A dynamic vehicle weighing system for toll stations based on multi-field thermal imaging technology

Abstract

This invention discloses a dynamic vehicle weighing system for toll stations based on multi-field-of-view thermal imaging technology, comprising an activation module, a first load prediction module, a second load prediction module, and a correction module. The first load prediction module calculates the first load of each tire based on the tire's mechanical deformation parameters and the tire size and pressure information sent by the activation module. The second load prediction module includes an infrared thermal imaging device and a temperature-sensitive material weighing device located above the lane, and calculates the second load of each tire by combining tire pressure information and the contact area between a single tire and the ground. The correction module performs temperature and hardness coefficient corrections on the first and second loads, and calculates the final measurement result of the tire load based on the corrected first and second loads. This invention can improve the efficiency and accuracy of tire load and vehicle weight measurement, enhance the system's all-weather usability, and reduce its cost and subsequent maintenance costs.

Description

Technical Field

[0001] This invention belongs to the field of image recognition weighing technology, and in particular relates to a dynamic vehicle weighing system for toll stations based on multi-field thermal imaging technology. Background Technology

[0002] Statistics show that over 80% of road traffic accidents involving heavy trucks are caused by overloading. Therefore, accurately and promptly obtaining vehicle tire load and vehicle load information is crucial.

[0003] Currently, mainstream vehicle weighing technologies include static weighbridges, bridge dynamic weighing systems, and road dynamic weighing systems. With the rapid development of transportation informatization and intelligence, newer image recognition-based vehicle weighing technologies have also emerged. Their problems are as follows:

[0004] 1. Static weighbridges require the installation of weighing service stations, which are complex to construct, have high installation and operating costs, require frequent maintenance, are slow and inefficient, and can easily cause traffic congestion during peak hours.

[0005] 2. The bridge dynamic weighing system requires the installation of sensors on the bridge. The measurement results are greatly affected by the sensor placement location, which necessitates specific sensor placement, making subsequent maintenance and replacement difficult. Furthermore, this method weighs vehicles as they travel across the bridge, failing to prevent overloaded vehicles from crossing.

[0006] 3. The road dynamic weighing system requires sensors to be installed on the road surface to measure the instantaneous dynamic force when a moving vehicle passes the sensor. The static axle load is then obtained through signal processing, thus determining the vehicle's weight. This system requires traffic interruption and the excavation of concrete pits or grooves in the road surface during installation and subsequent maintenance, resulting in low efficiency. Furthermore, its measurement accuracy is significantly affected by the road surface stiffness and unevenness.

[0007] 4. Existing image recognition-based vehicle weighing technologies have low measurement accuracy under nighttime conditions, making it difficult to operate efficiently around the clock.

[0008] In summary, how to improve the efficiency and accuracy of tire load and vehicle load measurement, enhance the system's all-weather usability, and reduce its cost and subsequent maintenance costs are technical problems that urgently need to be solved by those skilled in the art. Summary of the Invention

[0009] Technical problem to be solved: This invention provides a dynamic vehicle weighing system and method for toll stations based on multi-field thermal imaging technology, which improves the efficiency and accuracy of tire load and vehicle weight measurement, enhances the system's all-weather usability, and reduces its cost and subsequent maintenance costs.

[0010] Technical solution:

[0011] A dynamic vehicle weighing system for toll stations based on multi-field thermal imaging technology, the dynamic vehicle weighing system for toll stations includes an activation module, a first load prediction module, a second load prediction module and a correction module;

[0012] The activation module is used to capture real-time images of the toll station lanes, extract tire images from them, process the extracted tire images to obtain the tire size and tire pressure information, and simultaneously activate the first load prediction module and the second load prediction module.

[0013] The first load prediction module includes infrared thermal imaging devices located on both sides of the lane, a deformation parameter calculation unit, and a load prediction model trained based on the XGBoost model. The infrared thermal imaging devices are used to capture temperature matrix data of the sidewalls of the vehicle's single-axle tires. The deformation parameter calculation unit calculates the mechanical deformation parameters of the tires based on the temperature matrix data of the single-axle tires. The load prediction model calculates the first load F of each tire based on the mechanical deformation parameters of the tires sent by the deformation parameter calculation unit and the tire size and tire pressure information sent by the activation module. a ;

[0014] The second load prediction module includes an infrared thermal imaging device and a temperature-sensitive material weighing device located above the lane. It records the number of contact areas between the tires and the ground, the contact area of ​​a single tire with the ground, calculates the number of vehicle axles based on the number of contact areas, and calculates the second load F for each tire by combining tire pressure information and the contact area of ​​a single tire with the ground. b ;

[0015] The correction module is used to adjust the first load F. a Second load F b Temperature and hardness coefficient corrections are applied, and the final measurement result of the tire load is calculated based on the corrected first and second loads.

[0016] Furthermore, the activation module includes a triggering device, an optical camera, a target detection model, an OCR text recognition model, and a tire information extraction unit;

[0017] The triggering device is located within the lane area and is used to detect whether a vehicle has appeared in the lane area; the optical camera is located on the side of the toll station lane and is used to capture real-time images of the toll station lane area; the target detection model is built based on the YOLO-v5 network and is used to detect the vehicle position in the real-time image captured by the optical camera, and to capture all tire images contained in the vehicle, and send the tire images to the OCR text recognition model, which identifies the tire model; the tire information extraction unit finds the tire size and tire pressure information according to the tire model sent by the OCR text recognition model, and at the same time activates the first load prediction module and the second load prediction module.

[0018] Furthermore, the deformation parameter calculation unit processes the pixel temperature matrix data based on the OpenCV image processing algorithm, and detects the mechanical deformation parameters of the target sample tire through geometric fitting and region growing algorithms of image iteration. The mechanical deformation parameters include: the maximum pixel radius R of the tire, the maximum pixel area S1 of the tire, the pixel radius r of the wheel hub, the pixel length l of the tire-ground contact, the pixel distance h from the tire center to the ground, the equivalent pixel area S2 of the tire after deformation, the difference in image pixel area ΔS before and after tire deformation, and the pixel length L of the tire-ground contact dividing line.

[0019] The process by which the deformation parameter calculation unit detects and obtains the mechanical deformation parameters of the target sample tire includes the following steps:

[0020] S11, Generate a temperature image that is linearly related to temperature based on pixel temperature matrix data; Use the Sobel edge detection operator to calculate the pixel gradient magnitude of the temperature image, and perform image segmentation around pixels in the temperature image where the temperature difference is greater than a preset temperature difference threshold.

[0021] S12, retain the first 5% of the gradient magnitude of the segmented image, and take out the point with the largest gradient magnitude and mark it with color; select the color-marked pixels sequentially from the bottom to the top of the image, and use the selected pixels as seed points for the first fitting of the outer contour of the tire. All the selected seed points are pixels on the interface between the tire and the air.

[0022] S13, using the seed points selected in step S12, the region growing algorithm is used to grow the seed points to find the points adjacent to the seed points as new seed points, and the second fitting of the tire outer contour is performed.

[0023] S14. Retain the seed points selected in steps S12 and S13. Using the tire outer contour of the second fitting as a reference, find the pixel point with the maximum gradient magnitude in the upper half of the tire from top to bottom, and perform the third tire outer contour fitting.

[0024] S15. Using the tire outer contour fitted in step S14 as a reference, find the pixel point with the largest gradient magnitude towards the center of the tire. Use this point as the seed point for fitting the outer contour of the wheel hub, and fit the outer contour of the wheel hub.

[0025] S16, repeat the iteration to fit the tire outer contour and wheel hub outer contour that meet the preset error standard;

[0026] S17. Find the pixel gradient amplitude point at the junction of the tire and the ground within the range of the angle between the lower left and lower right 45 degrees from the center of the tire. Draw the dividing line between the tire and the ground based on the found pixel gradient amplitude point to obtain the interface between the tire and the ground after the tire is deformed.

[0027] S18. Perform endpoint processing on the image pixels at the tire-ground interface from bottom to top, calculate the difference in Y coordinates between two adjacent pixels, and determine that the pixel is the endpoint of the tire-ground contact when the difference in Y coordinates between any pixel and its adjacent pixels is greater than a preset coordinate difference threshold, thus obtaining the actual contact pixel length between the tire and the ground.

[0028] Furthermore, the deformation parameter calculation unit uses a scaling factor α to correct the obtained mechanical deformation parameters of the tire:

[0029]

[0030] In the formula, Rim is the hub radius; r is the fitted hub pixel radius.

[0031] Furthermore, the load prediction model calculates the first load F for each tire based on the mechanical deformation parameters of the tire sent by the deformation parameter calculation unit and the tire size and tire pressure information sent by the activation module. a Then, the regression coefficient β is used to evaluate the first load F predicted by the machine learning model. a Perform temperature correction:

[0032]

[0033] In the formula, F represents the temperature-corrected true tire load. a The actual tire load predicted by the machine learning model. This represents the average temperature of the tire surface. n represents the number of pixels capturing the temperature of the tire surface using thermal imaging, T jLet represent the temperature of the j-th temperature data point on the tire surface; the regression coefficient β ranges from 0.9 to 1.5. When the tire surface temperature reaches 90% of the cold tire pressure temperature corresponding to the standard tire pressure, it is taken as 0.9, and linear interpolation is used in the middle. When the tire surface temperature reaches 150% of the cold tire pressure temperature corresponding to the standard tire pressure, it is taken as 1.5, and linear interpolation is used in the middle.

[0034] Furthermore, the second load prediction module combines the tire pressure P of the i-th tire. i and the contact area S between the i-th tire and the ground i The second load F of the i-th tire is calculated. bi :

[0035] F' bi =P i ×S i ;

[0036] In the formula, the tire pressure P is the tire pressure. i Use 1.1 to 1.2 times the maximum tire pressure indicated by the tire markings.

[0037] Furthermore, the correction module includes a temperature correction unit, a hardness coefficient correction unit, and a final load calculation unit;

[0038] The temperature correction unit corrects the first load using the following formula:

[0039] F' ai =F ai ×k1;

[0040] In the formula, k1 is the temperature correction coefficient for the first load, F ai For the first load of the i-th tire, F' ai The first load after correction;

[0041] The temperature correction unit corrects the second load using the following formula:

[0042]

[0043] In the formula, k2 is the temperature correction coefficient for the second load, and F bi For the first load of the i-th tire, F' bi The first load is after temperature correction, and α is the hardness correction factor. The corrected second load;

[0044] The final load calculation unit combines the corrected first load and second load to calculate the tire load F of the i-th tire. i And the total load G of the vehicle:

[0045]

[0046]

[0047] In the formula, N is the number of axles of the vehicle.

[0048] This invention also mentions a dynamic weighing method for vehicles at toll stations based on multi-field thermal imaging technology, wherein the dynamic weighing method for vehicles at toll stations is executed based on the weighing system described above.

[0049] The dynamic vehicle weighing method at toll stations includes the following steps:

[0050] S1, capture real-time images of the toll station lanes, extract tire images from them, process the extracted tire images to obtain the tire size and tire pressure information, and simultaneously activate the first load prediction module and the second load prediction module.

[0051] S2, infrared thermal imaging devices located on both sides of the lane record the temperature information of the vehicle's single-axle tires. Based on the temperature matrix data of the single-axle tires, the mechanical deformation parameters of the single-axle tires are calculated. A load prediction model is trained based on the XGBoost model. The calculated tire mechanical deformation parameters, tire size, and tire pressure information are all imported into the load prediction model to calculate the first load F for each tire. a ;

[0052] S3 employs an infrared thermal imaging device and a temperature-sensitive material weighing device located above the lane to record the number of contact areas between the tires and the ground, as well as the contact area of ​​each individual tire. Based on the number of contact areas, the number of vehicle axles is calculated. Combining tire pressure information and the contact area of ​​each individual tire, the second load F for each tire is calculated. b ;

[0053] S4, regarding the first load F in step S2 a Perform temperature correction on the second load F in step S3. b Perform temperature correction and hardness coefficient correction;

[0054] S5. The final measurement result of the tire load is calculated based on the corrected first load and second load, and the total load of the vehicle is calculated in combination with the number of vehicle axles.

[0055] Furthermore, the process of calculating the mechanical deformation parameters of a single-axle tire based on the temperature matrix data of the single-axle tire includes the following steps:

[0056] S11, Generate a temperature image that is linearly related to temperature based on pixel temperature matrix data; Use the Sobel edge detection operator to calculate the pixel gradient magnitude of the temperature image, and perform image segmentation around pixels in the temperature image where the temperature difference is greater than a preset temperature difference threshold.

[0057] S12, retain the first 5% of the gradient magnitude of the segmented image, and take out the point with the largest gradient magnitude and mark it with color; select the color-marked pixels sequentially from the bottom to the top of the image, and use the selected pixels as seed points for the first fitting of the outer contour of the tire. All the selected seed points are pixels on the interface between the tire and the air.

[0058] S13, using the seed points selected in step S12, the region growing algorithm is used to grow the seed points to find the points adjacent to the seed points as new seed points, and the second fitting of the tire outer contour is performed.

[0059] S14. Retain the seed points selected in steps S12 and S13. Using the tire outer contour of the second fitting as a reference, find the pixel point with the maximum gradient magnitude in the upper half of the tire from top to bottom, and perform the third tire outer contour fitting.

[0060] S15. Using the tire outer contour fitted in step S14 as a reference, find the pixel point with the largest gradient magnitude towards the center of the tire. Use this point as the seed point for fitting the outer contour of the wheel hub, and fit the outer contour of the wheel hub.

[0061] S16, repeat the iteration to fit the tire outer contour and wheel hub outer contour that meet the preset error standard;

[0062] S17. Find the pixel gradient amplitude point at the junction of the tire and the ground within the range of the angle between the lower left and lower right 45 degrees from the center of the tire. Draw the dividing line between the tire and the ground based on the found pixel gradient amplitude point to obtain the interface between the tire and the ground after the tire is deformed.

[0063] S18. Perform endpoint processing on the image pixels at the tire-ground interface from bottom to top, calculate the difference in Y coordinates between two adjacent pixels, and determine that the pixel is the endpoint of the tire-ground contact when the difference in Y coordinates between any pixel and its adjacent pixels is greater than a preset coordinate difference threshold, thus obtaining the actual contact pixel length between the tire and the ground.

[0064] Furthermore, the obtained mechanical deformation parameters of the tire are corrected using a scaling factor α:

[0065]

[0066] In the formula, Rim is the hub radius; r is the fitted hub pixel radius.

[0067] Beneficial effects:

[0068] The toll station vehicle dynamic weighing system and method based on multi-field thermal imaging technology of the present invention employs an infrared thermal imaging camera in the measuring equipment, which improves measurement accuracy, expands the scope of application, and can be used even at night. In terms of computational theory, it adopts a machine learning XGBoost model, considering more features of the machine learning model input, reducing errors, and improving prediction accuracy. The tire load acquisition method, vehicle weighing method, and system of the present invention have a wide range of applications, are easy to integrate information, and compared with optical cameras, the infrared equipment in the present invention can operate stably for a long time in all weather conditions. Attached Figure Description

[0069] Figure 1 This is a flowchart of a tire load acquisition method based on a multi-field thermal imaging system in a preferred embodiment of the present invention;

[0070] Figure 2 This is a schematic diagram showing the arrangement of some system devices and equipment in a preferred embodiment of the present invention;

[0071] Figure 3 This is a schematic diagram of the effect of an infrared camera placed on the side capturing the outer contour of a tire in a preferred embodiment of the present invention;

[0072] Figure 4 This is a schematic diagram of a temperature-sensing material weighing device in a preferred embodiment of the present invention;

[0073] Figure 5 This is a schematic diagram of the contact area between a single tire and the ground recorded by the temperature-sensitive material weighing device in a preferred embodiment of the present invention;

[0074] Figure 6 This is a schematic diagram of the contact area between the tire and the ground, captured and recorded by an infrared camera positioned at the top in a preferred embodiment of the present invention.

[0075] The accompanying reference numerals are as follows: 1. Infrared thermal imaging device A; 2. Computer terminal loaded with the software required for the toll vehicle dynamic weighing system (excluding the triggering device, infrared thermal imaging device, optical camera, and temperature-sensitive material weighing device); 3. Infrared thermal imaging device C; 4. Temperature-sensitive material weighing device; 5. Contact area between a single tire and the ground. Detailed Implementation

[0076] The following embodiments are provided to enable those skilled in the art to more fully understand the present invention, but do not limit the invention in any way.

[0077] See Figure 2This embodiment discloses a dynamic vehicle weighing system for toll stations based on multi-field thermal imaging technology. The dynamic vehicle weighing system for toll stations includes an activation module, a first load prediction module, a second load prediction module, and a correction module.

[0078] The activation module is used to capture real-time images of the toll station lanes, extract tire images from them, process the extracted tire images to obtain the tire size and tire pressure information, and simultaneously activate the first load prediction module and the second load prediction module.

[0079] The first load prediction module includes infrared thermal imaging devices located on both sides of the lane, a deformation parameter calculation unit, and a load prediction model trained based on the XGBoost model. The infrared thermal imaging devices are used to capture temperature matrix data of the sidewalls of the vehicle's single-axle tires. The deformation parameter calculation unit calculates the mechanical deformation parameters of the tires based on the temperature matrix data of the single-axle tires. The load prediction model calculates the first load F of each tire based on the mechanical deformation parameters of the tires sent by the deformation parameter calculation unit and the tire size and tire pressure information sent by the activation module. a .

[0080] The second load prediction module includes an infrared thermal imaging device and a temperature-sensitive material weighing device located above the lane. It records the number of contact areas between the tires and the ground, the contact area of ​​a single tire with the ground, calculates the number of vehicle axles based on the number of contact areas, and calculates the second load F for each tire by combining tire pressure information and the contact area of ​​a single tire with the ground. b .

[0081] The correction module is used to adjust the first load F. a Second load F b Temperature and hardness coefficient corrections are applied, and the final measurement result of the tire load is calculated based on the corrected first and second loads.

[0082] Accordingly, see Figure 1 This embodiment also mentions a dynamic weighing method for vehicles at toll stations based on multi-field thermal imaging technology, wherein the dynamic weighing method for vehicles at toll stations is executed based on the weighing system described above.

[0083] The dynamic vehicle weighing method at toll stations includes the following steps:

[0084] S1: Capture real-time images of the toll station lanes, extract tire images from them, process the extracted tire images to obtain the tire size and tire pressure information, and simultaneously activate the first load prediction module and the second load prediction module.

[0085] The activation module includes a triggering device, an optical camera, a target detection model, an OCR text recognition model, and a tire information extraction unit.

[0086] The triggering device is located within the lane area and is used to detect whether a vehicle has appeared in the lane area; the optical camera is located on the side of the toll station lane and is used to capture real-time images of the toll station lane area; the target detection model is built based on the YOLO-v5 network and is used to detect the vehicle position in the real-time image captured by the optical camera, and to capture all tire images contained in the vehicle, and send the tire images to the OCR text recognition model, which identifies the tire model; the tire information extraction unit finds the tire size and tire pressure information according to the tire model sent by the OCR text recognition model, and at the same time activates the first load prediction module and the second load prediction module.

[0087] The activation module utilizes a deep learning model of the YOLO-v5 series of object detection algorithms to detect tires. When a tire is detected, the system is activated, and the multi-field thermal imaging system begins operation. The YOLO-v5 detection process is as follows: First, the image size is adjusted to 448*448. Then, the image is processed within the network. Finally, non-maximum suppression is performed to obtain the result. Unlike traditional detection algorithms that use sliding windows to find targets, YOLO-v5 directly uses a single convolutional neural network to predict multiple outer boxes and class probabilities. This unified design allows for end-to-end training and prediction. Furthermore, YOLO-v5 has strong generalization ability and can be widely applied to other test sets, resulting in a low background prediction error rate.

[0088] S2, infrared thermal imaging devices located on both sides of the lane (infrared thermal imaging device A on the left side of the lane and infrared thermal imaging device B on the right side of the lane) record the temperature information of the tires on each axle of the vehicle. Based on the temperature matrix data of the tires on each axle, the mechanical deformation parameters of the tires are calculated. A load prediction model is trained based on the XGBoost model. The calculated mechanical deformation parameters of the tires, the tire size, and the tire pressure information are all imported into the load prediction model to calculate the first load F of each tire. a .

[0089] The deformation parameter calculation unit processes the pixel temperature matrix data based on the OpenCV image processing algorithm, and detects the mechanical deformation parameters of the target sample tire through geometric fitting and region growing algorithms of image iteration. The mechanical deformation parameters include: the maximum pixel radius R of the tire, the maximum pixel area S1 of the tire, the pixel radius r of the wheel hub, the pixel length l of the tire contact with the ground, the pixel distance h from the center of the tire to the ground, the equivalent pixel area S2 of the tire after deformation, the difference in image pixel area ΔS before and after tire deformation, and the pixel length L of the tire-ground contact dividing line.

[0090] In this embodiment, the thermal imaging acquisition component can be a high-resolution, high-frame-rate thermal imaging device with a resolution of 640*480 or higher. In this example, the K26HE25 high-speed, high-frame-rate infrared thermal imaging acquisition component is used. The optical image acquisition component uses a Nikon D5600 SLR camera (hereinafter referred to as the camera), and the remaining components are laptop computers.

[0091] In this implementation case, as a vehicle slowly passes by, infrared thermal imaging devices A and B respectively capture thermal images of the left and right tires of the target vehicle, obtaining raw CSV temperature data files. The temperature information corresponding to each pixel in the CSV temperature data file is extracted, and a temperature image is redrawn according to a linear relationship. Then, the Sobel operator is used to perform edge detection on the temperature image, calculate the temperature gradient magnitude in the image, find the point with the largest gradient magnitude, and retain the top 5% of pixels with the largest gradient magnitude, marking them with color.

[0092] Based on prior knowledge, the point with the maximum gradient amplitude at the contact surface between the lower half of the tire and the air best represents the tire's edge contour. Edge detection points at the tire-air interface are selected sequentially from bottom to top of the image as seed points for fitting the tire's outer contour circle. The tire's outer contour is fitted using these seed points, and a region growing operation is performed on them to select a second set of seed points for further fitting. From the upper half of the tire, a third set of seed points is selected from top to bottom to further fit the tire's outer contour. Using the tire's outer contour as the boundary, temperature gradient amplitude points are sequentially found towards the center of the circle; these points are used as edge pixels of the wheel hub for fitting the wheel hub's outer contour. The pixel gradient magnitude point at the tire-ground interface is found within a 45-degree range to the lower left and lower right of the tire's center. Using this point as a reference, a tire-ground segmentation line is drawn to obtain the interface after tire deformation. Endpoint processing is performed on the image pixels at the tire-ground interface from bottom to top, calculating the Y-coordinate difference between two adjacent pixels. When the difference exceeds a certain threshold, the point is determined to be the endpoint of tire-ground contact, obtaining the actual contact pixel length. This completes the tire and rim outer contour edge segmentation based on thermal imaging data, yielding the following parameters: maximum tire pixel radius R, maximum tire pixel area S1, rim pixel radius r, tire-ground contact pixel length l, pixel distance h from the tire center to the ground, equivalent pixel area S2 after tire deformation, image pixel area difference ΔS before and after tire deformation, and the pixel length L of the tire-ground contact segmentation line. The average temperature of the tire's outer surface is calculated based on the thermal information captured by thermal imaging. like Figure 5 As shown.

[0093] Optical images of the tire are acquired using an optical acquisition component, and then an OCR recognition algorithm is invoked. The OCR recognition algorithm uses deep learning-based character recognition technology. It trains the PSENET character localization network and the CRNN character recognition network through transfer learning to form a two-stage character recognition method. It obtains tire size information (tire section height H, tire section width b, rim radius Rim) and air pressure information atm from the identifier information of the tire sidewall.

[0094] Based on the wheel hub dimensions, the various deformation parameters measured in the thermal imaging are corrected for size, and the mechanical deformation parameters of the tire are corrected using a scaling factor α. The corrected mechanical deformation parameters are:

[0095] r trye =α×r

[0096] R true =α×R

[0097] S 1true =α 2 ×S1

[0098] l true =α×l

[0099] h true =α×h

[0100] S 2true =α 2 ×S2

[0101] ΔS true =α 2 ×ΔS

[0102] L true =α×L

[0103] In the formula, r true R true S 1true l true h true S 2true ΔS true and L true The parameters are: the corrected wheel hub pixel radius, the maximum tire pixel radius, the maximum tire pixel area, the tire-to-ground contact pixel length, the pixel distance from the tire center to the ground, the equivalent pixel area after tire deformation, the difference in image pixel area before and after tire deformation, and the pixel length of the tire-to-ground contact dividing line, respectively. R, S1, l, h, S2, ΔS, and L are the fitted maximum tire pixel radius, maximum tire pixel area, tire-to-ground contact pixel length, tire-to-ground contact pixel distance, equivalent pixel area after tire deformation, difference in image pixel area before and after tire deformation, and the pixel length of the tire-to-ground contact dividing line, respectively. Specifically, when a vehicle slowly passes by, infrared thermal imaging devices A and B capture and record the outer contours of the tires on the left and right sides of the vehicle body, respectively. This involves detecting the rim area and the tire area to obtain the pixel points of the rim area and the tire area; calculating the image scaling factor (assuming no deformation in the rim area) based on the pixel points of the rim area and the rim diameter; and calculating the tire deformation based on the pixel points of the tire area and the image scaling factor. See the schematic diagram of the tire deformation effect captured by infrared cameras A and B placed on both sides of the lane. Figure 3 .

[0104] The XGboost model was trained using datasets collected from indoor tests to obtain a load prediction model. Twelve mechanical features measured by the thermal imaging and character recognition modules were input into the XGboost load prediction model to obtain the predicted load of a single tire. Outdoor validation showed that the method's error was less than 5%, demonstrating good prediction performance.

[0105] Based on the characteristic that tire pressure is affected by temperature changes, a temperature correction is applied to the tire load to obtain the corrected load for a single tire.

[0106] If the implementation scheme has the capability to measure the actual load of the tire, the tire load measured by a third-party device can be combined with the 12 mechanical features proposed in this embodiment to construct a dataset for updating the model (collecting a large number of images of tires of different sizes, tire pressures, loads, and temperatures, and using OCR technology to identify tire manufacturers and models to obtain tire section height, section width, and tire pressure, forming a certain amount of sample data), thereby realizing the automatic updating and self-learning mechanism of the model and improving the versatility of the load prediction model.

[0107] S3 uses an infrared thermal imaging device C located above the lane and a temperature-sensitive material weighing device to record the number of contact areas between the tires and the ground, and the contact area of ​​each tire with the ground. The number of vehicle axles is calculated based on the number of contact areas, and the second load F for each tire is calculated by combining tire pressure information and the contact area of ​​each tire with the ground. b .

[0108] After the vehicle passes the temperature-sensitive material weighing device, record the number of axles. The temperature-sensitive material weighing device is as follows: Figure 4 As shown. After the vehicle passes the temperature-sensitive material weighing device, the device records the contact area between the tire and the ground. The infrared thermal imaging device C, positioned above, immediately photographs and identifies this contact area, counting the number of contact areas. This number is recorded as M, and the number of vehicle axles is then calculated. The contact area S between a single tire on the left and right sides of the vehicle and the ground is recorded using an infrared thermal imaging device C positioned at the top. i The specific implementation method is as follows: After the vehicle passes the temperature-sensitive material weighing device, the device records the contact area between the tire and the ground. The infrared thermal imaging device C, positioned above, immediately photographs and identifies this contact area, and then counts and records the area of ​​each individual contact area. A schematic diagram of the contact area between a single tire and the ground is shown below. Figure 5 The infrared camera C, positioned at the top, captures and records the contact area between a single tire and the ground, as shown in the image. Figure 6As shown. The measured contact area between a single tire and the ground is proportionally corrected. The specific correction process is as follows: The data processing device calculates the number of pixels M in the infrared image generated by the upper infrared thermal imaging device C. The product of the calculated number of pixels M and the tire-to-ground true contact area ratio factor K (determined by the relative position and distance between the upper infrared thermal imaging device C and the temperature-sensing material weighing device) is taken as the true contact area S between the tire and the ground. 真 S 真 = M×K, where M is the number of pixels in the infrared image, and K is the ratio factor of the actual contact area between the tire and the ground, determined by the relative position and distance between the infrared thermal imaging device C placed above and the heat conduction weighing device. The obtained tire pressure P i The contact area S between a single tire and the ground. i Multiplying the corresponding values ​​yields the tire load, F. bi =P i ×S i .

[0109] S4, regarding the first load F in step S2 a Perform temperature correction on the second load F in step S3. b Temperature correction and hardness coefficient correction are performed.

[0110] When measuring tire temperature using an infrared thermal imaging device from the tire sidewall, a temperature correction is applied. The temperature correction factor is k1. Therefore, the tire load obtained in step 2 should be corrected as follows: F' ai =F ai ×k1.

[0111] When measuring the tire-ground contact area using an infrared thermal imaging device combined with a temperature-sensitive material weighing method, a temperature correction is applied. The temperature correction coefficient is taken as k2. Therefore, the tire load obtained in step 3 should be corrected as follows: F' bi =F bi ×k2; When measuring the tire-ground contact area using an infrared thermal imaging device combined with heat conduction, a hardness correction should also be applied. The hardness correction factor is taken as α.

[0112] S5. The final measurement result of the tire load is calculated based on the corrected first load and second load, and the total load of the vehicle is calculated in combination with the number of vehicle axles.

[0113] The final tire load can be obtained from the above steps. Total vehicle weight

[0114] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. The above description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application is not limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention. Furthermore, the parts of the technical solutions provided in the embodiments of this application that are consistent with the implementation principles of corresponding technical solutions in the prior art have not been described in detail to avoid excessive elaboration.

Claims

1. A dynamic vehicle weighing system for toll stations based on multi-field thermal imaging technology, characterized in that, The toll station vehicle dynamic weighing system includes an activation module, a first load prediction module, a second load prediction module, and a correction module. The activation module is used to capture real-time images of the toll station lanes, extract tire images from them, process the extracted tire images to obtain the tire size and tire pressure information, and simultaneously activate the first load prediction module and the second load prediction module. The first load prediction module includes infrared thermal imaging devices located on both sides of the lane, a deformation parameter calculation unit, and a load prediction model trained based on the XGBoost model. The infrared thermal imaging devices are used to capture temperature matrix data of the sidewalls of the vehicle's single-axle tires. The deformation parameter calculation unit calculates the mechanical deformation parameters of the tires based on the temperature matrix data of the single-axle tires. The load prediction model calculates the first load F of each tire based on the mechanical deformation parameters of the tires sent by the deformation parameter calculation unit and the tire size and tire pressure information sent by the activation module. a ; The second load prediction module includes an infrared thermal imaging device and a temperature-sensitive material weighing device located above the lane. It records the number of contact areas between the tires and the ground, the contact area of ​​a single tire with the ground, calculates the number of vehicle axles based on the number of contact areas, and calculates the second load F for each tire by combining tire pressure information and the contact area of ​​a single tire with the ground. b ; The correction module is used to adjust the first load F. a Second load F b Temperature and hardness coefficient corrections are applied, and the final measurement result of the tire load is calculated based on the corrected first and second loads. The deformation parameter calculation unit processes the pixel temperature matrix data based on the OpenCV image processing algorithm, and detects the mechanical deformation parameters of the target sample tire through geometric fitting and region growing algorithms based on image iteration. The mechanical deformation parameters include: the maximum pixel radius R of the tire, the maximum pixel area S1 of the tire, the pixel radius r of the wheel hub, the pixel length l of the tire in contact with the ground, the pixel distance h from the center of the tire to the ground, the equivalent pixel area S2 of the tire after deformation, the difference in image pixel area ΔS before and after tire deformation, and the pixel length L of the tire-ground contact dividing line. The process by which the deformation parameter calculation unit detects and obtains the mechanical deformation parameters of the target sample tire includes the following steps: S11, Generate a temperature image that is linearly related to temperature based on pixel temperature matrix data; Use the Sobel edge detection operator to calculate the pixel gradient magnitude of the temperature image, and perform image segmentation around pixels in the temperature image where the temperature difference is greater than a preset temperature difference threshold. S12, retain the first 5% of the gradient magnitude of the segmented image, and take out the point with the largest gradient magnitude and mark it with color; select the color-marked pixels sequentially from the bottom to the top of the image, and use the selected pixels as seed points for the first fitting of the outer contour of the tire. All the selected seed points are pixels on the interface between the tire and the air. S13, using the seed points selected in step S12, the region growing algorithm is used to grow the seed points to find the points adjacent to the seed points as new seed points, and the second fitting of the tire outer contour is performed. S14. Retain the seed points selected in steps S12 and S13. Using the tire outer contour of the second fitting as a reference, find the pixel point with the maximum gradient magnitude in the upper half of the tire from top to bottom, and perform the third tire outer contour fitting. S15. Using the tire outer contour fitted in step S14 as a reference, find the pixel point with the largest gradient magnitude towards the center of the tire. Use this point as the seed point for fitting the outer contour of the wheel hub and fit the outer contour of the wheel hub. S16, repeat the iteration to fit the tire outer contour and wheel hub outer contour that meet the preset error standard; S17. Find the pixel gradient amplitude point at the junction of the tire and the ground within the range of the angle between the lower left and lower right 45 degrees from the center of the tire. Draw the dividing line between the tire and the ground based on the found pixel gradient amplitude point to obtain the interface between the tire and the ground after the tire is deformed. S18. Perform endpoint processing on the image pixels at the tire-ground interface from bottom to top, calculate the difference in Y coordinates between two adjacent pixels, and determine that the pixel is the endpoint of the tire-ground contact when the difference in Y coordinates between any pixel and its adjacent pixels is greater than a preset coordinate difference threshold, thus obtaining the actual contact pixel length between the tire and the ground.

2. The toll station vehicle dynamic weighing system based on multi-field thermal imaging technology according to claim 1, characterized in that, The activation module includes a triggering device, an optical camera, a target detection model, an OCR text recognition model, and a tire information extraction unit. The triggering device is located within the lane area and is used to detect whether a vehicle has appeared in the lane area; The optical camera is located on the side of the toll station lane and is used to capture real-time images of the toll station lane area; The target detection model is built on the YOLO-v5 network and is used to detect the vehicle position in the real-time image captured by the optical camera, and to capture all tire images contained in the vehicle. The tire images are then sent to the OCR text recognition model, which identifies the tire model. The tire information extraction unit finds the tire size and tire pressure information based on the tire model sent by the OCR text recognition model, and simultaneously activates the first load prediction module and the second load prediction module.

3. The toll station vehicle dynamic weighing system based on multi-field thermal imaging technology according to claim 1, characterized in that, The deformation parameter calculation unit uses a scaling factor α to correct the obtained mechanical deformation parameters of the tire: In the formula, Rim is the hub radius; r is the fitted hub pixel radius.

4. The toll station vehicle dynamic weighing system based on multi-field thermal imaging technology according to claim 1, characterized in that, The load prediction model calculates the first load F for each tire based on the mechanical deformation parameters of the tire sent by the deformation parameter calculation unit and the tire size and tire pressure information sent by the activation module. a Then, the regression coefficient β is used to evaluate the first load F predicted by the machine learning model. a Perform temperature correction: In the formula, F represents the temperature-corrected true tire load. a The actual tire load predicted by the machine learning model. This represents the average temperature of the tire surface. n represents the number of pixels capturing the temperature of the tire surface using thermal imaging, T j Let represent the temperature of the j-th temperature data point on the tire surface; the regression coefficient β ranges from 0.9 to 1.

5. When the tire surface temperature reaches 90% of the cold tire pressure temperature corresponding to the standard tire pressure, it is taken as 0.9, and linear interpolation is used in the middle. When the tire surface temperature reaches 150% of the cold tire pressure temperature corresponding to the standard tire pressure, it is taken as 1.5, and linear interpolation is used in the middle.

5. The toll station vehicle dynamic weighing system based on multi-field thermal imaging technology according to claim 1, characterized in that, The second load prediction module combines the tire pressure P of the i-th tire. i and the contact area S between the i-th tire and the ground i The second load F of the i-th tire is calculated. bi : F bi =P i ×S i ; In the formula, the tire pressure P is the tire pressure. i Use 1.1 to 1.2 times the maximum tire pressure indicated by the tire markings.

6. The toll station vehicle dynamic weighing system based on multi-field thermal imaging technology according to claim 1, characterized in that, The correction module includes a temperature correction unit, a hardness coefficient correction unit, and a final load calculation unit; The temperature correction unit corrects the first load using the following formula: F ai =F ai ×k1; In the formula, k1 is the temperature correction coefficient for the first load, F ai For the first load of the i-th tire, F' ai The first load after correction; The temperature correction unit corrects the second load using the following formula: In the formula, k2 is the temperature correction coefficient for the second load, and F bi For the first load of the i-th tire, F' bi The first load is after temperature correction, and α′ is the hardness correction factor. The corrected second load; The final load calculation unit combines the corrected first load and second load to calculate the tire load F of the i-th tire. i And the total load G of the vehicle: In the formula, N is the number of axles of the vehicle.

7. A method for dynamic vehicle weighing at toll stations based on multi-field thermal imaging technology, characterized in that, The toll station vehicle dynamic weighing method is executed based on the weighing system described in any one of claims 1-6; The dynamic vehicle weighing method at toll stations includes the following steps: S1, capture real-time images of the toll station lanes, extract tire images from them, process the extracted tire images to obtain the tire size and tire pressure information, and simultaneously activate the first load prediction module and the second load prediction module. S2, infrared thermal imaging devices located on both sides of the lane record the temperature information of the vehicle's single-axle tires. Based on the temperature matrix data of the single-axle tires, the mechanical deformation parameters of the single-axle tires are calculated. A load prediction model is trained based on the XGBoost model. The calculated tire mechanical deformation parameters, tire size, and tire pressure information are all imported into the load prediction model to calculate the first load F for each tire. a ; S3 employs an infrared thermal imaging device and a temperature-sensitive material weighing device located above the lane to record the number of contact areas between the tires and the ground, as well as the contact area of ​​each individual tire. Based on the number of contact areas, the number of vehicle axles is calculated. Combined with tire pressure information and the contact area of ​​each individual tire, the second load F for each tire is calculated. b ; S4, regarding the first load F in step S2 a Perform temperature correction on the second load F in step S3. b Perform temperature correction and hardness coefficient correction; S5. The final measurement result of the tire load is calculated based on the corrected first load and second load, and the total load of the vehicle is calculated in combination with the number of vehicle axles.

8. The method for dynamic vehicle weighing at toll stations based on multi-field thermal imaging technology according to claim 7, characterized in that, The process of calculating the mechanical deformation parameters of a single-axle tire based on the temperature matrix data of the tires on a single axle of a vehicle includes the following steps: S11, Generate a temperature image that is linearly related to temperature based on pixel temperature matrix data; Use the Sobel edge detection operator to calculate the pixel gradient magnitude of the temperature image, and perform image segmentation around pixels in the temperature image where the temperature difference is greater than a preset temperature difference threshold. S12, retain the first 5% of the gradient magnitude of the segmented image, and take out the point with the largest gradient magnitude and mark it with color; select the color-marked pixels sequentially from the bottom to the top of the image, and use the selected pixels as seed points for the first fitting of the outer contour of the tire. All the selected seed points are pixels on the interface between the tire and the air. S13, using the seed points selected in step S12, the region growing algorithm is used to grow the seed points to find the points adjacent to the seed points as new seed points, and the second fitting of the tire outer contour is performed. S14. Retain the seed points selected in steps S12 and S13. Using the tire outer contour of the second fitting as a reference, find the pixel point with the maximum gradient magnitude in the upper half of the tire from top to bottom, and perform the third tire outer contour fitting. S15. Using the tire outer contour fitted in step S14 as a reference, find the pixel point with the largest gradient magnitude towards the center of the tire. Use this point as the seed point for fitting the outer contour of the wheel hub and fit the outer contour of the wheel hub. S16, repeat the iteration to fit the tire outer contour and wheel hub outer contour that meet the preset error standard; S17. Find the pixel gradient amplitude point at the junction of the tire and the ground within the range of the angle between the lower left and lower right 45 degrees from the center of the tire. Draw the dividing line between the tire and the ground based on the found pixel gradient amplitude point to obtain the interface between the tire and the ground after the tire is deformed. S18. Perform endpoint processing on the image pixels at the tire-ground interface from bottom to top, calculate the difference in Y coordinates between two adjacent pixels, and determine that the pixel is the endpoint of the tire-ground contact when the difference in Y coordinates between any pixel and its adjacent pixels is greater than a preset coordinate difference threshold, thus obtaining the actual contact pixel length between the tire and the ground.

9. The method for dynamic vehicle weighing at toll stations based on multi-field thermal imaging technology according to claim 8, characterized in that, The obtained mechanical deformation parameters of the tire are corrected using a scaling factor α: In the formula, Rim is the hub radius; r is the fitted hub pixel radius.

Images