the effective tire, x is the height of the effective tire in the left image of the vehicle, and y is the height of the effective tire in the right image of the vehicle;

[0098] The average height of the effective tires is calculated using a height difference model by comparing the effective tire heights and their height differences. This represents the ideal average height of all tires on the vehicle. When this difference exceeds a preset threshold, the left or right side of the vehicle is significantly closer to the primary camera than the other, and the approach distance is less than the maximum approach distance achieved when the vehicle is fully loaded.

[0099] S14: After confirming that the vehicle is completely loaded onto the scale, the vehicle is weighed and the weighing result is recorded after the weighing result is stable;

[0100] If the vehicle is not fully weighed, the driver is prompted that the vehicle is not fully weighed, and the driver is prompted to make adjustments in the following ways: controlling the gate to remain closed, informing the driver of the analyzed reason why the vehicle is not fully weighed by voice, regularly checking whether the vehicle is fully weighed and updating the reason why the vehicle is not fully weighed until the vehicle is fully weighed;

[0101] After the vehicle is fully weighed, the system will continue to check whether the vehicle is fully weighed. The vehicle cannot make any further movements until the weighing result is stable. If the vehicle moves while waiting for the weighing result to be stable or the vehicle fails to be fully weighed again, the current weighing will be stopped and the vehicle will be adjusted until it enters the fully weighed state again.

[0102] S15: After weighing is completed, the gate is opened and the vehicle leaves the scale.

[0103] After weighing is completed, the gate entering the weighing area is kept closed, and the gate leaving the weighing area is opened until neither the main camera nor the auxiliary camera can recognize features related to the vehicle. At this time, it is determined that the vehicle has left and the gate leaving the weighing area is closed, and the process returns to S11 to wait for the arrival of new vehicles.

[0104] The above-described embodiments merely represent preferred embodiments of the present invention. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that a person skilled in the art would be able to make various modifications, improvements, and substitutions without departing from the spirit of the present invention, all of which fall within the scope of protection of the present invention. Therefore, the scope of protection of the present invention shall be determined by the appended claims.

Claims

1. The AI-based vehicle full weighing detection method is characterized by: The following steps are included: S11: Verify vehicle information and confirm through camera that the vehicle has arrived at the weighing location; S12: After the vehicle arrives at the weighing location, use cameras in different directions to capture images of the vehicle. AI analyzes the images captured by the cameras and identifies the position of the tires and the contact points between the tires and the ground. The tire positions and heights relative to the weighing scale are marked in the captured images. S13: Using historical vehicle weighing images for recognition training, obtaining a height difference model of the vehicle's tires in the image, substituting the relative height of the actual vehicle tires in the captured image into the height difference model for calculation, and comparing the positions of the tires captured by cameras in different directions and the contact points between the tires and the ground to determine whether the vehicle is fully weighed; if the vehicle is not fully weighed, prompting the vehicle driver to make adjustments; S14: After confirming that the vehicle is completely loaded onto the scale, the vehicle is weighed and the weighing result is recorded after the weighing result is stable; S15: After weighing is completed, the gate is opened and the vehicle leaves the scale; All tires of the vehicle having contact points with the ground are screened out from the left image and the right image of the vehicle and are called valid tires; The steps to determine whether the vehicle is fully weighed based on the positional relationship between the vehicle's tires and the scale are as follows: Frame the weighbridge area in the vehicle left image and the vehicle right image; Determine the position of the contact point between the effective tire and the ground in the left image and the right image of the vehicle according to the position of the effective tire; and confirm using AI recognition technology that the contact point between the effective tire and the ground in the left image and the right image of the vehicle is within the selected weighing scale area; Determining the height of the effective tire of the vehicle by using the left image and the right image of the vehicle; By calculating the height difference between the effective tires corresponding to the left image and the right image of the vehicle; The height differences of the effective tires are summed to calculate a gap value. If the gap value is greater than a preset threshold, it indicates that the vehicle is too close to a camera, causing the tire on one side of the vehicle to be located at the edge of the weighbridge. When the number of valid tires in the left image of the vehicle is equal to and corresponds one to one with the number in the right image of the vehicle, and the difference value of the vehicle is less than a threshold, it is determined that the vehicle is completely weighed.

2. The AI-based vehicle full weighing detection method according to claim 1 is characterized in that: The vehicle information comes from the vehicle information uploaded by the driver when registering. The vehicle information includes: the license plate number of the vehicle and the driver's information. After confirming that the image taken by the camera is consistent with the registered information, the vehicle information is displayed on the large screen for the driver of the vehicle to confirm.

3. The AI-based vehicle full weighing detection method according to claim 2 is characterized in that: The cameras include a set of main cameras and a set of auxiliary cameras. The main cameras are installed on both sides of the scale to capture side images of the vehicle when it is weighed. Auxiliary cameras are installed at the exit and entrance of the weighbridge to capture images of the vehicle entering and exiting the weighbridge and to assist in determining the position of the vehicle relative to the weighbridge.

4. The AI-based vehicle full weighing detection method according to claim 3 is characterized in that: The steps to determine whether the vehicle is correctly weighed are: S21. After verifying the vehicle information, the auxiliary camera captures images of the entrance and exit of the scale, and after determining that there are no other vehicles on the scale, the gate is opened to allow the current vehicle to enter; S22. After the vehicle enters the weighbridge area, the auxiliary camera determines that the vehicle has entered the weighbridge area and closes the gate, and after closing the gate, activates the main camera to capture and identify the vehicle; S23. After closing the barrier, the driver of the vehicle adjusts the vehicle's position until the vehicle is no longer visible in the image captured by the auxiliary camera. The main camera then uses the main camera to identify the left and right sides of the vehicle. Using a deep learning model, the curvature angles of the left and right sides of the vehicle are adjusted to obtain left and right side images of the vehicle at the same angle. The positions of the vehicle's tires are identified from the left and right sides of the vehicle and marked with boxes. S24. Further analyze and calculate the height of the tires in the left and right images of the vehicle, and use AI to identify the contact points between the tires and the ground; S25. Filter out all tires of the vehicle having contact points with the ground from the left image and the right image of the vehicle, and refer to them as valid tires; When the vehicle cannot be identified in the image taken by the auxiliary camera and the number of valid tires in the left image of the vehicle and the right image of the vehicle obtained after being taken and processed by the main camera is equal, it means that the vehicle has reached the correct weighing position.

5. The AI-based vehicle full weighing detection method according to claim 4 is characterized in that: After the vehicle is correctly weighed, the effective tire distribution pattern of the vehicle is obtained by using the height difference model, thereby determining the corresponding effective tires located at the same position in the left and right vehicle images, wherein the training data used by the height difference model includes the heights of the effective tires in the left and right vehicle images of different vehicles that have been completely weighed; The formula for determining whether the valid tires in the left and right vehicle images correspond to each other is: ； Among them, the To compare the results, When it is less than a certain threshold, it means that the group of valid tires participating in the comparison corresponds to each other. 、 、 is the weight parameter, and the sum of the three is 1. 、 is the height of a set of effective tires involved in the comparison, is the height of the tire axle center of a set of effective tires involved in the comparison, is the height of the contact point between a set of effective tires involved in the comparison and the ground; When the comparison result is less than the threshold, it indicates that the valid tires involved in the comparison correspond to each other on the left and right and are not damaged.

6. The AI-based vehicle full weighing detection method according to claim 5 is characterized in that: The calculation formula of the gap value is: ; in, is the gap value, is the distance parameter, The size is related to the distance between the scale and the main camera. is the average height of the effective tire, is the height of the effective tire in the left image of the vehicle, is the height of the effective tire in the right image of the vehicle; when the gap value is greater than a preset threshold, the left side or the right side of the vehicle is significantly closer to the main camera than the other side, and the approach distance is less than the maximum approach distance that the vehicle can reach when it is fully weighed.

7. The AI-based vehicle full weighing detection method according to claim 6 is characterized in that: The image captured by the main camera needs to be pre-processed before it can be used to determine whether the vehicle is fully weighed. The image processing method is as follows: After the vehicle is weighed and comes to rest, performing preliminary processing on the vehicle weighing image acquired by the auxiliary camera, including noise reduction and contrast optimization; The Mask R-CNN model is used to perform AI recognition on the locations of the vehicle tires and the weighing scale, and the vehicle tires and the weighing scale are distinguished from other objects through feature extraction; According to the outline of the vehicle, the center line of the vehicle is marked, and based on the tire edges of the vehicle, the geometric center of each tire is calculated using an ellipse fitting algorithm.

8. The AI-based vehicle full weighing detection method according to claim 7 is characterized in that: The method for extracting features of parameters for determining whether the vehicle is fully loaded from the processed image is as follows: During the vehicle weighing process, an image of the vehicle generated during the weighing process is collected in real time, and the image of the vehicle weighing is corrected by perspective transformation; Correcting the vehicle boundary lines and center lines in the image captured by the auxiliary camera using camera calibration, and establishing a rectangular coordinate system for the image; Using the calibrated image captured by the main camera, marking the outline of the vehicle, the geometric center of the tires of the vehicle, determining the positions of the tires of the vehicle relative to the vehicle, and selecting tires located in the correct position; The image captured by the main camera is input into the machine learning model for size transformation, the left image and the right image of the vehicle are adjusted to the left image and the right image of the vehicle with the same angle, the tires on the left and right sides of the vehicle are set in the same rectangular coordinate system, and the coordinates of the screened tires and the coordinates of the contact points between the tires and the ground are extracted; the coordinate position of the tire is used to determine the position of the tire relative to the vehicle and calculate the height of the tire relative to the ground, and the coordinates of the contact point between the tire and the ground are used to determine the position of the contact point relative to the ground.

9. The AI-based vehicle full weighing detection method according to claim 8, characterized in that: The main camera is installed on both sides of the weighing scale, and the auxiliary camera is installed at the gate for entering and exiting the weighing scale area. The main camera is used to determine whether the vehicle has been weighed, and the auxiliary camera is used to assist in opening and closing the gate and ensure that only a single vehicle is weighed and to assist in determining that the vehicle is completely weighed. During the vehicle weighing process, the gate is controlled to close to prohibit other vehicles from entering the weighing scale area, ensuring that the vehicle body is in a stable state during the weighing process.

10. The AI-based vehicle full weighing detection method according to claim 9, characterized in that: The method of prompting the driver of the vehicle to make adjustments includes: controlling the gate to remain closed, informing the driver of the vehicle of the analyzed reason why the vehicle is not fully weighed through voice, regularly checking whether the vehicle is fully weighed and updating the reason why the vehicle is not fully weighed until the vehicle is fully weighed.

Images