A road and bridge overload early warning system and method based on computer vision technology

Abstract

The application discloses a road and bridge overload early warning system and method based on computer vision technology. First, a data acquisition system is arranged, that is, a camera is arranged on the top and side of a target lane to collect images. Second, the overall information of a vehicle is identified according to the image of the top camera, such as the type of the vehicle and the axle; and the information of the tire image collected by the side camera is identified, such as the tire model, the contact area between the tire and the ground and the like. Then, the probability distribution characteristics of the vehicle load are calculated based on the uncertainty of the tire pressure corresponding to the tire model, the number of vehicle axles and the contact area. Finally, the risk early warning of the overloaded vehicle is carried out by using the reliability analysis method according to the condition that the weight of the vehicle beam is less than the weight limit of the road and bridge. The application focuses on the easy-overloaded vehicle driving on the side lane of the bridge, and based on the research result of the tire pressure distribution characteristics of the vehicle, the relationship among the contact area, the pressure and the stress size is used, and simple physical knowledge is used to identify the dynamic vehicle load, so that a road and bridge overload early warning system and method which is stable, simple and reliable and can be adjusted according to the needs is provided for the bridge vehicle identification.

Description

Technical Field

[0001] This invention relates to the field of vehicle load recognition technology, specifically a road and bridge overload early warning system and method based on computer vision technology. Background Technology

[0002] Road transport is a crucial component of my country's comprehensive transportation system. With my country's economic development and the continuous improvement of people's living standards, road transport construction has progressed rapidly. However, due to disorderly competition in the transport market in recent years, some transport companies and individuals, driven by profit, have overloaded and oversized cargo, damaging road and bridge facilities, disrupting the road transport market order, and causing numerous traffic accidents. To prevent road fatigue and serious damage to transportation infrastructure caused by vehicle overloading, it is essential to closely monitor the load conditions of heavy vehicles. Strengthening the management of overloaded transport on bridges and highways, and developing early warning technologies for detecting overloaded vehicles on roads and bridges, are crucial for the safety of highways and bridges.

[0003] Existing vehicle weight measurement methods include static weighing and road surface dynamic response-based weighing. Static weighing, also known as traditional weighing, involves fixed weighbridges typically located at roadside vehicle overload checkpoints or highway toll booths. Due to inherent limitations of static monitoring technology, its detection efficiency is low, it requires manual intervention, and it cannot achieve 24 / 7 non-stop monitoring and dynamic measurement of all vehicles. It primarily uses electronic weighbridges installed on the ground to weigh passing vehicles; the process is complex, resulting in slow detection speeds, inconvenient maintenance, and potential traffic congestion. Existing dynamic weighing systems indirectly acquire dynamic vehicle data by obtaining bridge responses through high-precision sensors. This method requires fixing sensors under the bridge, placing high technical demands on bridge design and construction, and incurring significant construction costs. Summary of the Invention

[0004] To address the shortcomings of the aforementioned methods and technologies, this invention provides a road and bridge overload early warning system and method based on computer vision technology, aiming to achieve automatic identification and early warning of dynamic vehicle loads.

[0005] This invention is achieved through the following technical solution: This invention proposes a road and bridge overload early warning system and method based on computer vision technology, specifically including the following steps:

[0006] Step 1: Deploy the data acquisition system: Select the lane to be detected, and use the existing cameras on the top of the lane to collect overall images of the vehicles in the lane; deploy cameras on the side of the lane to collect overall images of the side tires of the vehicles.

[0007] Step 2: Overall vehicle feature recognition: Using images captured by the vehicle's roof camera, the vehicle information is automatically determined through a deep learning object detection algorithm: license plate, vehicle model, and number of axles.

[0008] Step 3: Local Tire Feature Recognition: Using images captured by lane-side cameras, identify the tire model and tire-to-ground contact length. From the tire model, obtain the rim diameter and tire width, and thus the tire-to-ground contact area. Actual vehicle tire pressure follows a normal distribution P:N(μ p ,σ p 2 ), where μ p It is the mean tire pressure, σ p This represents the standard deviation of tire pressure.

[0009] Step 4: Identify vehicle weight: Calculate the probability density function of the target vehicle load by taking into account the contact area between each tire and the ground, tire pressure characteristics, and the number of axles.

[0010] Step 5: Risk Warning for Overweight Vehicles: Identify vehicles whose weight requirement is less than the weight limit of the road or bridge. Calculate the probability P of the target vehicle being overweight. f Based on risk appetite, a warning percentage is set to implement a four-level warning system.

[0011] Furthermore, step one specifically includes:

[0012] Step 1.1: Adjust the field of view of the high-definition high-speed camera installed on the side of the bridge lane to match the field of view of the existing capture camera at the top of the lane, so as to ensure that the two cameras capture the same vehicle.

[0013] Step 1.2: Use the lane-top camera to capture an image of the overall outline of the target vehicle; use the lane-side camera to capture images of passing vehicles as tire images of the target vehicle.

[0014] Step 1.3: Filter the captured images and upload the clear and identifiable images to the computer.

[0015] Furthermore, step two specifically involves: using deep learning to extract features from the captured target vehicle outline image, and further performing vehicle type recognition, axle counting, and license plate recognition on the target vehicle.

[0016] Furthermore, step three specifically includes:

[0017] Step 3.1: Identify the special markings of different tire models through big data collection to determine the tire model, thereby obtaining the tire width and rim diameter;

[0018] Step 3.2: Perform grayscale processing on the target vehicle tire image, perform edge detection on the tire edge, extract an approximate straight line segment as the contact length between the tire and the ground, and define the relative length of this straight line segment on the captured image as the pixel length of the tire contact with the ground, which will be used for the next step of conversion.

[0019] Step 3.3: Use the identified wheel hub diameter to calibrate the image, and then use an algorithm based on the similarity principle to convert the pixel length of the tire contact with the ground into the actual contact length between the tire and the ground;

[0020] Step 3.4: The contact length between the vehicle tire and the ground is analyzed using a computer vision recognition system. The corresponding tire width is automatically entered based on the tire model. Then, the contact area between the vehicle tire and the ground is calculated based on the approximate rectangular shape of the tire-ground contact area. The formula for expressing the tire-ground contact area is:

[0021] A i =a i ·b

[0022] Among them, A i Let a be the contact area between the i-th tire and the road surface. i Let be the contact length between the i-th tire and the road surface, and b be the contact width between the tire and the ground automatically matched by the system.

[0023] Step 3.5: Based on the target vehicle model, match the tire pressure P:N (μ) for that tire from the database. p ,σ p 2 ) and the weight limit for roads and bridges for this vehicle type [W] r ].

[0024] Furthermore, step four specifically involves:

[0025] Considering vehicle tire pressure fluctuations, the tire pressure N (μ) p ,σ p 2 By combining the number of vehicle axles and the contact area between a single tire of the target vehicle and the ground, the probability density function of the vehicle load is derived. The relationship between the probability density function of the vehicle load and the contact area and tire pressure is as follows:

[0026]

[0027] Where f(W) is the probability density function of the vehicle load, and W is the vehicle load. N represents the number of tires on the right side of the vehicle, i.e., the number of axles. A i Let be the contact area between the i-th tire and the road surface.

[0028] Furthermore, step five specifically includes:

[0029] Step 5.1, when W > [W r When, [W] r[This refers to] weight limits for roads and bridges. If a vehicle exceeds its weight limit, the probability P of the target vehicle exceeding its weight limit can be obtained from the probability density function of the vehicle load. f Among them, the probability P of the vehicle being overweight f The expression is:

[0030]

[0031] Among them, P f The probability of the target vehicle being overweight;

[0032] Step 5.2: The warning signal level is divided according to the probability range of the target vehicle. It can be set by the relevant management unit based on factors such as the highway bridge grade and the current condition of the bridge. For example, a newly built secondary highway bridge can generally be set with four warning signals: Level IV (General, The low possible warning, the probability of the target vehicle being overweight is [0, 25%]), Level III (Slightly heavy, The high possible warning, the probability of the target vehicle being overweight is (25%, 50%)), Level II (Severe, The higher possible warning, the probability of the target vehicle being overweight is (25%, 50%)), and Level I (Extremely severe, The highest possible warning, the probability of the target vehicle being overweight is (75%, 100%)), represented by blue, yellow, orange, and red respectively, and labeled in both Chinese and English.

[0033] The beneficial effects of this invention are as follows:

[0034] 1. This invention can be widely applied to various bridge monitoring scenarios, and there are no special requirements for the installation position and angle of the monitoring camera.

[0035] 2. The road and bridge overload early warning system and method based on computer vision technology of the present invention provides a novel vehicle identification method. Compared with the traditional static weighing method and the existing dynamic weighing method, the information identification end adopts a simple image acquisition device and is not affected by the bridge or its own conditions. Therefore, it is more economical and easier to install.

[0036] 3. The road and bridge overload early warning system and method based on computer vision technology can identify vehicle load and overload status at the same time, realizing the effective combination of identification and early warning systems, and laying the foundation for bridge structural damage identification based on vehicle overload. Attached Figure Description

[0037] Figure 1 This is a flowchart of the road and bridge overload early warning system and method based on computer vision technology, which is part of this invention patent.

[0038] Figure 2 This is a schematic diagram of the deployment of the data acquisition system in an embodiment of the present invention;

[0039] Figure 3 These are tire images acquired in an embodiment of the present invention, wherein (a) is a front axle tire image and (b) is a rear axle tire image;

[0040] Figure 4 This is a flowchart illustrating the process of identifying the contact length between the tire and the ground in an embodiment of the present invention.

[0041] Figure 5 This is a tire pressure normal distribution diagram in an embodiment of the present invention;

[0042] Figure 6 This is a probability density function graph of the measured vehicle load in an embodiment of the present invention;

[0043] Figure 7 This is a schematic diagram of a graded early warning system in an embodiment of the present invention. Detailed Implementation

[0044] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described examples are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making any creative work are within the protection scope of the present invention.

[0045] In the broad field of bridge health monitoring, the identification of heavy truck loads and overload warnings are crucial for assessing and managing the load-bearing status of bridge structures. Computer vision-based methods and systems for identifying moving vehicle loads on bridges, focusing on heavy truck loads on sidewalks, represent a promising strategy.

[0046] The purpose of this invention is to meet the needs of the field of bridge health monitoring by establishing a computer vision-based method for identifying bridge vehicle moving loads. Combined with... Figure 1 This invention proposes a computer vision-based method for identifying moving loads on bridge vehicles, specifically including the following steps:

[0047] Step 1: Deploy the data acquisition system: Select the lane to be detected, and use the existing cameras on the top of the lane to collect overall images of the vehicles in the lane; deploy cameras on the side of the lane to collect overall images of the side tires of the vehicles.

[0048] Step 2: Overall vehicle feature recognition: Using images captured by the vehicle's roof camera, the vehicle information is automatically determined through a deep learning object detection algorithm: license plate, vehicle model, and number of axles.

[0049] Step 3: Local Tire Feature Recognition: Using images captured by lane-side cameras, identify the tire model and tire-to-ground contact length. From the tire model, obtain the rim diameter and tire width, and thus the tire-to-ground contact area. Actual vehicle tire pressure follows a normal distribution P:N(μ p ,σ p 2 ), where μ p It is the mean tire pressure, σ p This represents the standard deviation of tire pressure.

[0050] Step 4: Identify vehicle weight: Calculate the probability density function of the target vehicle load by taking into account the contact area between each tire and the ground, tire pressure characteristics, and the number of axles.

[0051] Step 5: Risk Warning for Overweight Vehicles: Identify vehicles whose weight requirement is less than the weight limit of the road or bridge. Calculate the probability P of the target vehicle being overweight. f Based on risk appetite, a warning percentage is set to implement a four-level warning system.

[0052] Furthermore, step one specifically includes:

[0053] Step 1.1: Adjust the field of view of the high-definition high-speed camera installed on the side of the bridge lane to match the field of view of the existing capture camera at the top of the lane, so as to ensure that the two cameras capture the same vehicle.

[0054] Step 1.2: Use the lane-top camera to capture an image of the overall outline of the target vehicle; use the lane-side camera to capture images of passing vehicles as tire images of the target vehicle.

[0055] Step 1.3: Filter the captured images and upload the clear and identifiable images to the computer.

[0056] Furthermore, step two specifically involves: using deep learning to extract features from the captured target vehicle outline image, and further performing vehicle type recognition, axle counting, and license plate recognition on the target vehicle.

[0057] Furthermore, step three specifically includes:

[0058] Step 3.1: Identify the special markings of different tire models through big data collection to determine the tire model, thereby obtaining the tire width and rim diameter;

[0059] Step 3.2: Perform grayscale processing on the target vehicle tire image, perform edge detection on the tire edge, extract an approximate straight line segment as the contact length between the tire and the ground, and define the relative length of this straight line segment on the captured image as the pixel length of the tire contact with the ground, which will be used for the next step of conversion.

[0060] Step 3.3: Use the identified wheel hub diameter to calibrate the image, and then use an algorithm based on the similarity principle to convert the pixel length of the tire contact with the ground into the actual contact length between the tire and the ground;

[0061] Step 3.4: The contact length between the vehicle tire and the ground is analyzed using a computer vision recognition system. The corresponding tire width is automatically entered based on the tire model. Then, the contact area between the vehicle tire and the ground is calculated based on the approximate rectangular shape of the tire-ground contact area. The formula for expressing the tire-ground contact area is:

[0062] A i =a i ·b

[0063] Among them, A i Let a be the contact area between the i-th tire and the road surface. i Let be the contact length between the i-th tire and the road surface, and b be the contact width between the tire and the ground automatically matched by the system.

[0064] Step 3.5: Based on the target vehicle model, match the tire pressure P:N (μ) for that tire from the database. p ,σ p 2 ) and the weight limit for roads and bridges for this vehicle type [W] r ].

[0065] Furthermore, step four specifically involves:

[0066] Considering vehicle tire pressure fluctuations, the tire pressure N (μ) p ,σ p 2 By combining the number of vehicle axles and the contact area between a single tire of the target vehicle and the ground, the probability density function of the vehicle load is derived. The relationship between the probability density function of the vehicle load and the contact area and tire pressure is as follows:

[0067]

[0068] Where f(W) is the probability density function of the vehicle load, and W is the vehicle load. N represents the number of tires on the right side of the vehicle, i.e., the number of axles. A i Let be the contact area between the i-th tire and the road surface.

[0069] Furthermore, step five specifically includes: /

[0070] Step 5.1, when W > [W r When, [Wr ] For weight limits on roads and bridges, if a vehicle is overweight, the probability P of the target vehicle being overweight can be obtained from the probability density function of the vehicle load.f Among them, the probability P of the vehicle being overweight f The expression is:

[0071]

[0072] Among them, P f The probability of the target vehicle being overweight;

[0073] Step 5.2: The warning signal level is divided according to the probability range of the target vehicle. It can be set by the relevant management unit based on factors such as the highway bridge grade and the current condition of the bridge. For example, a newly built secondary highway bridge can generally be set with four warning signals: Level IV (General, The low possible warning, the probability of the target vehicle being overweight is [0, 25%]), Level III (Slightly heavy, The high possible warning, the probability of the target vehicle being overweight is (25%, 50%)), Level II (Severe, The higher possible warning, the probability of the target vehicle being overweight is (25%, 50%)), and Level I (Extremely severe, The highest possible warning, the probability of the target vehicle being overweight is (75%, 100%)), represented by blue, yellow, orange, and red respectively, and labeled in both Chinese and English.

[0074] This invention utilizes computer vision technology and deep learning to quickly and easily acquire vehicle tire information, integrating and calculating the tire-ground contact area and tire pressure distribution. It exhibits good adaptability to bridges and does not impact traffic flow. Therefore, the load identification method based on computer vision technology can conveniently, quickly, and effectively identify vehicle load distribution. Based on vehicle load distribution and combined with road and bridge weight limits, it enables graded early warning of overload on bridges.

[0075] Example:

[0076] This implementation method applies the invention to the entrance and exit sections of a bridge. Since large vehicles are a major consideration in assessing the load-bearing state of bridge structures, this example uses heavy-duty trucks as a case study.

[0077] Step one specifically involves: setting up a data acquisition system, such as... Figure 2 As shown. The side lane frequently used by heavy trucks was selected as the test lane. The high-definition high-speed camera installed beside the lane was adjusted to fit the field of view of the capture camera on the gantry. The side tire images and the overall image of the vehicle were taken and uploaded to the computer.

[0078] Step two specifically involves: inspecting the camera installed on the gantry, i.e. Figure 2The No. 1 camera in the middle collects an overall image of the vehicle, which is preprocessed to extract the vehicle edge map. The vehicle features are then identified as a whole using a deep learning object detection algorithm, and the vehicle information is automatically determined: the vehicle type is a heavy truck with 2 axles.

[0079] Step three specifically involves: using lane-side cameras, i.e. Figure 2 The tire images of the vehicle under test, captured by camera number 1, include the front axle tire images as shown below. Figure 3 As shown in (a), the rear axle tire image is as follows: Figure 3 As shown in (b), vehicle data is extracted using tire feature points. The tires on both the front and rear axles of the vehicle under test are 10.00R 20. The tire width and rim diameter are obtained and converted to millimeters. For the rim diameter, the recognition result is 10 inches, which is converted to millimeters and output as 1 in = 25.4 mm, resulting in a rim diameter of 508 mm. For the tire width, the recognition result is either inches or millimeters, requiring a judgment. If the value is less than 50, the recognition result is in inches, converted to millimeters, and output as 1 in = 25.4 mm; if the value is greater than 50, the recognition result is in millimeters, and the result is directly output. In this example, the tire width recognition result for the vehicle under test is 10.00, which is less than 50, resulting in a tire width output of 254 mm.

[0080] Combination Figure 4 The target vehicle tire image is processed in grayscale. After edge detection, an approximate straight line segment is extracted as the contact length between the tire and the ground. The relative length of this straight line segment in the captured image is defined as the pixel length of the tire-ground contact. The identified wheel hub diameter is used to calibrate the image. Then, an algorithm based on the similarity principle is used to convert the pixel length of the tire-ground contact into the actual contact length between the tire and the ground. In this example, the actual contact length between the front axle tire and the ground is 280mm, and the actual contact length between the front axle tire and the ground is 230mm.

[0081] The computer vision recognition system analyzes the contact length between the vehicle tire and the ground, automatically fills in the corresponding tire width based on the tire model, and then calculates the contact area between the tire and the ground, in square millimeters, based on the approximate rectangular shape of the tire-ground contact. The formula for expressing the tire-ground contact area is:

[0082] A i =a i ·b

[0083] Among them, A i Let a be the contact area between the i-th tire and the road surface. i Let be the contact length between the i-th tire and the road surface, and b be the contact width between the tire and the ground automatically matched by the system.

[0084] Calculations show that in this example, the actual contact area between the front axle tire and the ground is 7.112 × 10⁻⁶. 4 mm 2 The actual contact area between the front axle tire and the ground is 5.842 × 10⁻⁶. 4 mm 2 ;

[0085] Based on the target vehicle model, match the tire pressure P:N (μ) for that tire from the database. p ,σ p 2 ) and the weight limit for roads and bridges for this vehicle type [W] r The normal distribution curve of tire pressure is shown in the figure. Figure 5 As shown, for the heavy-duty truck under study, μ is taken as... p =0.7MPa,σ p =0.05MPa, [W r ]=18000kg×9.8N / kg=176400kN.

[0086] Step four specifically involves: considering the fluctuation of vehicle tire pressure, adjusting the tire pressure N (μ) p ,σ p 2 By combining the number of vehicle axles and the contact area between the target vehicle's individual tires and the ground, the probability density function of the vehicle load is derived, as shown in the graph. Figure 6 As shown. The probability density function of vehicle load is related to the contact area and tire pressure as follows:

[0087]

[0088] Where f(W) is the probability density function of the vehicle load, and W is the vehicle load. N represents the number of tires on the right side of the vehicle, i.e., the number of axles. A i Let μ be the contact area between the i-th tire and the road surface. Calculations show that μ... W =181356kN,σ W =12954kN.

[0089] Step five specifically involves: calculating the overweight probability P of the target vehicle based on the probability density function of the vehicle load. f ,Right now Figure 6 The area of ​​the shaded region in the diagram, where the probability of the vehicle being overweight, P, is... f The expression is:

[0090]

[0091] Among them, P f The probability of the target vehicle being overweight;

[0092] According to the computer calculation results, the probability of the vehicle being overweight in this embodiment is 64.90%.

[0093] The warning percentage is set according to risk preference, and a four-level warning system is implemented. The warning signal level is divided according to the probability interval of the target vehicle and can be set by the relevant management unit based on factors such as the highway bridge grade and the current condition of the bridge. In this example, the warning signal is set to four levels: Level IV (General, The low possible warning, the probability of the target vehicle being overweight is [0, 25%]), Level III (Slightly Heavy, The high possible warning, the probability of the target vehicle being overweight is (25%, 50%)), Level II (Severe, The higher possible warning, the probability of the target vehicle being overweight is (25%, 50%)), and Level I (Extremely Severe, The highest possible warning, the probability of the target vehicle being overweight is (75%, 100%)), represented by blue, yellow, orange, and red respectively, and labeled in both Chinese and English, such as... Figure 6 As shown. Accordingly, the vehicle overload warning level measured in this embodiment is Level II (Severe, The higher possible warning), indicated by orange.

Claims

1. A method for early warning of overload on roads and bridges based on computer vision technology, characterized in that: Includes the following steps: Step 1: Deploy the data acquisition system: Select the lane to be detected, and use the existing cameras on the top of the lane to collect overall images of the vehicles in the lane; deploy cameras on the side of the lane to collect overall images of the side tires of the vehicles. Step 2: Overall vehicle feature recognition: Using images captured by the vehicle's roof camera, the vehicle information is automatically determined through a deep learning object detection algorithm: license plate, vehicle model, and number of axles; Step 3: Local Tire Feature Recognition: Using images captured by lane-side cameras, identify the tire model and tire-to-ground contact length. From the tire model, obtain the rim diameter and tire width, and thus the tire-to-ground contact area. Actual vehicle tire pressure follows a normal distribution. ,in This is the average tire pressure. The standard deviation of tire pressure includes: Step 3.1: Identify the special markings of different tire models through big data collection to determine the tire model, thereby obtaining the tire width and rim diameter; Step 3.2: Perform grayscale processing on the target vehicle tire image, perform edge detection on the tire edge, extract an approximate straight line segment as the contact length between the tire and the ground, and define the relative length of this straight line segment on the captured image as the pixel length of the tire contact with the ground, which will be used for the next step of conversion. Step 3.3: Use the identified wheel hub diameter to calibrate the image, and then use an algorithm based on the similarity principle to convert the pixel length of the tire contact with the ground into the actual contact length between the tire and the ground; Step 3.4: Analyze the contact length between the vehicle tire and the ground using a computer vision recognition system, automatically fill in the corresponding tire width based on the tire model, and then calculate the contact area between the vehicle tire and the ground based on the approximate rectangular shape of the tire-ground contact. The formula for expressing the tire-ground contact area is: in, For the first The contact area between the tire and the road surface. For the first The length of contact between the tire and the road surface. The system automatically matches the contact width value between the tire and the ground; Step 3.5: Match the tire pressure of the target tire from the database based on the vehicle model. and the weight limit for this vehicle type on roads and bridges. ; Step 4: Identify Vehicle Weight: Calculate the probability density function of the target vehicle's load by considering the contact area between each tire and the ground, tire pressure characteristics, and the number of axles; specifically: taking into account tire pressure fluctuations, adjust the tire pressure... By combining the number of vehicle axles and the contact area between a single tire of the target vehicle and the ground, the probability density function of the vehicle load is derived. The relationship between the probability density function of the vehicle load and the contact area and tire pressure is as follows: in, The probability density function of vehicle load. For vehicle load, , , This refers to the number of tires on the right side of the vehicle, i.e., the number of axles. For the first The contact area between the tire and the road surface; Step 5: Risk Warning for Overweight Vehicles: Identify vehicles whose weight requirement is less than the weight limit of the road or bridge; calculate the probability of the target vehicle exceeding the weight limit. Based on risk appetite, a warning percentage is set, and four levels of warnings are implemented, specifically including: Step 5.1, when hour, For weight limits on roads and bridges, if a vehicle exceeds its weight limit, the probability of the target vehicle exceeding its weight limit can be obtained from the probability density function of the vehicle load. Among them, the probability of a vehicle being overweight The expression is: in, The probability of the target vehicle being overweight; Step 5.2: The warning signal level is determined based on the probability range of the target vehicle and can be set by the relevant management unit according to the highway bridge grade and current bridge conditions. For newly built secondary highway bridges, the warning signal is set to Level IV: Level IV, generally the lowest possible warning, indicating a low probability of the target vehicle being overweight. Level III, Relatively Heavy, The high possible warning: The target vehicle has a high probability of being overweight. Level II, Severe: The higher possible warning is that the target vehicle is likely to be overweight. Level I, Extremely Serious: The highest possible warning, indicating the target vehicle has a high probability of being overweight. The colors are represented by blue, yellow, orange, and red in that order, and are indicated in both Chinese and English.

2. The method for early warning of overloaded roads and bridges based on computer vision technology according to claim 1, characterized in that, Step one specifically involves: Step 1.1: Adjust the field of view of the high-definition high-speed camera arranged on the side of the bridge lane to match the field of view of the existing capture camera at the top of the lane, so as to ensure that the two capture the same vehicle. Step 1.2: Use the lane-top camera to capture an image of the overall outline of the target vehicle; use the lane-side camera to capture images of passing vehicles as tire images of the target vehicle. Step 1.3: Filter the captured images and upload the clear and identifiable images to the computer.

3. The method for early warning of overload on roads and bridges based on computer vision technology according to claim 1, characterized in that, Step two specifically involves: using deep learning to extract features from the captured target vehicle outline image, and further performing vehicle type recognition, axle counting, and license plate recognition on the target vehicle.

Images