Vehicle brake adjustment coefficient calibration method based on visual perception and multi-source data fusion

Abstract

The invention discloses a vehicle brake adjustment coefficient calibration method based on visual perception and multi-source data fusion, and belongs to the field of traffic engineering. The method comprises the following steps: firstly, in a sensing identification stage, sensing the current driving environment and state of a vehicle by utilizing a sensing device carried by the vehicle, acquiring the vehicle type number and load information by combining a carrying shipping order, and calling historical driving data of a driver; then capturing a road image through a vehicle-mounted camera, and identifying the type of a road surface material by applying a YOLO-v11 algorithm; in the adjustment coefficient calibration stage, the vehicle load condition is analyzed according to the load information, the load adjustment coefficient for braking deceleration is calibrated, the road condition is analyzed by combining the road material and the environment information, and the road adjustment coefficient for braking deceleration is calibrated. The driving style is analyzed according to the historical behavior data of the driver, and a driving style adjustment coefficient aiming at the reaction time is calibrated; and finally, integrating the step-by-step adjustment coefficients of the influence factors, and providing a total adjustment parameter under multi-factor coupling for the calculation of the ideal car-following distance of the vehicle.

Description

TECHNICAL FIELD

[0001] The application belongs to the field of traffic engineering, and specifically relates to the cross application of vehicle braking control and visual perception technology, and researches a type identification and adjustment coefficient calibration method for key influencing factors in the vehicle braking process. The method realizes effective identification and collaborative correlation of braking influencing factors such as vehicle driving environment, self load state, road surface material type and driver driving style by fusing vehicle-mounted sensing device perception, visual image recognition and driver historical driving data analysis technology, and then completes scientific calibration of adjustment coefficients in each dimension based on the multi-factor correlation results, and finally forms total adjustment parameters under the coupling of multiple factors to provide core support for ideal following distance calculation of the vehicle. The method is suitable for vehicle active safety control, intelligent following system optimization and other scenes, and can especially solve the braking decision deviation problem caused by single factor independent consideration in traditional technology, improve the accuracy of vehicle following distance calculation under complex driving conditions, and provide technical support for dynamic adaptation of vehicle braking decision and actual working conditions. BACKGROUND

[0002] With the rapid development of intelligent traffic technology, vehicle active safety control has become a core research direction to improve road traffic safety and reduce traffic accident rate, and the precise calculation of ideal following distance in the vehicle following process directly determines the safety of braking decision. In actual driving scenes, the calculation result of following distance is jointly affected by multiple factors, including temperature and humidity, rainfall conditions of the vehicle driving environment, load state of the vehicle itself, material differences of the driving road surface, and operation habits and reaction characteristics of the driver, etc. These factors are interrelated and influence each other, and jointly determine the actual play effect of the vehicle braking performance.

[0003] The prior art has the problem of dispersed consideration of each factor when dealing with the above braking influencing factors, and it is difficult to form a comprehensive decision basis that fits the actual working conditions. On the one hand, in the road surface condition evaluation link, the traditional technology often only relies on environmental perception data to judge the road surface adhesion state, and does not fully combine the inherent characteristics of the road surface material. The microstructure and porosity of the OGFC surface layer, SMA surface layer, AC surface layer and cement road surface of the asphalt road surface are obviously different, which makes their anti-skid performance and braking adhesion coefficient under the same environmental conditions show significant differences. However, the traditional method often uses a unified standard when setting the road surface adjustment parameter, and does not adapt to the characteristics of different road surface materials, which eventually leads to a large deviation between the influence of road conditions on braking deceleration and the actual situation. On the other hand, in the driver reaction time selection link, the existing technology often uses a fixed reaction time threshold, without considering the differences in operating habits formed by different drivers in the long-term driving process. Some drivers operate gently, with low frequency of sudden acceleration and sudden deceleration, and the reaction time is relatively stable; some drivers operate aggressively, with frequent vehicle state switching, and the reaction time fluctuates greatly. The fixed reaction time threshold cannot adapt to the braking reaction characteristics of different drivers, further exacerbating the deviation of the following distance calculation.

[0004] In the calibration and application of the braking adjustment coefficient, for the load condition adjustment coefficient, the traditional method only sets the load condition as a single factor, without considering the actual influence of the road adhesion performance. Under the same load condition, the braking deceleration of dry road surface and water accumulation road surface has different sensitivity to load condition changes, and a single-dimensional coefficient cannot accurately reflect the actual braking performance difference. For the road surface adjustment coefficient, the absolute value of the adhesion coefficient is often directly used for calibration, without considering the driver reaction time, resulting in independent adjustment of the braking deceleration and the reaction time, and it is difficult to form a coordinated parameter system. More importantly, the existing technology lacks a multi-dimensional coefficient integration mechanism, and cannot couple the load condition, road surface, and driver operation characteristics, etc. The output following distance parameter is only suitable for a single working condition, and is prone to collision risk or reduces the traffic efficiency in complex scenarios.

[0005] In addition, the current road surface material identification based on YOLO series algorithm, vehicle-mounted multi-sensor fusion collection of environmental and vehicle data, and driver historical driving analysis technology are relatively mature, but the existing braking control technology does not fully integrate these achievements, and each sensing and data processing link is independent, multi-source data is difficult to link, and the collaborative value cannot be realized, further limiting the accuracy and adaptability of the following distance calculation.

[0006] In summary, the core problem of current vehicle following distance calculation technology is the failure to achieve the collaborative fusion of multiple braking influencing factors and the lack of scientific coupling mechanism of multi-dimensional adjustment coefficient. It is urgent to integrate vehicle-mounted sensing, visual recognition and historical data analysis technology, to complete the coefficient coupling calibration by establishing the correlation logic between factors, to improve the accuracy of following distance calculation, to provide support for braking decision that fits the actual working conditions, and finally to ensure the active safety performance of intelligent vehicles in complex scenarios. SUMMARY

[0007] The present application aims to solve the problem of lack of collaborative fusion of multiple braking influencing factors and insufficient coupling of adjustment coefficient in existing vehicle following distance calculation technology, and provides a vehicle braking influencing factor identification and adjustment coefficient calibration method based on visual perception. By integrating multi-source perception technology and data analysis means, the correlation identification of braking influencing factors and the scientific calibration of adjustment coefficient are realized, and finally the accuracy of vehicle ideal following distance calculation is improved, and technical support for braking decision that fits the actual working conditions is provided.

[0008] The technical solution of the present application specifically includes the following three core steps, and the overall process is shown in the accompanying Figure 1

[0009] (1) Multi-source perception of braking influencing factors

[0010] Regarding the identification of road surface material and environmental information, the camera and other visual perception devices carried by the vehicle are used to capture road images, and the YOLO-V11 model is used to process the images to accurately identify the type of road surface material. At the same time, through the temperature and humidity, rainfall sensors in the vehicle-mounted multi-sensor fusion system, the humidity, rainfall and temperature information of the driving environment is collected in real time, and the overall perception of the road environment state is completed.

[0011] Regarding the identification of vehicle load information, the vehicle-mounted cabin single is accessed to obtain the vehicle model and load basic data, and after calculating the actual load proportion, the K-means clustering algorithm is used to classify the load proportion, which is divided into 6 categories to cover the full state from empty load to overload. The clustering analysis process and results are shown in the accompanying Figure 3 , and Figure 4 , finally the current load category of the vehicle is determined.

[0012] Regarding the identification of driver driving style, the historical data of the driver's last 3 times of driving the same type of vehicle is accessed, the frequency of vehicle rapid acceleration / deceleration state under a specific road service level is counted, and based on the frequency distribution, the driver's driving style is judged as cautious, conservative and aggressive.

[0013] (2) Multi-dimensional braking adjustment coefficient calibration

[0014] ​Regarding the load condition adjustment coefficient calibration. Based on the vehicle load category obtained by K-means clustering, the brake deceleration adjustment coefficient corresponding to different load conditions is calibrated to realize the precise adaptation of the load factor to the brake deceleration.

[0015] Regarding the road adjustment coefficient calibration. The road adhesion coefficient is determined in combination with the identified road surface material and driving environment information. The road adjustment coefficient is calibrated according to the relative proportion of the actual adhesion coefficient and the reference value based on the adhesion coefficient of dry asphalt pavement. The front and rear vehicles are assumed to drive under the same road conditions, and the adjustment coefficient is not distinguished.

[0016] Regarding the driving style adjustment coefficient calibration, based on the driving style classification results of the driver, the driving style adjustment coefficient for the reaction time of the rear vehicle is calibrated by referring to the experimental values and experience values in existing literature data, to realize the differentiated adaptation of different driving styles to the brake reaction time.

[0017] (3) Total adjustment parameter fusion output

[0018] The above-mentioned load condition, road and driving style adjustment coefficients are integrated to form a total adjustment parameter system under the coupling of multiple factors. The load condition and road adjustment coefficients act on the adjustment of the brake deceleration of the front and rear vehicles, and the driving style adjustment coefficient acts on the adjustment of the reaction time of the rear vehicle alone. Finally, the total adjustment parameter adapted to the current driving condition is output, providing the core input parameter for the calculation of the ideal following distance of the vehicle.

[0019] Compared with the prior art, the present application has the following obvious advantages and beneficial effects:

[0020] Firstly, the present application solves the problem of incomplete processing of brake influencing factors and adjustment coefficients in traditional technology. The existing technology mainly analyzes road conditions, load conditions, drivers and other factors respectively, and lacks coordination logic between adjustment coefficients. However, the present application establishes an associated analysis logic between factors by integrating multiple data sources, calibrates each coefficient in combination with actual data and literature experience values, and forms a coupling system, so that the adjustment parameter can accurately reflect the brake performance under the combined action of multiple factors, and fundamentally solve the problem of following distance calculation deviating from the actual working condition.

[0021] Secondly, the technical process of the present application realizes integrated design, and has high practicability and easy landing. From the identification of brake influencing factors to the calibration of multi-dimensional adjustment coefficients, and then to the coupling output of total adjustment parameters, the whole process is closely connected and closely linked. And the hardware relied on by the process is the existing configuration or easily installed components of the vehicle. The YOLO-V11 image recognition model and K-means clustering algorithm used are mature and low in deployment difficulty, which greatly reduces the cost and implementation threshold of technology landing, and can easily adapt to various intelligent vehicles and be widely applied.

[0022] Finally, the application can significantly enhance the active safety capability of intelligent vehicles. Relying on the collaborative identification of multiple braking influencing factors such as road surface, load condition, driving style, etc., and the scientific coupling of corresponding adjustment coefficients, the total adjustment parameters output by the application can accurately optimize the calculation results of the ideal following distance of the vehicle. Even in complex driving scenarios such as variable weather, different load, and various driving habits, it can provide stable and reliable technical support for vehicle active safety control, and truly realize the dynamic balance of road traffic safety and efficiency. BRIEF DESCRIPTION OF DRAWINGS

[0023] Figure 1 Process schematic diagram of vehicle braking adjustment coefficient calibration method based on visual perception and multi-source data fusion

[0024] Figure 2 Schematic diagram of road surface identification result

[0025] Figure 3 Schematic diagram of vehicle load condition category clustering result

[0026] Figure 4 Schematic diagram of vehicle load condition category clustering result rationality test DETAILED DESCRIPTION

[0027] According to the above description, the following is a specific implementation process, but the scope of protection of this patent is not limited to this implementation process.

[0028] Step 1: Multi-source perception identification of braking influencing factors

[0029] This step completes the comprehensive collection and identification of relevant data of vehicle driving environment, load condition, road surface material and driver driving style through existing physical sensing devices of vehicle, load cabin single data, driver historical driving records and visual perception equipment, providing basic data support for subsequent adjustment coefficient calibration.

[0030] Step 1.1: Collection of driving environment and vehicle basic information

[0031] The existing physical sensing devices carried by the vehicle are enabled to collect real-time humidity, rainfall and environmental temperature data of the current driving environment. The data is stored to the central processing unit through the vehicle data transmission module; the load cabin single is retrieved through the vehicle data terminal to extract the specific model and load condition related information of the vehicle; the historical driving data of the driver of the vehicle in the last 3 times of driving the same model vehicle is synchronously retrieved, and the data content needs to include the coordinate points of each time in the journey, the driving speed, and the single journey data time length is not less than 30 minutes, ensuring the data integrity and effectiveness of subsequent driving style analysis.

[0032] Step 1.2: Road surface material identification

[0033] Step 1.2.1: Road surface material image database construction

[0034] The database contains actual collected images of four types of road surfaces: asphalt OGFC surface layer, asphalt SMA surface layer, asphalt AC surface layer, and cement pavement. The cement pavement is no longer classified into specific categories. For each type of road surface, more than 10 road images are selected for training and learning. Each road image is taken from the perspective of a driver looking straight ahead, and covers three scenarios: normal lighting conditions, night lighting conditions, and rainy lighting conditions, to adapt to the identification needs in different environments. The road surface image data of the same type is imported into the Roboflow platform, and the road surface part in the picture is selected using a manual annotation tool, and the corresponding label is added according to the definition of the road surface type. After labeling, all images are exported in YOLO-v11 format, and the export file contains train, valid, and test subdirectories and corresponding data.yaml configuration files, meeting the format requirements of model training. The exported data set is used for model training, and the best.pt weight file is finally obtained, which is used for subsequent accurate detection of road surface types.

[0035] Step 1.2.2: Real-time identification of road surface material

[0036] The road image information is captured by the visual perception device mounted on the vehicle, and the image is input into the pre-trained model to identify the road surface elements and further accurately determine the road surface material type, ensuring that the identification results cover the four types of road surfaces: asphalt OGFC surface layer, asphalt SMA surface layer, asphalt AC surface layer, and cement pavement.

[0037] Step 1.3: Vehicle load condition and driver behavior data extraction

[0038] According to the actual load obtained from the cargo manifest and the rated load of the vehicle, the actual load ratio is calculated, and the formula is: actual load ratio = (actual load / rated load) x 100%. At the same time, from the retrieved historical driving data of the driver, the road sections with a road service level of B are selected, and the vehicle speed information at each time in the road section is extracted for subsequent acceleration calculation and driving style analysis.

[0039] Step 2: Multi-dimensional adjustment coefficient calibration

[0040] Step 2.1: Vehicle load condition analysis and adjustment coefficient calibration

[0041] Step 2.1.1: Cluster analysis of load condition

[0042] K-means clustering analysis is performed with actual load ratio as an index. In the clustering process, 207 groups of load ratio data collected in the past three years are used as training samples, and the number of clusters is set to 6. Through iterative calculation of the Euclidean distance between the sample and each cluster center, the sample is assigned to the nearest cluster category until the cluster center no longer changes. Finally, the cluster center, critical value and corresponding load of each category are determined as follows: when the actual value of vehicle load is 0.00% (including) -14.00% (not including) of the rated load, the vehicle load condition is determined to be empty load, and the cluster center of this category is 5.06%; when the actual value of vehicle load is 14.00% (including) -36.00% (not including) of the rated load, the vehicle load condition is determined to be light load, and the cluster center of this category is 23.93%; when the actual value of vehicle load is 36.00% (including) -62.00% (not including) of the rated load, the vehicle load condition is determined to be half load, and the cluster center of this category is 47.77%; when the actual value of vehicle load is 62.00% (including) -88.00% (not including) of the rated load, the vehicle load condition is determined to be standard load, and the cluster center of this category is 76.00%; when the actual value of vehicle load is 88.00% (including) -110.00% (not including) of the rated load, the vehicle load condition is determined to be overload, and the cluster center of this category is 99.83%; when the actual value of vehicle load is 110% (including) or more of the rated load, the vehicle load condition is determined to be overload, and the cluster center of this category is 149.04%.

[0043] Step 2.1.2: Adjustment coefficient calibration of load condition

[0044] According to the disclosed truck real vehicle dry braking experimental results, combined with the above load classification conditions, the load adjustment coefficient A for adjusting the braking deceleration of the front and rear vehicles is calibrated load , and the specific values are as follows: when the vehicle load condition is empty load, A lo = 1.000; when the vehicle load condition is light load, A load = 0.960; when the vehicle load condition is half load, A load = 0.920; when the vehicle load condition is standard load, A load = 0.880; when the vehicle load condition is full load, A load = 0.839; when the vehicle load condition is overload, A load = 0.000; this adjustment coefficient is used to adjust the braking deceleration of the vehicle and the front vehicle.

[0045] Among them, the adjustment coefficients corresponding to empty load and full load are experimental values, and the adjustment coefficients corresponding to light load, half load, standard load and overload are inferred based on the experimental values.

[0046] Step 2.2: Road condition analysis and adjustment coefficient calibration

[0047] Step 2.2.1: Road surface adjustment coefficient determination

[0048] In combination with the previously identified road surface material type and the collected driving environment information, including dry, wet, water accumulation, snow accumulation, ice, and water sliding, the corresponding road adhesion coefficient μ is determined by table lookup. In this method, the default value of the road adhesion coefficient μ is only 0.700, 0.500, and 0.300. When the actual value of the road adhesion coefficient corresponds to the default value, it is matched according to the nearest principle. The specific correspondence is as follows:

[0049] When the vehicle is driving on a road with no precipitation and the air temperature is greater than or equal to 0°C, and the road surface water film thickness is 0, it is determined that the road surface environment is dry. At this time, when the road surface material is asphalt OFGC surface layer, the road adhesion coefficient default value μ = 0.700; when the road surface material is asphalt SMA surface layer, the road adhesion coefficient default value μ = 0.700; when the road surface material is asphalt AC surface layer, the road adhesion coefficient default value μ = 0.700; when the road surface material is cement, the road adhesion coefficient default value μ = 0.500;

[0050] When the vehicle is driving on a road with precipitation and the air temperature is greater than or equal to 0°C, and the road surface water film thickness is less than 1mm, it is determined that the road surface environment is wet. At this time, when the road surface material is asphalt OFGC surface layer, the road adhesion coefficient default value μ = 0.700; when the road surface material is asphalt SMA surface layer, the road adhesion coefficient default value μ = 0.700; when the road surface material is asphalt AC surface layer, the road adhesion coefficient default value μ = 0.500; when the road surface material is cement, the road adhesion coefficient default value μ = 0.300;

[0051] When the vehicle is driving on a road with precipitation and the air temperature is greater than or equal to 0°C, and the road surface water film thickness is greater than 1mm, it is determined that the road surface environment is water accumulation. At this time, when the road surface material is asphalt OFGC surface layer, the road adhesion coefficient default value μ = 0.500; when the road surface material is asphalt SMA surface layer, the road adhesion coefficient default value μ = 0.500; when the road surface material is asphalt AC surface layer, the road adhesion coefficient default value μ = 0.500; when the road surface material is cement, the road adhesion coefficient default value μ = 0.300;

[0052] When the vehicle is driving on a road with snowfall, it is determined that the road surface environment is snow accumulation. At this time, when the road surface material is asphalt OFGC surface layer, the road adhesion coefficient default value μ = 0.300; when the road surface material is asphalt SMA surface layer, the road adhesion coefficient default value μ = 0.300; when the road surface material is asphalt AC surface layer, the road adhesion coefficient default value μ = 0.300; when the road surface material is cement, the road adhesion coefficient default value μ = 0.300;

[0053] When the vehicle travels on the road weather condition accompanied by precipitation, and the air temperature is less than 0℃, it is determined that the road surface environment is possibly icy at this time; at this time, when the road surface material is asphalt OFGC surface layer, the default value of the road surface adhesion coefficient μ = 0.300; when the road surface material is asphalt SMA surface layer, the default value of the road surface adhesion coefficient μ = 0.300; when the road surface material is asphalt AC surface layer, the default value of the road surface adhesion coefficient μ = 0.300; when the road surface material is cement, the default value of the road surface adhesion coefficient μ = 0.300.

[0054] Step 2.2.2: Calibration of road surface adjustment coefficient

[0055] Taking the adhesion coefficient μ = 0.700 of the dry asphalt road surface as the benchmark, according to the relative value of the actual road surface adhesion coefficient compared with the benchmark value, and combining the truck real vehicle braking test results, the road surface adjustment coefficient A for adjusting the front and rear vehicle braking deceleration is calibrated road , and the specific values are as follows: when the road surface adhesion coefficient is 0.700, the relative value of the road surface adhesion coefficient relative to the road surface adhesion coefficient benchmark value is 1.000, A road = 1.000; when the road surface adhesion coefficient is 0.500, the relative value of the road surface adhesion coefficient relative to the road surface adhesion coefficient benchmark value is 0.714, A road = 0.773; when the road surface adhesion coefficient is 0.300, the relative value of the road surface adhesion coefficient relative to the road surface adhesion coefficient benchmark value is 0.429, A road = 0.554; when the road surface adhesion coefficient is 0.000, the relative value of the road surface adhesion coefficient relative to the road surface adhesion coefficient benchmark value is 0.000, A road = 0.000; The adjustment coefficient is used to adjust the braking deceleration of the vehicle and the front vehicle.

[0056] Among them, the adjustment coefficient of the actual road surface adhesion coefficient of 0.700 and 0.300 is the experimental value, and the adjustment coefficient corresponding to the rest of the road surface is inferred from the experimental value. It should be noted that the following vehicles are assumed to travel on the same road surface conditions during the following process, i.e. the road surface material and the driving environment are the same, so the adjustment coefficient does not distinguish between the front and rear vehicles.

[0057] Step 2.3: Analysis of the driving style of the rear driver and calibration of the driving style adjustment coefficient Step 2.3.1: Analysis of the driving style

[0058] Based on the historical driving data of the driver extracted in the early stage, the acceleration value a history of the vehicle in the road service level B section (the above definition is from the HCM manual) is calculated. history When the absolute value of the acceleration |a 2When a vehicle is in a state of rapid acceleration or deceleration, this criterion is determined by acceleration clustering analysis of existing truck driver driving data. Positive values ​​are used in the analysis, and then the frequency 'a' of the vehicle entering this state is extracted and counted. account The unit is f / h. When performing statistics, note that the condition |a| must be satisfied continuously between adjacent time points. history |≥0.501m / s 2 In cases where frequency is not repeated, the frequency will not be counted again. This frequency will be used as an indicator to classify driver driving styles, with the following specific criteria: when a... account When the value is less than 8 (excluding 8), the driver's driving style is judged to be cautious; when a account When the value is greater than or equal to 8 and less than 20 (excluding 20), the driver's driving style is judged to be conservative; when a account When the value is greater than or equal to 20, the driver's driving style is determined to be aggressive.

[0059] Step 2.3.2: Calibration of Driving Style Adjustment Coefficient

[0060] Based on existing literature data and empirical values, the coefficient A used for adjusting the reaction time of the following vehicle is calibrated. style The specific values ​​are as follows: A is determined when the driver's driving style is aggressive. style =1.25; When the driver's driving style is aggressive, A style =1.00; When the driver's driving style is aggressive, A style =0.80; all values ​​are taken from experience. This adjustment factor is adjusted for the braking reaction time of this vehicle.

[0061] Step 3: Total Adjustment Coefficient Fusion Output

[0062] Based on the above-calibrated load adjustment factor A load Road surface adjustment coefficient A road and driving style adjustment coefficient A style This establishes a comprehensive adjustment parameter system for calculating vehicle following distance, specifically divided into adjustments for the braking deceleration of the front and rear vehicles and the reaction time of the following vehicle. First, a comprehensive adjustment coefficient for the braking deceleration of the front and rear vehicles is given. This coefficient is the product of the load condition adjustment coefficient and the road condition adjustment coefficient, used to adjust the braking reaction time of the vehicle itself. Then, a comprehensive adjustment coefficient for the vehicle's reaction time is given, which is equivalent to the driving style adjustment coefficient, used to adjust the braking deceleration of both the vehicle itself and the vehicle in front.

Claims

1. A method for calibrating vehicle braking adjustment coefficients based on visual perception and multi-source data fusion, characterized in that: First, in the perception and recognition stage, the vehicle's existing physical sensing devices are used to perceive the vehicle's current driving status and environment, including the humidity or rainfall and ambient temperature of the driving environment. Then, the vehicle's cargo manifest is used to obtain the vehicle's specific model and load information. At the same time, the driver's historical driving data records are obtained. Subsequently, the vehicle's visual perception equipment is used to capture road image information and identify the road surface material. Secondly, in the adjustment coefficient calibration stage, the vehicle load information and vehicle model obtained from the environmental information perceived in the previous steps are combined to analyze the current vehicle load situation and calibrate the load situation adjustment coefficient; then, the road condition information and environmental information perceived in the previous steps are combined to analyze the road condition situation and calibrate the road condition adjustment coefficient; combined with the driver information obtained in the previous steps, the driving style of the vehicle is analyzed and the driving style adjustment coefficient is calibrated; finally, the step-by-step adjustment coefficients of various influencing factors are combined to give the total adjustment coefficient under a specific state.

2. The vehicle braking adjustment coefficient calibration method based on visual perception and multi-source data fusion according to claim 1, characterized in that, The specific methods for identifying road surface materials include: Step 1: Construct a road surface material image database The database contains image information for asphalt pavement OGFC surface layer, asphalt pavement SMA surface layer, asphalt pavement AC surface layer, and cement pavement. Each type of pavement image is trained using more than 10 images of that type. Each image is taken from a level driving perspective and consists of three images: one under normal lighting conditions, one under nighttime lighting conditions, and one under cloudy / rainy lighting conditions. The purpose of using a large number of similar pavement images for training is to improve recognition speed and accuracy. The aforementioned pavement images were collected from actual roads. Road surface image data of the same type were entered into Roboflow. A manual annotation tool was used to select the road surface portion of each image and add corresponding labels, with label names defined according to the road surface type. All annotated images were exported in YOLO-v11 format, including train, valid, and test subdirectories and the corresponding data.yaml configuration file, to meet model training requirements. The exported dataset was used for training, resulting in a best.pt weight file, which was used during the model inference stage to achieve accurate road surface type detection. Step 2: Road surface material identification The identification range includes four types of asphalt pavement: OGFC surface layer, SMA surface layer, AC surface layer, and cement pavement. Cement pavement is no longer distinguished by specific category. Based on the captured road image information, the YOLO-v11 algorithm is first used to identify road elements, and then the road material type is identified.

3. The vehicle braking adjustment coefficient calibration method based on visual perception and multi-source data fusion according to claim 1, characterized in that, The specific methods for analyzing vehicle load conditions and calibrating load condition adjustment coefficients include: The vehicle load analysis considers only the specific vehicle's load situation, relying on the load information obtained in previous steps to determine the corresponding load conditions. The actual load ratio is calculated based on the load capacity obtained from existing truck manifest data and the vehicle's rated load capacity. This ratio is used as an indicator for K-means clustering. When the number of clusters is greater than 6, the total sum of squared errors significantly decreases; therefore, the vehicle is clustered into 6 classes. When the actual load value is between 0.00% (inclusive) and 14.00% (exclusive) of the rated load, the vehicle is classified as unloaded, with a cluster center of 5.06%. When the actual load value is between 14.00% (inclusive) and 36.00% (exclusive) of the rated load, the vehicle is classified as lightly loaded, with a cluster center of 23. 0.93%; when the actual vehicle load is 36.00% (inclusive) to 62.00% (exclusive) of the rated load, the vehicle load is classified as half-loaded, and the cluster center for this category is 47.77%; when the actual vehicle load is 62.00% (inclusive) to 88.00% (exclusive) of the rated load, the vehicle load is classified as standard load, and the cluster center for this category is 76.00%; when the actual vehicle load is 88.00% (inclusive) to 110.00% (exclusive) of the rated load, the vehicle load is classified as overloaded, and the cluster center for this category is 99.83%; when the actual vehicle load is 110% (inclusive) or more of the rated load, the vehicle load is classified as overloaded, and the cluster center for this category is 149.04%. Based on the publicly available results of truck braking tests on dry ground, an adjustment factor A for vehicle load conditions is given. load This adjustment factor is used to adjust the braking deceleration of the front and rear vehicles when calculating the following distance. The values ​​are as follows: When the vehicle is unloaded, A load =1.000; When the vehicle is lightly loaded, A load =0.960; When the vehicle is half-loaded, A load =0.920; When the vehicle load is standard load, A load =0.880; When the vehicle is fully loaded, A load =0.839; When the vehicle is overloaded, A load =0.000; This adjustment factor is applied to the braking deceleration of both the vehicle in front and the vehicle itself.

4. The vehicle braking adjustment coefficient calibration method based on visual perception and multi-source data fusion according to claim 1, characterized in that, The analysis of road surface conditions and the calibration method for road surface condition adjustment coefficients include: Based on the specific road surface material type identified in the previous steps, and combined with the vehicle driving environment, the default value μ of the road surface adhesion coefficient is matched. In this method, the default value μ of the road surface adhesion coefficient is only 0.700, 0.500, and 0.

300. When the actual value of the road surface adhesion coefficient corresponds to the default value, the nearest value is matched. When the weather conditions on the road where the vehicle is traveling are dry, with no precipitation, an air temperature greater than or equal to 0℃, and a water film thickness of 0 on the road surface, the road surface environment is considered dry. In this case, the default value for the road adhesion coefficient is μ = 0.700 when the road surface material is OFGC asphalt, SMA asphalt, AC asphalt, or cement. When the weather conditions on the road where the vehicle is traveling are accompanied by precipitation, and the air temperature is greater than or equal to 0℃ and the water film thickness on the road surface is less than 1mm, the road surface environment is judged to be wet. At this time, the default value of the road adhesion coefficient is μ = 0.700 when the road surface material is asphalt OFGC surface layer; μ = 0.700 when the road surface material is asphalt SMA surface layer; μ = 0.500 when the road surface material is asphalt AC surface layer; and μ = 0.300 when the road surface material is cement. When the weather conditions on the road where the vehicle is traveling are accompanied by precipitation, and the air temperature is greater than or equal to 0℃ and the water film thickness on the road surface is greater than 1mm, the road surface environment is judged to be waterlogged. At this time, the default value of the road adhesion coefficient is μ = 0.500 when the road surface material is asphalt OFGC surface layer; μ = 0.500 when the road surface material is asphalt SMA surface layer; μ = 0.500 when the road surface material is asphalt AC surface layer; and μ = 0.300 when the road surface material is cement. When the weather conditions on the road where the vehicle is traveling include snowfall, the road surface environment is determined to be snow-covered. At this time, the default value of the road adhesion coefficient is μ = 0.300 when the road surface material is asphalt OFGC, SMA, AC, or cement. When the weather conditions on the road where the vehicle is traveling are accompanied by precipitation and the temperature is below 0℃, the road surface environment is judged to be likely to freeze. At this time, the default value of the road adhesion coefficient is μ = 0.300 when the road surface material is OFGC asphalt surface layer; μ = 0.300 when the road surface material is SMA asphalt surface layer; μ = 0.300 when the road surface material is AC asphalt surface layer; and μ = 0.300 when the road surface material is cement. Based on the braking test results of the truck under different road surface adhesion coefficients, the vehicle road condition adjustment coefficient A is given. road This adjustment factor is used to adjust the braking deceleration of the front and rear vehicles in the following distance calculation; Using μ = 0.700 as the baseline, the adjustment factor A road The value is given based on the relative value of the actual road surface adhesion coefficient to μ = 0.700, not directly based on the actual road surface adhesion coefficient. Meanwhile, during the following process, the vehicles in front and behind are assumed to be traveling on the same road surface, i.e., the road surface material and driving environment are the same; therefore, the adjustment coefficient A is... road Without distinguishing between front and rear vehicles; A road The possible values ​​are as follows; When the road surface adhesion coefficient is 0.700, the relative value of the road surface adhesion coefficient to the reference value is 1.000, A road =1.000; When the road surface adhesion coefficient is 0.500, the relative value of the road surface adhesion coefficient to the reference value is 0.714, A road =0.773; When the road surface adhesion coefficient is 0.300, the relative value of the road surface adhesion coefficient to the reference value is 0.429, A road =0.554; When the road surface adhesion coefficient is 0.000, the relative value of the road surface adhesion coefficient compared to the reference value is 0.000, A road =0.000; This adjustment factor is applied to the braking deceleration of both the vehicle in front and the vehicle itself.

5. The vehicle braking adjustment coefficient calibration method based on visual perception and multi-source data fusion according to claim 1, characterized in that, The method for analyzing the driver's driving style and calibrating the driving style adjustment coefficient of this vehicle includes: Based on the driver's historical behavior data from the last three trips involving the same model of vehicle, including coordinates and speed at various points along the journey, the vehicle's acceleration value 'a' was calculated when the road service level was B during the historical journey. history ; when |a history |≥0.501m / s 2 When a vehicle is in a state of rapid acceleration or deceleration, this criterion is determined by acceleration clustering analysis of existing truck driver driving data, with positive values ​​used in the analysis. Subsequently, the frequency 'a' of vehicles entering a state of rapid acceleration or deceleration is extracted and counted. account The unit is f / h; If |a| persists in adjacent moments history |≥0.501m / s 2 In cases where frequencies are not counted repeatedly, the final count is based on a. account As an indicator for analyzing driver driving style; when a account When the value is less than 8 (excluding 8), the driver's driving style is judged to be cautious; when a account When the value is greater than or equal to 8 and less than 20 (excluding 20), the driver's driving style is judged to be conservative; when a account When the value is greater than or equal to 20, the driver's driving style is determined to be aggressive. Considering the reaction time (PRT), the adjustment factor A for different driving styles of the driver of this vehicle is calculated. style This adjustment factor is used to adjust the vehicle's reaction time when calculating the following distance, and its value is as follows: When the driver's driving style is aggressive, A style =1.25; When the driver's driving style is aggressive, A style =1.00; When the driver's driving style is aggressive, A style =0.80; all values ​​are taken from experience. This adjustment factor is applied to the braking response time of this vehicle.

6. The vehicle braking adjustment coefficient calibration method based on visual perception and multi-source data fusion according to claim 1, characterized in that, The methods for calibrating the total adjustment factor include: The total adjustment coefficient for calculating the following distance of a vehicle includes two aspects: adjustment of the braking deceleration of the vehicles in front and behind, and adjustment of the vehicle's own reaction time. First, the total adjustment coefficient for the braking deceleration of the vehicles in front and behind is given. This total adjustment coefficient is the product of the load condition adjustment coefficient and the road condition adjustment coefficient, which is used to adjust the braking reaction time of the vehicle itself. Then, the total adjustment coefficient for the vehicle's own reaction time is given. This total adjustment coefficient is equivalent to the driving style adjustment coefficient, which is used to adjust the braking deceleration of both the vehicle itself and the vehicle in front.

Images