Bayesian estimation system and method for automotive wheel alignment parameters

Abstract

The application discloses a kind of automobile wheel positioning parameter bayesian estimation system and method, to solve automobile wheel positioning parameter bayesian estimation and need pre-calibration before detection problem.Automobile wheel positioning parameter bayesian estimation system mainly by camera (1), wheel target (2), target connection clamp (3), clamp (4) and tripod (5).Tribrach (5) is placed on horizontal ground, target connection clamp (3) is the section U-shaped cuboid of steel plate processing and two solid and 120-140 degree obtuse angle cylinder welding parts, automobile wheel positioning parameter bayesian estimation method is by automobile wheel positioning parameter bayesian estimation data acquisition, automobile wheel positioning parameter bayesian estimation training and test, automobile wheel positioning parameter bayesian estimation weight vector optimization etc.composition of step, provide a kind of structure simple, detection precision is high, need not pre-calibrated automobile wheel positioning parameter bayesian estimation system and method.

Description

Technical Field

[0001] This invention relates to an automotive testing device and testing method, and more specifically, to a Bayesian estimation system and method for automotive wheel alignment parameters. Background Technology

[0002] With the continuous development of the automotive industry, people are paying more and more attention to the field of automotive inspection and diagnosis. Good vehicle performance is paramount for safe driving. As one of the important automotive inspection targets, vehicle wheel alignment parameters have a significant impact on vehicle safety and stability. For example, when vehicle wheel alignment parameters are inaccurate, it can cause dangerous driving phenomena such as vehicle skidding and the steering wheel failing to return to center automatically. Therefore, accurate measurement and timely adjustment of vehicle wheel alignment parameters are crucial for safe driving. Currently, visual inspection methods are commonly used, utilizing images of a target rotating with the wheel, and then analyzing and calculating the changes in its feature points to obtain the corresponding alignment parameters. This method belongs to the maximum likelihood estimation method, but it has high computational costs, the accuracy of the results is affected by many factors, and it requires calibration of the camera's intrinsic and extrinsic parameters. To improve detection accuracy and reduce detection complexity, this invention designs a maximum a posteriori Bayesian estimation method for vehicle wheel alignment parameters. Summary of the Invention

[0003] This invention addresses the problems of complex equipment and cumbersome calibration in existing technologies for measuring automotive wheel alignment parameters, and the fact that most existing methods fall under the category of maximum likelihood methods, which suffer from overfitting and other errors. This invention proposes a Bayesian estimation method for automotive wheel alignment parameters, which belongs to the maximum a posteriori (MAP) method. MAP methods do not suffer from overfitting, thus reducing prediction errors and achieving high-precision measurement of automotive wheel alignment parameters. Furthermore, this Bayesian estimation method does not require pre-calibration of the camera's intrinsic and extrinsic parameters. The method mainly consists of an industrial camera, a wheel target composed of a regular circular pattern, a target connecting clamp, and a fixture connecting the wheel target and the wheel rim. The method trains the Bayesian estimation of automotive wheel alignment parameters by analyzing the image coordinates of the center point on the wheel target and the corresponding wheel alignment parameter values ​​on the automotive wheel alignment parameter calibration device. Then, the wheel target located at the test position is tested to obtain the automotive wheel alignment parameters.

[0004] Referring to the accompanying drawings, the present invention is implemented using the following technical solution:

[0005] The Bayesian estimation system for vehicle wheel alignment parameters includes a camera, wheel targets, target connecting clamps, fixtures, and tripods.

[0006] The tripod is placed on a level ground. The bottom of the camera is threadedly fixed to the top of the tripod. The wheel target is placed in the inner long groove of the target connecting clamp and fixed with the target connecting clamp by interference fit. The cylinder of the target connecting clamp is inserted into the circular through hole in the middle part of the clamp and is connected with the clamp by interference fit.

[0007] The wheel target described in the technical solution is a slender rectangular part made of steel plate, and the surface of the wheel target is covered with regular geometric patterns.

[0008] The target connecting clamp described in the technical solution is a part made by welding a U-shaped cuboid made of steel plate and two solid cylinders with an obtuse angle of 120-140 degrees.

[0009] The fixture described in the technical solution is a triangular part machined from a steel plate. The side of the fixture has three small cylinders with recesses on the outer side, and the middle part of the fixture has a circular through hole.

[0010] The specific steps of the Bayesian estimation method for automobile wheel alignment parameters are as follows:

[0011] Step 1: Data collection for Bayesian estimation of vehicle wheel alignment parameters:

[0012] A tripod is placed on a level surface, a camera is fixed to the top of the tripod, and a clamp is fixed to the rim of the car wheel to be inspected. The wheel target is then fixed to the target connector, and the wheel target and connector are fixed together to the clamp. The camera position is adjusted to capture images of the wheel target moving with the wheel. The camera acquires U images, where the wheel target contains V centers. The coordinate vector of the V centers in a single target image is represented as C. u =(x u1 y u1 x u2 y u2 , ..., x uV y uV ), where u = 1, 2, ..., U, v = 1, 2, ..., V, and record the corresponding wheel alignment parameter values ​​on the vehicle wheel alignment parameter calibration device for this U images, denoted as Θ = (Θ1, Θ2, ..., Θ). U Based on the coordinate vector of the circle center image and the corresponding wheel positioning parameter values, the Bayesian estimation of the vehicle wheel positioning parameters is determined.

[0013] Step 2: Training and testing of Bayesian estimation of vehicle wheel alignment parameters:

[0014] First, the prior probability of the weight vector h of the Bayesian estimation of the vehicle wheel alignment parameters is:

[0015] p(h|μ)=N(h|0,μ -1 I)

[0016] Where μ is a hyperparameter, and its likelihood function with respect to h follows a Gaussian distribution, i.e.

[0017]

[0018] Where ρ is a hyperparameter, and according to Bayes' theorem, the posterior distribution of the weight vector h is:

[0019] p(h|{C1,C2,...,C U}, Θ, μ, ρ)∝p(Θ|{C1, C2,..., C U},h,ρ)p(h|μ)

[0020] Input the target center image coordinate vector {C1, C2, ..., C} of the training set U} and its corresponding wheel alignment parameter values ​​Θ=(Θ1,Θ2,...,Θ U The target center image coordinate vector c during testing, and the joint probability distribution of the training and testing set data in the Bayesian estimation of vehicle wheel alignment parameters, are as follows:

[0021] p(θ,c,Θ,{C1,C2,...,C U})=∫p(θ|c, Θ, {C1, C2,...,C U},h)p(c,Θ,{C1,C2,...,C U},h)dh

[0022] Where p(c, Θ, {C1, C2, ..., C...}) U}, h) can be written as

[0023] p(c, Θ, {C1, C2, ..., C...} U},h)=p(h|c,Θ,{C1,C2,...,C U})p(c,Θ,{C1,C2,...,C U})

[0024] According to the multiplication rule, p(θ, c, Θ, {C1, C2, ..., C...}) U}) can be written as

[0025] p(θ,c,Θ,{C1,C2,...,C U})=p(θ|c, Θ, {C1, C2,...,C U})p(c,Θ,{C1,C2,...,C U})

[0026] From the above equations, we can obtain the wheel alignment parameters θ corresponding to the test image c output by the Bayesian estimation of the vehicle wheel alignment parameters.

[0027] p(θ|c, Θ, {C1, C2,...,C U})=∫p(θ|c, Θ, {C1, C2,...,C U},h)p(h|c,Θ,{C1,C2,...,C U})dh

[0028] The Gaussian distribution in the above equation has the following form:

[0029] p(θ|c, Θ, {C1, C2,...,C U})=N(θ|m(c), s 2 (c))

[0030] in, s 2 (c)=ρ -1 +c T Sc, m U =ρ(μI+ρc T c) -1 c T Θ, γ=(λ+μ) / λ, λ is the matrix ρc T eigenvalues ​​of c Then, the weight vector h of the Bayesian estimation of the vehicle wheel alignment parameters is solved using an iterative method.

[0031] Step 3: Optimization of the weight vector for Bayesian estimation of vehicle wheel alignment parameters:

[0032] By using the alternating direction multiplier method, the weight vector h obtained after Bayesian estimation of the vehicle wheel alignment parameters is further optimized, and a more accurate result is obtained when a new test image is input according to the following formula.

[0033] p(θ|c, Θ, {C1, C2,...,C U})=∫p(θ|c, Θ, {C1, C2,...,C U},h)p(h|c,Θ,{C1,C2,...,C U})dh

[0034] The beneficial effects of this invention are:

[0035] 1. This invention employs a Bayesian method to estimate vehicle wheel alignment parameters. First, the wheel target 2 is fixedly connected to the wheel rim using a target connector 3 and a clamp 4. Camera 1 captures images of the wheel target 2 moving with the wheel. A training set is constructed using the coordinate vectors of the center images on the target pattern and their corresponding wheel alignment parameter values. New target center image coordinate vectors used for testing are then input into the Bayesian estimation model for vehicle wheel alignment parameters, thus obtaining the corresponding vehicle wheel alignment parameters.

[0036] 2. This invention utilizes a Bayesian estimation method for vehicle wheel alignment parameters to measure the alignment parameters of vehicle wheels. This method falls under the maximum a posteriori (MAP) category, while most current measurement methods are maximum likelihood methods, which suffer from overfitting and other problems. The Bayesian estimation method for vehicle wheel alignment parameters can effectively avoid overfitting by using a training set to determine the Bayesian estimation model for vehicle wheel alignment parameters, reducing computational errors and improving the measurement accuracy of vehicle wheel alignment parameters.

[0037] 3. This invention utilizes the Bayesian estimation method for vehicle wheel alignment parameters. Theoretically, as long as a training set with the coordinate vector of the circle center image and its corresponding wheel alignment parameter values ​​and the input test image are constructed, the corresponding test results can be obtained. This testing process does not require pre-calibration of the camera's intrinsic and extrinsic parameters, which, unlike the traditional complex measurement process, reduces the error loss in the intermediate calculation process. Attached Figure Description

[0038] Figure 1 It is the overall axle projection of the Bayesian estimation system for automobile wheel alignment parameters;

[0039] Figure 2 This is the isometric view of camera 1 in the Bayesian estimation system for vehicle wheel alignment parameters;

[0040] Figure 3 It is the axonometric view of wheel target 2 in the Bayesian estimation system for automobile wheel alignment parameters;

[0041] Figure 4 This is the axonometric view of target connector 3 in the Bayesian estimation system for automobile wheel alignment parameters;

[0042] Figure 5 This is the axonometric view of fixture 4 in the Bayesian estimation system for automobile wheel alignment parameters;

[0043] Figure 6 It is the axonometric view of tripod 5 in the Bayesian estimation system for automobile wheel alignment parameters;

[0044] Figure 7 This is a flowchart of the Bayesian estimation method for automobile wheel alignment parameters;

[0045] In the picture: 1. Camera, 2. Wheel target, 3. Target connecting clamp, 4. Fixture, 5. Tripod. Detailed Implementation

[0046] The present invention will now be described in further detail with reference to the accompanying drawings:

[0047] See Figures 1 to 6 The Bayesian estimation system for vehicle wheel alignment parameters includes a camera 1, a wheel target 2, a target connecting clamp 3, a fixture 4, and a tripod 5.

[0048] Tripod 5 is placed on a horizontal surface. The bottom of camera 1 is threadedly fixed to the top of tripod 5. Wheel target 2 is a slender rectangular part made of steel plate. The surface of wheel target 2 is covered with regular geometric patterns. Target connecting clamp 3 is a part made of steel plate with a U-shaped cross section and two solid cylinders with an obtuse angle of 120-140 degrees welded together. Wheel target 2 is placed in the inner long groove of target connecting clamp 3 and fixed with target connecting clamp 3 with interference fit. Fixture 4 is a triangular part made of steel plate. The side of fixture 4 is machined with three small cylinders with pits on the outside. The middle part of fixture 4 is machined with a circular through hole. The cylinder of target connecting clamp 3 is inserted into the circular through hole in the middle part of fixture 4 and connected with fixture 4 with interference fit.

[0049] See Figure 7 The specific process of the Bayesian estimation method for automobile wheel alignment parameters provided by this invention is as follows:

[0050] Step 1: Data collection for Bayesian estimation of vehicle wheel alignment parameters:

[0051] Tripod 5 is placed on a horizontal surface. Camera 1 is fixed to the top of tripod 5. Clamp 4 is fixed to the rim of the car wheel to be inspected. Then, wheel target 2 is fixed to target connecting clamp 3, and wheel target 2 and target connecting clamp 3 are fixed together to clamp 4. The position of camera 1 is adjusted so that it can capture images of wheel target 2 moving with the wheel. Camera 1 acquires U images. The wheel target 2 in the image contains V centers. The coordinate vector of the V centers in a target image is represented as C. u =(x u1 y u1 x u2 y u2 , ..., x uV y uV ), where u = 1, 2, ..., U, v = 1, 2, ..., V, and record the corresponding wheel alignment parameter values ​​on the vehicle wheel alignment parameter calibration device for this U images, denoted as Θ = (Θ1, Θ2, ..., Θ). UBased on the coordinate vector of the circle center image and the corresponding wheel positioning parameter values, the Bayesian estimation of the vehicle wheel positioning parameters is determined.

[0052] Step 2: Training and testing of Bayesian estimation of vehicle wheel alignment parameters:

[0053] First, the prior probability of the weight vector h of the Bayesian estimation of the vehicle wheel alignment parameters is:

[0054] p(h|μ)=N(h|0,μ -1 I)

[0055] Where μ is a hyperparameter, and its likelihood function with respect to h follows a Gaussian distribution, i.e.

[0056]

[0057] Where ρ is a hyperparameter, and according to Bayes' theorem, the posterior distribution of the weight vector h is:

[0058] p(h|{C1,C2,...,C U}, Θ, μ, ρ)∝p(Θ|{C1, C2,..., C U},h,ρ)p(h|μ)

[0059] Input the target center image coordinate vector {C1, C2, ..., C} of the training set U} and its corresponding wheel alignment parameter values ​​Θ=(Θ1,Θ2,...,Θ U The target center image coordinate vector c during testing, and the joint probability distribution of the training and testing set data in the Bayesian estimation of vehicle wheel alignment parameters, are as follows:

[0060] p(θ,c,Θ,{C1,C2,...,C U})=∫p(θ|c, Θ, {C1, C2,...,C U},h)p(c,Θ,{C1,C2,...,C U},h)dh

[0061] Where p(c, Θ, {C1, C2, ..., C...}) U}, h) can be written as

[0062] p(c, Θ, {C1, C2, ..., C...} U},h)=p(h|c,Θ,{C1,C2,...,C U})p(c,Θ,{C1,C2,...,C U})

[0063] According to the multiplication rule, p(θ, c, Θ, {C1, C2, ..., C...})U}) can be written as

[0064] p(θ,c,Θ,{C1,C2,...,C U})=p(θ|c, Θ, {C1, C2,...,C U})p(c,Θ,{C1,C2,...,C U})

[0065] From the above equations, we can obtain the wheel alignment parameters θ corresponding to the test image c output by the Bayesian estimation of the vehicle wheel alignment parameters.

[0066] p(θ|c, Θ, {C1, C2,...,C U})=∫p(θ|c, Θ, {C1, C2,...,C U},h)p(h|c,Θ,{C1,C2,...,C U})dh

[0067] The Gaussian distribution in the above equation has the following form:

[0068] p(θ|c, Θ, {C1, C2,...,C U})=N(θ| m(c), s 2 (c))

[0069] in, s 2 (c)=ρ -1 +c T Sc, m U =ρ(μI+ρc T c) -1 c T Θ, γ=(λ+μ) / λ, λ is the matrix ρc T eigenvalues ​​of c Then, the weight vector h of the Bayesian estimation of the vehicle wheel alignment parameters is solved using an iterative method.

[0070] Step 3: Optimization of the weight vector for Bayesian estimation of vehicle wheel alignment parameters:

[0071] By using the alternating direction multiplier method, the weight vector h obtained after Bayesian estimation of the vehicle wheel alignment parameters is further optimized, and a more accurate result is obtained when a new test image is input according to the following formula.

[0072] p(θ|c, Θ, {C1, C2,...,C U})=∫p(θ|c, Θ, {C1, C2,...,C U},h)p(h|c,Θ,{C1,C2,...,CU})dh.

Claims

1. A Bayesian estimation method for automobile wheel alignment parameters, characterized in that, The Bayesian estimation method for the vehicle wheel positioning parameters includes a camera (1), a wheel target (2), a target connecting clamp (3), a fixture (4), and a tripod (5). The tripod (5) is placed on a horizontal ground. The bottom of the camera (1) is threadedly fixed to the top of the tripod (5). The wheel target (2) is placed in the inner long groove of the target connecting clamp (3) and fixed with the target connecting clamp (3) by interference fit. The cylindrical part of the target connecting clamp (3) is inserted into the circular through hole in the middle part of the clamp (4) and connected with the clamp (4) by interference fit. The specific steps are as follows: Step 1: Data collection for Bayesian estimation of vehicle wheel alignment parameters: The tripod (5) is placed on a horizontal surface. The camera (1) is fixed to the top of the tripod (5). The clamp (4) is fixed to the rim of the car wheel to be inspected. Then, the wheel target (2) is fixed to the target connecting clamp (3), and the wheel target (2) and the target connecting clamp (3) are fixed together to the clamp (4). The position of the camera (1) is adjusted so that it can capture an image of the wheel target (2) moving with the wheel. The camera (1) acquires U images. The wheel target (2) in the image contains V centers. The coordinate vector of the V centers in a target image is represented as follows: ,in And record the corresponding wheel alignment parameter values ​​on the vehicle wheel alignment parameter calibration device for this U-shaped image, denoted as: Based on the coordinate vector of the circle center image and the corresponding wheel positioning parameter values, the Bayesian estimation of the vehicle wheel positioning parameters is determined. Step 2: Training and testing of Bayesian estimation of vehicle wheel alignment parameters: First, the prior probability of the weight vector h of the Bayesian estimation of the vehicle wheel alignment parameters is: ， Where μ is a hyperparameter, and its likelihood function with respect to h follows a Gaussian distribution, i.e.: ， Where ρ is a hyperparameter, and according to Bayes' theorem, the posterior distribution of the weight vector h is: ， Input the target center image coordinate vector {C1, C2, …, C} of the training set U } and its corresponding wheel alignment parameter values ​​Θ=(Θ1, Θ2, …, Θ U Given the target center image coordinate vector c during testing, the joint probability distribution of the training and testing set data in the Bayesian estimation of vehicle wheel positioning parameters is as follows: ， Where p(c, Θ, {C1, C2, …, C U }, h) can be written as: ， According to the multiplication rule, p(θ, c, Θ, {C1, C2, …, C U }) can be written as: ， From the above equations, we can obtain the wheel alignment parameters θ corresponding to the test image c output by the Bayesian estimation of the vehicle wheel alignment parameters as follows: ， The Gaussian distribution in the above equation has the following form: ， in, , , , , , γ=(λ+μ) / λ, λ is the matrix ρc T Eigenvalues ​​of c Then, the weight vector h of the Bayesian estimation of the vehicle wheel alignment parameters is solved using an iterative method; Step 3: Optimization of the weight vector for Bayesian estimation of vehicle wheel alignment parameters: By using the alternating direction multiplier method, the weight vector h obtained after Bayesian estimation of the vehicle wheel alignment parameters is further optimized. When a new test image is input, a more accurate result is obtained according to the following formula. 。 2. The Bayesian estimation method for automobile wheel alignment parameters according to claim 1, characterized in that... The wheel target (2) is a slender rectangular part made of steel plate, and the surface of the wheel target (2) is covered with regular geometric patterns.

3. The Bayesian estimation method for automobile wheel alignment parameters according to claim 1, characterized in that... The target connecting clip (3) is a part made of a U-shaped cuboid made of steel plate and two solid cylinders with an obtuse angle of 120-140 degrees welded together.

4. The Bayesian estimation method for automobile wheel alignment parameters according to claim 1, characterized in that... The fixture (4) is a triangular part machined from steel plate. The side of the fixture (4) is machined with three small cylinders with pits on the outside, and the middle part of the fixture (4) is machined with a circular through hole.

Images