A multi-source fusion estimation method for road surface adhesion coefficient based on vehicle-road cooperation

Abstract

The present invention relates to a multi-source fusion estimation method for road surface adhesion coefficient based on vehicle-road cooperation, including: S1. Obtain the road surface image of a specified section by using a road-side camera, input the road surface image into a pre-constructed road surface classification model, and output the road surface classification result at the road side; S2. The in-vehicle terminal obtains and recognizes the road surface image in front of the vehicle to obtain the road surface classification result at the vehicle side; S3. Fuse the road surface classification result at the vehicle side and the road surface classification result at the road side to obtain the fused road surface classification result; S4. Perform spatio-temporal conversion on the fused classification result to obtain the current visual estimation value θ of the peak adhesion coefficient image ; S5. Establish a Burckhardt tire model, calculate the dynamic estimation value θ of the peak adhesion coefficient dynamics , and fuse the visual estimation value θ image and the dynamic estimation value θ dynamics to obtain the fused estimation result θ of the peak adhesion coefficient fusion . Compared with the prior art, the present invention has the advantages of high estimation accuracy, good real-time performance, and little influence by the occlusion of the vehicle in front

Description

Technical Field

[0001] The present invention relates to the technical field of electric vehicle control, and particularly to a multi-source fusion estimation method for road surface adhesion coefficient based on vehicle-road cooperation. Background Art

[0002] The adhesion coefficient is the ratio of the adhesion force to the wheel normal (the direction perpendicular to the road surface) pressure, and it can be regarded as the static friction coefficient between the tire and the road surface. The larger this coefficient is, the greater the available adhesion force is, and the less likely the vehicle is to skid. The adhesion coefficient mainly depends on the roughness and wet and muddy degree of the road surface, the pattern and air pressure of the tire, as well as the vehicle speed and load, etc.

[0003] If the peak adhesion coefficient can be accurately and real-time estimated, and the vehicle active safety system can change the control strategy in real time according to the estimated road surface information, it is beneficial to improve the control effect of the system on the vehicle on the road surface with changing states, which has important significance for the active safety of the vehicle.

[0004] According to the recognition principle, the estimation methods of the peak adhesion coefficient can be divided into two categories: cause-based estimation methods (Cause-Based) and effect-based estimation methods (Effect-Based). Among them, the cause-based method needs to obtain road surface related parameters with the help of sensors (such as cameras, lidar, etc.), and establish a mathematical model representing the mapping relationship between these parameters and the peak adhesion coefficient, and then use the designed mathematical model to calculate the magnitude of the peak adhesion coefficient. The estimation result obtained by this method has foresight and predictability, which is beneficial for the vehicle decision-making system to adjust the vehicle control strategy in advance according to the estimated value. However, this method cannot fully reflect the real wheel-ground interaction relationship, and the on-vehicle camera will be affected by factors such as occlusion and motion blur, resulting in a decrease in image quality and affecting the accuracy.

[0005] The effect-based estimation method estimates the magnitude of the peak adhesion coefficient by measuring the dynamic response of the whole vehicle or the wheel caused by the change of road surface excitation. This method can obtain a relatively accurate numerical solution when the road surface excitation is large enough, and the accuracy of the estimation result is relatively high. However, when the road surface excitation is small, the estimation result has a large estimation error. In addition, this method can only estimate the road surface that the vehicle has passed, and cannot predict the road surface conditions ahead, resulting in little reference value of the estimation result for vehicle decision-making and planning. Summary of the Invention

[0006] The purpose of the present invention is to overcome the defects existing in the above-mentioned prior art and provide a multi-source fusion estimation method for road surface adhesion coefficient based on vehicle-road cooperation, which can obtain the peak adhesion coefficient estimation result in real time and accurately, so as to be beneficial for the vehicle active safety system to change the control strategy in real time according to the estimation result and improve the vehicle control effect.

[0007] The object of the present invention can be achieved by the following technical solutions: A multi-source fusion estimation method for road surface adhesion coefficient based on vehicle-road cooperation, comprising the following steps:

[0008] S1. Use a roadside camera to obtain a road surface image of a specified section, input the road surface image into a pre-constructed road surface classification model, and output the road surface classification result at the roadside;

[0009] S2. The in-vehicle terminal obtains and identifies the road surface image in front of the vehicle to obtain the road surface classification result at the vehicle end;

[0010] S3. Fuse the road surface classification result at the vehicle end and the road surface classification result at the roadside to obtain a road surface fusion classification result;

[0011] S4. Perform spatio-temporal conversion on the fusion classification result to obtain the current peak adhesion coefficient visual estimation value θ image ;

[0012] S5. Establish a Burckhardt tire model, calculate the dynamic estimation value θ of the peak adhesion coefficient dynamics , and fuse the visual estimation value θ image and the dynamic estimation value θ dynamics to obtain the fusion estimation result θ of the peak adhesion coefficient fusion .

[0013] Further, the road surface classification model in step S1 is specifically constructed through the following process:

[0014] Obtain historical road surface images of different types, including dry asphalt roads, wet asphalt roads, snow-covered roads, and concrete roads;

[0015] Label the road surface types of the obtained historical road surface images;

[0016] Use the labeled road surface images to train a deep convolutional neural network to obtain a road surface classification model.

[0017] Further, the specific process of outputting the road surface classification result at the roadside in step S1 is as follows:

[0018] Determine the region of interest (ROI) for roadside unit image processing, which is a rectangular road surface area of a1m × a2 m in front of the camera;

[0019] Obtain a bird's-eye view of the road surface image in the ROI through inverse perspective transformation;

[0020] Divide the road surface image into n×n grids;

[0021] Each small grid is input into the road surface classification network, and combined with the position information corresponding to the grid, a road surface type distribution result matrix A is obtained.

[0022] Further, the specific process of obtaining the road surface classification result at the vehicle end in step S2 is as follows:

[0023] Determine the ROI for on-vehicle terminal image processing, which is a square road surface area of b m×b m at a distance of l m in front of the vehicle;

[0024] The road surface image in the ROI is transformed into a bird's-eye view through inverse perspective transformation;

[0025] The ROI is divided into m×m grids, and each grid is respectively input into the classification network to obtain the classification result of each grid, and then the road surface classification result matrix X at the vehicle end is obtained.

[0026] Further, step S3 specifically includes the following steps:

[0027] S31. The roadside unit transmits the road surface type distribution result matrix A to the on-vehicle terminal through DSRC (Dedicated Short Range Communication);

[0028] S32. Perform spatial synchronization: Using the vehicle position and the position of the roadside unit, respectively determine the vehicle coordinate system and the roadside unit coordinate system, and determine the position (xV, yV) of the vehicle in the roadside unit coordinate system;

[0029] S33. Calculate the position of the vehicle ROI in the roadside unit coordinate system, that is, the four-point coordinates of the ROI;

[0030] S34. Calculate the roadside unit road surface type result matrix Y in the vehicle ROI position area, and calculate the road surface classification result matrix C of the vehicle ROI;

[0031] S35. For the road surface classification result matrix C, each row votes according to the relative majority voting method to obtain the final classification result, and finally obtain a matrix D containing the fusion recognition result of the position information and the corresponding road surface type.

[0032] Further, the vehicle position in step S32 is specifically obtained through an on-vehicle positioning system.

[0033] Further, the specific process of step S34 is as follows:

[0034] S341. Calculate the road surface type results of each small grid in the ROI respectively. Each small grid will cover 1 to 4 grid areas in the distribution map, corresponding to the road surface type results I1, I2, I3, I4 in the distribution map, as shown in the following formula, and then obtain the matrix Y;

[0035]

[0036] Among them, S v (iden) represents the classification result of the small grid, and s iv (iden) represents the area occupied by each road surface in the small grid;

[0037] S342. As shown in the following formula, calculate the road surface classification result matrix C of the vehicle ROI;

[0038]

[0039] Among them, the matrices Y1 and Y2 are obtained by repeating step S341 after longitudinally moving the position of the ROI in the distribution map by ±x m.

[0040] Furthermore, the elements of the matrix D containing the fusion recognition result of the position information and the corresponding road surface type in step S35 are specifically:

[0041]

[0042] Among them, H v (iden) is the voting prediction result, and h iv (iden) is the classification result of the single-grid visual fusion classifier.

[0043] Furthermore, the specific process of step S4 is:

[0044] Record the cumulative displacement s of the vehicle t ;

[0045] The vehicle displacement corresponding to the current ROI is s t1 = s t + l + b / 3, s t2 = s t + l + 2b / 3, s t3 = s t + l + b;

[0046] Obtain a new road surface type matrix D', and the i-th row of D' is (s ti , I i ), where s ti is the vehicle displacement, and I i is the corresponding road surface classification result;

[0047] When I m in the matrix D' is different from I m-1 , record s tm and I m ;

[0048] Final output result I of road surface type t as follows:

[0049]

[0050] According to the final output result of the road surface type, the visual estimated value θ of the peak adhesion coefficient is obtained image : The visual estimated values of the peak adhesion coefficients corresponding to dry asphalt pavement, wet asphalt pavement, snow pavement and concrete pavement are 0.8, 0.6, 0.4 and 0.8 respectively.

[0051] Furthermore, the step S5 specifically includes the following steps:

[0052] S51. Establish the Burckhardt tire model:

[0053]

[0054] where μ(θ,λ) is the utilization adhesion coefficient of the current tire to the ground, θ is the peak adhesion coefficient, λ is the slip ratio, c1 is the longitudinal slip stiffness of the tire, and c2, c3, c4 are the curve shape control parameters after the curve crosses the peak point;

[0055] S52. According to the current tire longitudinal force and slip ratio, obtain the dynamic estimated value θ of the current peak adhesion coefficient dynamics :

[0056]

[0057] F x = T m / R

[0058] where T m is the wheel-end torque, R is the tire rolling radius, F x is the longitudinal force received by the tire, is the estimated value of the peak adhesion coefficient, θ(λ,F x ) is the peak adhesion coefficient obtained according to the longitudinal force and slip ratio, and γ is the dynamic estimator gain;

[0059] S53. Establish a fuzzy logic, with the input being the difference |θ image -θ fusion | between the current fusion estimator estimated value and the visual estimated value and the visual estimated value θ image , and the output being the dynamic estimator γ, thereby obtaining the fusion estimated result θ fusion .

[0060] Compared with the prior art, the present invention has the following advantages:

[0061] 1. The present invention obtains road surface images through in-vehicle cameras, obtains the road surface type results at the vehicle end through image processing, combines with roadside units for classification to obtain the road surface type results at the roadside end, so as to obtain a fused road surface type recognition result, and then performs spatio-temporal conversion on the fused classification result to obtain a visual estimated value of the peak adhesion coefficient. Compared with traditional visual classification methods, it has the advantages of high accuracy and good robustness.

[0062] 2. Based on the road surface estimation of tire longitudinal dynamics, the present invention calculates the dynamic estimated value of the peak adhesion coefficient by establishing a Burckhardt tire model, and then combines it with the visual estimated value, which can correct the parameters of the dynamic estimator. Compared with traditional estimation methods, it has a faster convergence speed and better real-time performance.

[0063] 3. The present invention estimates the peak adhesion coefficient by establishing fuzzy logic to achieve the fusion of vision and dynamics. When the excitation of the tire is small, the fused estimated value can be maintained at an empirical value close to the true value with the help of the visual estimated value, ensuring that the estimation error of the estimator is small and improving the accuracy of the estimation result. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] Figure 1 is a schematic diagram of the method flow of the present invention;

[0065] Figure 2 is a schematic diagram of the application framework of the embodiment;

[0066] Figure 3 is a position diagram of the vehicle in the roadside unit coordinate system in the embodiment;

[0067] Figure 4 is a schematic diagram of the ROI in the road surface coordinate system in the embodiment. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0068] The present invention will be described in detail below with reference to the drawings and specific embodiments.

[0069] Embodiment

[0070] As Figure 1 shown, a multi-source fusion estimation method for road surface adhesion coefficient based on vehicle-road cooperation includes the following steps:

[0071] S1. Use a roadside camera to obtain road surface images of a specified section, and input the road surface images into a road surface classification model to obtain the road surface classification results at the roadside end;

[0072] S2. The in-vehicle terminal obtains and recognizes the road surface images in front of the vehicle to obtain the road surface classification results at the vehicle end;

[0073] S3. Combine the road surface classification results at the vehicle end and the road surface classification results at the roadside end to obtain the fused road surface classification result;

[0074] S4. Perform spatio-temporal conversion on the fused classification result to obtain the current visual estimated value θ of the peak adhesion coefficient image ;

[0075] S5. Combine the visual estimated value θ image of the peak adhesion coefficient and the dynamic estimated value θ dynamics to obtain the fused estimated result θ fusion of the peak adhesion coefficient.

[0076] In this embodiment, the above method is applied to build an application framework as shown in Figure 2 and mainly includes the following five parts:

[0077] I. Use the roadside camera to obtain the road surface image of the specified section, and input the road surface image into the road surface classification model to obtain the road surface classification result at the roadside end. The specific process includes:

[0078] (1.1) Determine that the region of interest (ROI) for roadside unit image processing is the road surface area 20 - 90 m longitudinally and 7 m laterally in front of the camera;

[0079] (1.2) Obtain the bird's-eye view of the road surface image in the ROI through inverse perspective transformation;

[0080] (1.3) Segment the road surface image, and segment the road surface image into a 4×40 grid area, and the road surface area corresponding to each grid is a 1.75 m×1.75 m square;

[0081] (1.4) Input each small grid into the road surface classification network, and combine the position information corresponding to the grid to obtain the road surface type distribution result matrix A.

[0082] II. Use the in-vehicle camera to obtain the road surface image in front of the vehicle, and input the road surface image into the road surface classification model to obtain the road surface classification result at the vehicle end. The specific process includes:

[0083] (2.1) First, obtain historical road surface images of different types, including dry asphalt roads, wet asphalt roads, snow-covered roads, and concrete roads;

[0084] (2.2) Label the road surface types of the obtained historical road surface images;

[0085] (2.3) Use the labeled road surface images to train the deep convolutional neural network to obtain the road surface classification model;

[0086] (2.4) Determine the region of interest (ROI) as the area longitudinally 3 - 6.6 m in front of the vehicle and transversely -1.8 m to 1.8 m;

[0087] (2.5) Obtain the bird's-eye view of the road surface image in the ROI through inverse perspective transformation;

[0088] (2.6) Divide the ROI into a 3×3 grid, and input each grid into the classification network respectively to obtain the classification result of each grid, and further obtain the road surface classification result matrix X at the vehicle end.

[0089] III. According to the vehicle position, fuse the road surface classification result at the vehicle end and the road surface classification result at the roadside end to obtain the visual estimated value θ of the peak adhesion coefficient image , and its specific process includes:

[0090] (3.1) The roadside unit transmits the road surface type distribution result matrix A to the in-vehicle terminal through DSRC communication;

[0091] (3.2) Perform spatial synchronization. Obtain the vehicle position through the vehicle-mounted positioning system, determine the vehicle coordinate system and the roadside unit coordinate system respectively using the vehicle position and the position of the roadside unit, and determine the position (x V , y V ) of the vehicle in the roadside unit coordinate system, as shown in Figure 3 ;

[0092] (3.3) Calculate the position of the vehicle ROI in the roadside unit coordinate system, that is, the four-point coordinates of the ROI;

[0093] (3.4) Calculate the roadside unit road surface type result matrix Y in the vehicle ROI position area. Calculate the road surface type results of each small grid in the ROI respectively. As shown in Figure 4 , each small grid will cover 1 - 4 grid areas in the distribution map, corresponding to the road surface type results I1, I2, I3, I4 in the distribution map, as shown in the following formula, and further obtain the matrix Y;

[0094]

[0095] where, S v (iden) represents the classification result of the small grid, and s iv (iden) represents the area occupied by each road surface in the small grid.

[0096] (3.5) As shown in the following formula, calculate the road surface classification result matrix C of the vehicle ROI;

[0097]

[0098] Among them, matrices Y1 and Y2 are obtained by repeating step (3.4) after longitudinally moving the position of the ROI in the distribution map by ±0.1 m respectively.

[0099] (3.6) To simplify the recognition results of the visual fusion classifier, for the obtained final result matrix, each row votes according to the relative majority voting method to obtain the final classification result, as shown in the following formula, and finally a matrix D containing the fusion recognition results of the position information and the corresponding road surface type is obtained.

[0100]

[0101] Among them, H v (iden) represents the voting prediction result, and h iv (iden) represents the classification result of the single-grid visual fusion classifier.

[0102] IV. Perform spatio-temporal conversion on the fusion recognition results, and the specific process includes:

[0103] (4.1) Record the cumulative displacement s of the vehicle t ;

[0104] (4.2) The vehicle displacement corresponding to the current ROI is s t1 = s t + 3.6, s t2 = s t + 4.8, s t3 = s t + 6;

[0105] (4.3) Obtain a new road surface type matrix D', and the i-th row of D' is (s ti , I i ), where s ti is the vehicle displacement, and I i is the corresponding road surface classification result;

[0106] (4.4) When I m in matrix D' is different from I m-1 , record s tm and I m ;

[0107] (4.5) The final output result I of the road surface type is as follows: t as follows:

[0108]

[0109] (4.6) Obtain the visual estimated value θ of the peak adhesion coefficient according to the fusion recognition results of the road surface type image, where the visually estimated values of the peak adhesion coefficients for dry asphalt pavement, wet asphalt pavement, snow pavement, and concrete pavement are 0.8, 0.6, 0.4, and 0.8 respectively.

[0110] V. Combining the visually estimated value θ image and the kinematic estimated value θ dynamics , the fused estimated result θ fusion of the peak adhesion coefficient is obtained. The specific process includes:

[0111] (5.1) Establish the Burckhardt tire model as shown in the following equation:

[0112]

[0113] where μ(θ,λ) represents the utilized adhesion coefficient of the current tire to the ground, θ is the peak adhesion coefficient, λ is the slip ratio, c1 is the longitudinal slip stiffness of the tire, and c2, c3, c4 are the curve shape control parameters after the curve crosses the peak point.

[0114] (5.2) According to the current tire longitudinal force and slip ratio, the kinematic estimated value θ dynamics of the current peak adhesion coefficient is obtained.

[0115] F x = T m / R

[0116]

[0117] where T m is the wheel-end torque, R is the tire rolling radius, F x is the longitudinal force exerted on the tire, is the estimated value of the peak adhesion coefficient, θ(λ,F x ) is the peak adhesion coefficient obtained according to the longitudinal force and slip ratio using the formula in step (5.1), and γ is the kinematic estimator gain.

[0118] (5.3) Establish a fuzzy logic with the input being the difference |θ image -θ fusion | between the current fused estimator estimated value and the visually estimated value and the visually estimated value θ image , and the output being the kinematic estimator γ. In this embodiment, the domain of θ image is set to [0.15, 0.85], the domain of |θ image -θ fusion | is set to [0.1, 1.0], and the domain of γ is set to [3, 30]. Table 1 shows the fuzzy rule logic in this embodiment.

[0119] Table 1

[0120]

Claims

1. A multi-source fusion estimation method for road surface adhesion coefficient based on vehicle-road cooperation, characterized in that It includes the following steps: S1. Use the roadside camera to obtain the road surface image of the specified section, input the road surface image into the pre-constructed road surface classification model, and output the road surface classification result at the roadside; S2. The in-vehicle terminal obtains and recognizes the road surface image in front of the vehicle to obtain the road surface classification result at the vehicle end; S3. Fuse the road surface classification result at the vehicle end and the road surface classification result at the roadside to obtain the road surface fusion classification result; S4. Perform spatio-temporal conversion on the fused classification result to obtain the visual estimated value θ of the current peak adhesion coefficient image ; S5. Establish the Burckhardt tire model and calculate the kinetic estimated value θ of the peak adhesion coefficient dynamics , and fuse the visual estimated value θ image and the kinetic estimated value θ dynamics to obtain the fused estimated result θ of the peak adhesion coefficient fusion ; Step S5 specifically includes the following steps: S51. Establish the Burckhardt tire model: Among them, μ(θ,λ) is the utilization adhesion coefficient of the current tire to the ground, θ is the peak adhesion coefficient, λ is the slip ratio, c1 is the longitudinal slip stiffness of the tire, and c2, c3, c4 are the curve shape control parameters after the curve crosses the peak point; S52. Obtain a dynamic estimated value θ of the current peak adhesion coefficient based on the current longitudinal tire force and slip ratio dynamics : F x = T m / R Among them, T m is the wheel-end torque, R is the tire rolling radius, and F x is the longitudinal force acting on the tire, is the estimated value of the peak adhesion coefficient, and θ(λ, F x ) is the peak adhesion coefficient obtained based on the longitudinal force and the slip ratio, and γ is the gain of the dynamic estimator; S53. Establish a fuzzy logic with the input being the difference |θ| between the estimated value of the current fusion estimator and the visual estimated value image -θ fusion | and the visual estimated value θ image and the output being the dynamic estimator γ, from which the fusion estimation result θ of the peak adhesion coefficient is obtained fusion .

2. The multi-source fusion estimation method for road surface adhesion coefficient based on vehicle-road cooperation according to claim 1, wherein, In step S1, the road surface classification model is specifically constructed through the following process: Obtain historical road surface images of different types, including dry asphalt roads, wet asphalt roads, snow-covered roads, and concrete roads; Label the road surface types of the obtained historical road surface images; Use the labeled road surface images to train the deep convolutional neural network to obtain the road surface classification model.

3. A multi-source fusion estimation method for road surface adhesion coefficient based on vehicle-road cooperation according to claim 1, characterized in that, The specific process of outputting the road surface classification result at the roadside in step S1 is as follows: Determine the region of interest ROI for roadside unit image processing, which is a rectangular road surface area of a1 m×a2 m in front of the camera; Obtain the bird's-eye view of the road surface image in the ROI through inverse perspective transformation; Divide the road surface image into n×n grids; Input each small grid into the road surface classification network, and combine the position information corresponding to the grid to obtain the road surface type distribution result matrix A.

4. A multi-source fusion estimation method for road surface adhesion coefficient based on vehicle-road cooperation according to claim 3, characterized in that The specific process of obtaining the road surface classification result at the vehicle end in step S2 is as follows: Determine the ROI for in-vehicle terminal image processing, which is a square road surface area of b m×b m at a distance of l m in front of the vehicle; Obtain the bird's-eye view of the road surface image in the ROI through inverse perspective transformation; Divide the ROI into m×m grids, and input each grid into the classification network respectively to obtain the classification result of each grid, and then obtain the road surface classification result matrix X at the vehicle end.

5. A method for multi-source fusion estimation of road surface adhesion coefficient based on vehicle-road cooperation according to claim 4, characterized in that, Step S3 specifically includes the following steps: S31. The roadside unit transmits the road surface type distribution result matrix A to the in-vehicle terminal; S32. Perform spatial synchronization: Using the vehicle position and the position of the roadside unit, respectively determine the vehicle coordinate system and the roadside unit coordinate system, and determine the position of the vehicle in the roadside unit coordinate system (x V , y V ); S33. Calculate the position of the vehicle ROI in the roadside unit coordinate system, that is, the four-point coordinates of the ROI; S34. Calculate the roadside unit road surface type result matrix Y in the vehicle ROI position area, and calculate the road surface classification result matrix C of the vehicle ROI; S35. For the road surface classification result matrix C, vote for the final classification result according to the relative majority voting method for each row, and finally obtain a matrix D containing the fusion recognition result of the position information and the corresponding road surface type.

6. The multi-source fusion estimation method of road surface adhesion coefficient based on vehicle-road cooperation according to claim 5, characterized in that In step S32, the vehicle position is specifically obtained through the vehicle positioning system.

7. A multi-source fusion estimation method for road surface adhesion coefficient based on vehicle-road cooperation according to claim 5, characterized in that, The specific process of step S34 is as follows: S341. Calculate the road surface type results of each small grid in the ROI respectively. Each small grid will cover 1-4 grid areas in the distribution map, corresponding to the road surface type results I1, I2, I3, I4 in the distribution map, as shown in the following formula, and then obtain the matrix Y; Among them, S v (iden) represents the classification result of the small grid, and s iv (iden) represents the area occupied by each type of road surface in the small grid; S342. As shown in the following formula, calculate the road surface classification result matrix C of the vehicle ROI; Among them, matrices Y1 and Y2 are obtained by repeating step S341 after longitudinally moving the position of the ROI in the distribution map by ±x m.

8. A method for multi-source fusion estimation of road surface adhesion coefficient based on vehicle-road cooperation according to claim 7, characterized in that The specific elements of matrix D containing the position information and the fusion recognition result of the corresponding road surface type in step S35 are: Among them, H v (iden) is the voting prediction result, and h iv (iden) is the classification result of a single grid visual fusion classifier.

9. The multi-source fusion estimation method of road surface adhesion coefficient based on vehicle-road cooperation according to claim 8, characterized in that The specific process of step S4 is: Record the cumulative displacement s of the vehicle t ; The vehicle displacement corresponding to the current ROI is s t1 = s t + l + b / 3, s t2 = s t + l + 2b / 3, s t3 = s t + l + b; Obtain a new road surface type matrix D', where the i-th row of D' is (s ti , I i ), where s ti is the vehicle displacement and I i is the corresponding road surface classification result; When I in matrix D' m is different from I m-1 , record s tm and I m ; Final output result I of pavement type t As follows: According to the final output result of the road surface type, the visual estimated value θ of the peak adhesion coefficient is obtained image : The visual estimated values of the peak adhesion coefficients corresponding to dry asphalt pavement, wet asphalt pavement, snow pavement, and concrete pavement are 0.8, 0.6, 0.4, and 0.8 respectively.

Images