Vehicle body height difference detection method and system based on vehicle body point cloud features

By using a detection method based on the point cloud features of the car body, and employing three depth cameras and advanced algorithms, non-contact and automated detection of the left and right height differences of the car body was achieved. This solved the problems of low detection accuracy and component damage, and improved detection accuracy and efficiency.

CN117870558BActive Publication Date: 2026-11-10JILIN UNIVERSITY
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410060380.1
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-01-12
Publication Date
2026-11-10
Estimated Expiration
2044-01-12

AI Technical Summary

Technical Problem

Existing methods for detecting the left-right height difference of automobile bodies suffer from low accuracy, slow speed, and the potential for damage to components due to contact-based measurement.

Method used

A detection method based on the point cloud features of the car body is adopted. Point cloud data is collected using three depth cameras. Through techniques such as k-nearest neighbor algorithm, Gaussian filtering, DBSCAN clustering and Bayesian classification, non-contact and automated identification and height difference calculation of car body parts are achieved.

Benefits of technology

It achieves high-precision and rapid detection of the left and right height difference of the car body, improving detection accuracy and efficiency, and avoiding human error and damage to parts.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN117870558B_ABST
    Figure CN117870558B_ABST
Patent Text Reader

Abstract

The application discloses a kind of based on automobile body point cloud feature's body height difference detection method and system, to solve the problem of based on automobile body point cloud feature's body height difference detection.The body height difference detection method based on automobile body point cloud feature mainly includes the complete body point cloud of the vehicle (1) to be detected and detection datum plane point cloud, the different components of the complete body point cloud collected are extracted, the left and right symmetrical parts of automobile point cloud are identified and classified, and the height difference of the left and right symmetrical parts of automobile body is calculated, etc.Based on automobile body point cloud feature's body height difference detection system is mainly composed of first depth camera (2), second depth camera (3), third depth camera (4) and the like.A kind of based on automobile body point cloud feature's body height difference detection method and system can be used for large-scale body height difference measurement, easy to solve, and the performance of stable based on automobile body point cloud feature's body height difference detection method and system.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to a method and apparatus for automobile inspection, and more specifically, to a method and system for detecting the height difference of a vehicle body based on point cloud features of the vehicle body. Background Technology

[0002] Vehicle body height difference detection is a crucial aspect of automotive inspection. Ensuring the flatness and symmetry of the vehicle body is essential for overall quality and safety during vehicle use. Therefore, accurate detection and measurement of the height difference between symmetrical components, such as headlights and mirrors, is critical. Traditional manual inspection methods suffer from low accuracy and speed, while mechanical contact measurement methods can easily damage the surfaces of automotive parts. With the rapid development of 3D point cloud acquisition technology, point cloud data has become a vital method for non-contact measurement of vehicle shape and structure. However, simply acquiring vehicle point cloud data is insufficient to obtain detailed information on the left-right height difference of the vehicle body. Therefore, further analysis and processing of complex vehicle point cloud data are necessary to extract the point cloud feature information of components related to the body height difference. To fully utilize the viewpoint feature information in the acquired vehicle point cloud data and achieve non-contact, high-precision detection of the left-right height difference of the vehicle body, a method and system for vehicle body height difference detection based on vehicle body point cloud features were designed. Summary of the Invention

[0003] This invention addresses the problems of low detection accuracy, slow detection speed, and potential scratches on car body parts caused by contact measurements in existing automotive inspection methods for detecting left-right height differences. It proposes a method and system for detecting left-right height differences based on car body point cloud features, achieving non-contact, high-precision, and fully automated detection. The system mainly consists of three depth cameras and three camera mounts. Depth cameras placed on the left and right sides of the detection line and directly in front of the detection line acquire car body point clouds. The point cloud data is then processed to obtain highly accurate results for detecting the left-right height difference of the car body.

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

[0005] The specific steps of the vehicle height difference detection method based on vehicle body point cloud features are as follows:

[0006] Step 1: Collect the complete vehicle body point cloud and the inspection reference surface point cloud of the vehicle to be inspected:

[0007] The first camera mount is placed directly in front of the inspection line, while the second and third camera mounts are placed opposite each other on either side of the inspection line. The first depth camera is fixed to the first camera mount, the second depth camera to the second camera mount, and the third depth camera to the third camera mount. The first, second, and third depth cameras are activated to begin acquiring the complete vehicle body point cloud and the point cloud of the horizontal plane containing the inspection line. Upon receiving the start signal, the driver drives the vehicle at a constant speed of 2–4 km / h from the starting point of the inspection line. After the vehicle passes the inspection line, the inspection system obtains the complete vehicle body point cloud T and the point cloud H of the horizontal plane containing the inspection line. The plane π containing the point cloud H of the horizontal plane containing the inspection line is then obtained using the least squares method. H , π H This is the detection reference surface;

[0008] Step 2: Extract features from different components of the complete vehicle body point cloud obtained from the data collection.

[0009] Point X on the complete vehicle body point cloud T obtained in the first step i The k-nearest neighbor algorithm is used to select X. i The neighborhood set N centered on the center i This includes X i Calculate point X by considering itself and its k nearest neighbors. i With the neighborhood point set N i Each point X j The distance d between ij for

[0010]

[0011] Among them, (x i y i , z i () is point X i The coordinates, (x j y j , z j () is point X j The coordinates;

[0012] Based on point X on the vehicle body point cloud T i Neighborhood point set N i Each point X in j The distance d obtained from the above formula ij The point Xi after noise reduction is obtained using a Gaussian filtering algorithm based on statistical analysis. for

[0013]

[0014] in, It is point X i With the neighborhood point set N i Each point X j The weights between them, σ is the covariance parameter of the Gaussian filtering algorithm, and the points after noise reduction. The filtered point cloud T of the vehicle body of the vehicle under test C ;

[0015] The filtered vehicle body point cloud T obtained from the above formula C Each point in Use its k nearest neighbors The surface A containing it is fitted using the least squares method.

[0016]

[0017] Using the SVD method to perform eigenvalue decomposition on surface A, we obtain eigenvalues ​​λ1≤λ2≤λ3. The eigenvector corresponding to the smallest eigenvalue λ1 is the eigenvector of surface A at point [0, 1]. normal vector n i =(n ix n iy n iz )

[0018] The vehicle point cloud T obtained by the above formula C Each point in normal vector n i As feature vectors, the DBSCAN clustering algorithm is used to analyze the vehicle body point cloud T. C Divide into M classes, and for each cluster part M i Select each point normal vector n i x component n ix As viewpoint features, the number of intervals Q in the discretized feature space is adaptively determined based on the distribution of the point cloud data, and each point is used as a viewpoint feature. normal vector n i x component n ix The allocation is based on the discretized interval, and the points are determined using the following function. The range of belonging is

[0019]

[0020] Where q1, q2, ..., q Q The interval boundary value is determined after the interval number Q is determined, based on n. ix Given a defined range and density distribution of the data, after all points have been assigned to specific intervals, the number of points m within each discrete interval is counted. q The value of the feature description map H(M) for each cluster part is determined according to the following formula. i )

[0021]

[0022] Where, N T Dot the car body with cloud T C The total number of points in the cluster, for each cluster part M i H(M) obtained using the above formula i ) serves as a feature description for its corresponding part;

[0023] Step 3: Identify and classify the symmetrical parts in the car point cloud:

[0024] The feature description H(M) obtained in the second step i If the following likelihood function is input to the classification model, then

[0025]

[0026] Where the target vector t is the category label L i Binary encoding, category label L i Each cluster represents a component that is symmetrical on the left or right side. φ is the basic function, Φ is the basic function matrix, w is the weight vector of the model, and β is the error precision.

[0027] The prior probability distribution of weight w is defined using a Gaussian distribution as follows:

[0028]

[0029] Where Ω is the covariance matrix of the weights;

[0030] The posterior probability distribution of the weight w is calculated using Bayes' theorem.

[0031] p(w|Φ,t,α,β)∝p(t|Φ,w,β)p(w|Ω)

[0032] Where α is a hyperparameter;

[0033] The posterior probability of weight w is updated iteratively using the following formula.

[0034] Ω=αI+βΦΦT

[0035] v=βΩ -1 Φ T t

[0036] Where I is a unit vector and v is the posterior mean vector of the weights;

[0037] Using the posterior probability distribution of the weights w obtained above, the posterior mean vector v of the weights, the basis functions φ and Φ, and the feature description H(M) obtained in the second step... iThe posterior probability distribution of the target vector t is calculated using the covariance matrix Ω of the weights, the hyperparameter α, and the error precision β.

[0038] p(t|φ(H(M i )),Φ,v,Ω)∝p(t|Φ,w,β)p(w|Φ,t,α,β)

[0039] Based on the above formula, the posterior probability of the target vector t is obtained. The category label with the highest posterior probability is selected, and the part point cloud is classified into the corresponding part type L according to the following formula. i

[0040] L i =arg max p(L i |φ(H(M i )), Φ, v, Ω)

[0041] Step 4: Calculate the height difference between the symmetrical parts of the car body:

[0042] Based on the part type L corresponding to the part point cloud obtained in step three i Establish a left-side point cloud set Right-side point cloud set The reference plane π obtained in the first step H Based on the baseline, the average height H of the symmetrical point clouds on the left and right sides left H right for

[0043]

[0044]

[0045] Where, N left and N right These represent the point cloud sets on the left. And the number of points in the point cloud set on the right y i and y j These are the point cloud sets on the left. and the point cloud set on the right The coordinates of each point in the plane perpendicular to the reference plane;

[0046] The average height H on the left side obtained from the above formula left and the average height H on the right side right The difference in left and right height D of the vehicle under inspection is obtained as follows:

[0047] D = H left -H right

[0048] Where D is a positive value, it means the left side of the vehicle is higher, and a negative value means the right side of the vehicle is higher.

[0049] The vehicle height difference detection system based on vehicle body point cloud features includes a first depth camera, a second depth camera, a third depth camera, a first camera bracket, a second camera bracket, and a third camera bracket.

[0050] The first camera bracket is placed at the shooting position directly in front of the detection line, and the second and third camera brackets are placed at shooting positions on both sides of the detection line. The first, second, and third depth cameras are fixedly connected to the bolts of the first, second, and third camera brackets respectively through the threaded holes at the bottom.

[0051] The beneficial effects of this invention are:

[0052] 1. The vehicle body height difference detection method and system based on vehicle body point cloud features designed in this invention utilizes viewpoint features and other information from vehicle body point cloud data to accurately identify and classify components on the left and right symmetrical parts of the vehicle body. By accurately measuring the height difference between these vehicle body components and a reference plane, accurate measurement of the left-right height difference of the vehicle body can be achieved, improving the accuracy of vehicle inspection.

[0053] 2. Compared with traditional visual inspection and mechanical measurement methods, the vehicle height difference detection method and system based on vehicle body point cloud features designed in this invention fully utilizes point cloud technology to quickly acquire three-dimensional vehicle data, and rapidly locates and measures symmetrical parts of the vehicle body through automated recognition and classification methods. This improves the efficiency of vehicle height difference detection and saves time and labor costs.

[0054] 3. The vehicle body height difference detection method and system designed in this invention based on vehicle body point cloud features are not affected by the subjective factors of the inspector, providing more objective and reliable measurement results. Through the calculation of the reference plane and height difference, accurate numerical results can be obtained, providing objective conclusions for vehicle inspection. This ensures the symmetry of the vehicle body and the safety of vehicle operation, and has beneficial effects on the widespread application of point cloud technology in the field of vehicle inspection and the accurate measurement of vehicle dimensional parameters. Attached Figure Description

[0055] Figure 1 This is a flowchart of a method for detecting vehicle height difference based on point cloud features of the vehicle body;

[0056] Figure 2 This is an isometric view of a vehicle height difference detection system based on vehicle body point cloud features;

[0057] Figure 3 This is an isometric view of the first depth camera 2 in a vehicle height difference detection system based on vehicle body point cloud features;

[0058] Figure 4 This is an isometric view of the first camera bracket 5 in a vehicle height difference detection system based on vehicle body point cloud features;

[0059] In the image: 1. Vehicle to be inspected, 2. First depth camera, 3. Second depth camera, 4. Third depth camera, 5. First camera bracket, 6. Second camera bracket, 7. Third camera bracket. Detailed Implementation

[0060] The present invention will now be described in further detail with reference to the accompanying drawings:

[0061] See Figure 1 The method for detecting vehicle height difference based on vehicle body point cloud features can be divided into the following four steps:

[0062] Step 1: Collect the complete body point cloud and the detection reference surface point cloud of vehicle 1 to be inspected:

[0063] The first camera bracket 5 is placed in front of the detection line, and the second camera bracket 6 and the third camera bracket 7 are placed opposite each other on either side of the detection line. The first depth camera 2 is fixed to the first camera bracket 5, the second depth camera 3 is fixed to the second camera bracket 6, and the third depth camera 4 is fixed to the third camera bracket 7. The first depth camera 2, the second depth camera 3, and the third depth camera 4 are turned on to start acquiring the complete body point cloud of the vehicle under inspection 1 and the point cloud of the horizontal plane where the detection line is located. After receiving the start signal, the driver drives the vehicle under inspection 1 from the starting point of the detection line at a constant speed of 2-4 km / h across the detection line. After the vehicle under inspection 1 passes the detection line, the detection system obtains the complete body point cloud T of the vehicle under inspection 1 and the point cloud H of the horizontal plane where the detection line is located. The plane π of the point cloud H of the horizontal plane where the detection line is located is obtained by fitting using the least squares method. H , π H This is the detection reference surface;

[0064] Step 2: Extract features from different components of the complete vehicle body point cloud obtained from the data collection.

[0065] For the complete point cloud T of the vehicle body of the vehicle to be inspected obtained in the first step, point X i The k-nearest neighbor algorithm is used to select X. i The neighborhood set N centered on the center i This includes X i Calculate point X by considering itself and its k nearest neighbors. i With the neighborhood point set N i Each point X j The distance d between ij for

[0066]

[0067] Among them, (x i y i , z i () is point X i The coordinates, (x j y j , z j () is point X j The coordinates;

[0068] Based on point X on the vehicle body point cloud T i Neighborhood point set N i Each point X j The distance d obtained from the above formula ij Point X is obtained using a Gaussian filtering algorithm based on statistical analysis. i Noise reduction points for

[0069]

[0070] in, It is point X i With the neighborhood point set N i Each point X j The weights between them, σ is the covariance parameter of the Gaussian filtering algorithm, and the points after noise reduction. The filtered point cloud TC of the vehicle body of the vehicle under test 1 is formed.

[0071] The filtered vehicle body point cloud T obtained from the above formula C Each point in Use its k nearest neighbors The surface A containing it is fitted using the least squares method.

[0072]

[0073] Using the SVD method to perform eigenvalue decomposition on surface A, we obtain eigenvalues ​​λ1≤λ2≤λ3. The eigenvector corresponding to the smallest eigenvalue λ1 is the eigenvector of surface A at point [0, 1]. normal vector n i =(n ix n iy n iz )

[0074] The vehicle point cloud T obtained by the above formula C Each point in normal vector n i As feature vectors, the DBSCAN clustering algorithm is used to analyze the vehicle body point cloud T. C Divide into M classes, and for each cluster part M i Select each point normal vector ni x component n ix As viewpoint features, the number of intervals Q in the discretized feature space is adaptively determined based on the distribution of the point cloud data, and each point is used as a viewpoint feature. normal vector n i x component n ix The allocation is based on the discretized interval, and the points are determined using the following function. The range of belonging is

[0075]

[0076] Where q1, q2, ..., q Q The interval boundary value is determined after the interval number Q is determined, based on n. ix Given a defined range and density distribution of the data, after all points have been assigned to specific intervals, the number of points m within each discrete interval is counted. q The value of the feature description map H(M) for each cluster part is determined according to the following formula. i )

[0077]

[0078] Where, N T Dot the car body with cloud T C The total number of points in the cluster, for each cluster part M i H(M) obtained using the above formula i ) serves as a feature description for its corresponding part;

[0079] Step 3: Identify and classify the symmetrical parts in the car point cloud:

[0080] The feature description H(M) obtained in the second step i If the following likelihood function is input to the classification model, then

[0081]

[0082] Where the target vector t is the category label L i Binary encoding, category label L i Each cluster represents a component that is symmetrical on the left or right side. φ is the basic function, Φ is the basic function matrix, w is the weight vector of the model, and β is the error precision.

[0083] The prior probability distribution of weight w is defined using a Gaussian distribution as follows:

[0084]

[0085] Where Ω is the covariance matrix of the weights;

[0086] The posterior probability distribution of the weight w is calculated using Bayes' theorem.

[0087] p(w|Φ,t,α,β)∝p(t|Φ,w,β)p(w|Ω)

[0088] Where α is a hyperparameter;

[0089] The posterior probability of weight w is updated iteratively using the following formula.

[0090] Ω=αI+βΦΦT

[0091] v=βΩ -1 Φ T t

[0092] Where I is a unit vector and v is the posterior mean vector of the weights;

[0093] Using the posterior probability distribution of the weights w obtained above, the posterior mean vector v of the weights, the basis functions φ and Φ, and the feature description H(M) obtained in the second step... i The posterior probability distribution of the target vector t is calculated using the covariance matrix Ω of the weights, the hyperparameter α, and the error precision β.

[0094] p(t|φ(H(M i )),Φ,v,Ω)∝p(t|Φ,w,β)p(w|Φ,t,α,β)

[0095] Based on the above formula, the posterior probability of the target vector t is obtained. The category label with the highest posterior probability is selected, and the part point cloud is classified into the corresponding part type L according to the following formula. i

[0096] L i =arg max p(L i |φ(H(M i )), Φ, v, Ω)

[0097] Step 4: Calculate the height difference between the symmetrical parts of the car body:

[0098] Based on the part type L corresponding to the part point cloud obtained in step three i Establish a left-side point cloud set Right-side point cloud set The reference plane π obtained in the first step H Based on the baseline, the average height H of the symmetrical point clouds on the left and right sides lef t、H right for

[0099]

[0100]

[0101] Where, N left and N right These represent the point cloud sets on the left. And the number of points in the point cloud set on the right y i and y j These are the point cloud sets on the left. and the point cloud set on the right The coordinates of each point in the plane perpendicular to the reference plane;

[0102] The average height H on the left side obtained from the above formula left and the average height H on the right side right The left-right height difference D of the vehicle under inspection 1 is obtained as follows:

[0103] D = H left -H right

[0104] Where D is a positive value, it means the left side of the vehicle is higher, and a negative value means the right side of the vehicle is higher.

[0105] See Figures 2 to 4 The vehicle height difference detection system based on the point cloud features of the vehicle body includes a first depth camera 2, a second depth camera 3, a third depth camera 4, a first camera bracket 5, a second camera bracket 6, and a third camera bracket 7.

[0106] The first camera bracket 5 is placed at the shooting position directly in front of the detection line, and the second camera bracket 6 and the third camera bracket 7 are placed at shooting positions on both sides of the detection line. The first depth camera 2, the second depth camera 3 and the third depth camera 4 are fixedly connected to the first camera bracket 5, the second camera bracket 6 and the third camera bracket 7 respectively through the threaded holes at the bottom.

Claims

1. A method for detecting vehicle height difference based on vehicle body point cloud features, characterized in that, The specific steps of the vehicle body height difference detection method based on vehicle body point cloud features are as follows: Step 1: Collect the complete body point cloud and the detection reference surface point cloud of the vehicle to be inspected (1): The first camera bracket (5) is placed at the shooting position directly in front of the detection line. The second camera bracket (6) and the third camera bracket (7) are placed at shooting positions on both sides of the detection line. The first depth camera (2) is fixed on the first camera bracket (5), the second depth camera (3) is fixed on the second camera bracket (6), and the third depth camera (4) is fixed on the third camera bracket (7). The first depth camera (2), the second depth camera (3), and the third depth camera (4) are turned on to start collecting the complete body point cloud of the vehicle to be inspected (1) and the point cloud of the horizontal plane where the detection line is located. After receiving the start signal, the driver drives the vehicle to be inspected (1) from the starting point of the detection line at a constant speed of 2~4km / h through the detection line. After the vehicle to be inspected (1) passes through the detection line, the detection system obtains the complete body point cloud T of the vehicle to be inspected (1) and the point cloud H of the horizontal plane where the detection line is located. The plane π where the point cloud H of the horizontal plane where the detection line is located is obtained by fitting using the least squares method. H , π H This is the detection reference surface; Step 2: Extract features from different components of the complete vehicle body point cloud obtained from the data collection. The point X on the complete vehicle body point cloud T of the vehicle to be inspected (1) obtained in the first step. i The k-nearest neighbor algorithm is used to select X. i The neighborhood set N centered on the center i This includes X i Calculate point X by considering itself and its k nearest neighbors. i With the neighborhood point set N i Each point X in j The distance d between ij for Among them, (x i ,y i ,z i () is point X i The coordinates, (x j ,y j ,z j () is point X j The coordinates; Based on point X on the vehicle body point cloud T i Neighborhood point set N i Each point X in j The distance d obtained from the above formula ij The point X is obtained using a Gaussian filtering algorithm based on statistical analysis. i Noise reduction points for in, It is point X i With the neighborhood point set N i Each point X in j The weights between them, where σ is the covariance parameter of the Gaussian filtering algorithm, and the points after noise reduction. The filtered point cloud T of the vehicle body of the vehicle under test (1) is formed. C ; The filtered vehicle body point cloud T obtained from the above formula C Each point in Use its k nearest neighbors The surface A containing it is fitted using the least squares method. Using the SVD method to perform eigenvalue decomposition on surface A, we obtain eigenvalues ​​λ1≤λ2≤λ3. The eigenvector corresponding to the smallest eigenvalue λ1 is the eigenvector of surface A at point [0, 1]. normal vector n i =(n ix ,n iy ,n iz ) The vehicle point cloud T obtained by the above formula C Each point in normal vector n i As feature vectors, the DBSCAN clustering algorithm is used to analyze the vehicle body point cloud T. C Divide into M classes, and for each cluster part M i Select each point normal vector n i x component n ix As viewpoint features, the number of intervals Q in the discretized feature space is adaptively determined based on the distribution of the point cloud data, and each point is used as a viewpoint feature. normal vector n i x component n ix The points are determined by assigning based on the discretized intervals and using the following function. The range of belonging is Where q1,q2,…,q Q The interval boundary value is determined after the interval number Q is determined, based on n. ix Given a defined range and density distribution of the data, after all points have been assigned to specific intervals, the number of points m within each discrete interval is counted. q The value of the feature description map H(M) for each cluster part is determined according to the following formula. i ) Where, N T Add cloud-shaped dots to the car body C The total number of points in the cluster, for each cluster part M i H(M) obtained using the above formula i ) serves as a feature description for its corresponding part; Step 3: Identify and classify the symmetrical parts in the car point cloud: The feature description H(M) obtained in the second step i If the following likelihood function is input to the classification model, then Where the target vector t is the category label L i Binary encoding, category label L i This indicates that each cluster is a type of component that is symmetrical about the left or right side. Basic functions, is the basic function matrix, w is the model's weight vector, and β is the error precision; The prior probability distribution of weight w is defined using a Gaussian distribution as follows: Where Ω is the covariance matrix of the weights; The posterior probability distribution of the weight w is calculated using Bayes' theorem. Where α is a hyperparameter; The posterior probability of weight w is updated iteratively using the following formula. Where I is a unit vector and v is the posterior mean vector of the weights; Using the above calculations, the posterior probability distribution of the weight w, the posterior mean vector v of the weights, and the basis function... Fundamental function matrix The second step yields the feature description H(M). i The posterior probability distribution of the target vector t is calculated using the covariance matrix Ω of the weights, the hyperparameter α, and the error precision β. Based on the above formula, the posterior probability of the target vector t is obtained. The category label with the highest posterior probability is selected, and the part point cloud is classified into the corresponding part type L according to the following formula. i Step 4: Calculate the height difference between the symmetrical parts of the car body: Based on the part type L corresponding to the part point cloud obtained in step three i Establish a left-side point cloud set Point cloud set on the right The reference plane π obtained in the first step H Based on the baseline, the average height H of the symmetrical point clouds on the left and right sides left H right for Where, N left and N right These represent the point cloud sets on the left. And the number of points in the point cloud set on the right y i and y j These are the point cloud sets on the left. and the point cloud set on the right The coordinates of each point in the direction perpendicular to the reference plane; The average height H on the left side obtained from the above formula left and the average height H on the right side right The left-right height difference D of the vehicle under inspection (1) is obtained as follows: Where D is a positive value, it means the left side of the vehicle is higher, and a negative value means the right side of the vehicle is higher.