Study On Vision Detection Method Of Wheel Alignment Parameters Based On Vehicle-body Reference Plane

The issue of driving safety has gained significant attention in response to the growing number of motor vehicles in China.Regular detection and adjustment of vehicle wheel alignment parameters are crucial to ensure vehicle handling stability and driving safety.However,the operation of traditional contact-type four-wheel aligners is relatively complex,and there are discrepancies between the measurement benchmark and national standard.In response to these issues,this study explores a visual detection method which enables the noncontact measurement for the wheel alignment parameters based on the vehicle-body reference.This study is built upon the techniques such as extracting the shape of the wheel rim,segmenting the point cloud of the wheel rim,matching the point cloud of the wheel rim,detecting the symmetry plane of the vehicle-body,and calibrating the camera globally because of the characteristics of the wheel alignment parameter detection technology.The study costructs a symmetrical plane calculation model for the vehicle-body.It also establishes a calculation model for the vehicle kingpin and wheel axis.Additionally,it calibrates the system of multiple cameras and performs visual inspection of wheel alignment parameters using the vehicle-body reference.To verify the accuracy of the wheel alignment parameter detection system,a verification device is used to analyze the accuracy of the detection system,and vehicle tests are conducted.The specific researches are as follows:(1)A method is proposed for detecting vehicle wheel ellipses based on three intersecting chord invariants of ellipses,which aims at the problem of difficult extraction of ellipses in vehicle wheel images due to different environmental lighting,complex wheel texture,and black wheel,Firstly,a dataset of wheel rim images is established which contains different wheel rim textures and shooting conditions.Secondly,the accurate rim edge arc segments are obtained by performing edge detection on the region of interest in the car wheel image and detecting the bifurcation and mutation points of the rim arc segments.Then,many nonhomologous arc segment combinations are quickly removed by applying quadrant constraints and midpoint distance constraints.The initial sets of ellipses are effectively generated based on the proposed three intersecting chord invariants of ellipses.Subsequently,elliptical validation and clustering are performed on the initial sets of ellipses to obtain the highprecision rim ellipses.Finally,the accuracy of the algorithm is verified using six publicly available datasets and two established wheel rim datasets.The wheel rim ellipse detection method,which is based on the three intersecting chord invariants of ellipses,has high accuracy by comparing with other ellipse detection algorithms.(2)In order to address the issue of obtaining accurate positions for the kingpin and wheel axis through precise and efficient matching of the wheel rim point clouds,a solution method is proposed which is based on Gaussian weighting of the extended Kalman hidden Markov conditional random field.Firstly,the position of the wheel rim center,the normal vector of the wheel rim plane,and the wheel rim radius are obtained by the imaging projection equation of the wheel rim and the elliptical conical surface equation,which is formed by the camera optical center and the spatial circle of the wheel rim.The wheel rim point clouds are segmented accurately based on these geometric features of the wheel rim and Kd＿Tree algorithm.Afterwards,the EKF algorithm is employed to track the center and normal vector of the wheel rim point cloud.It allows for a rough matching of the rim point cloud,and the resulting rough match serves as the initial value for the rim point cloud matching algorithm of the hidden Markov random field.A Gaussian weighted solution model for the kingpin and wheel axis is constructed,which is based on the motion characteristics of the wheel rim point cloud around the rotation axis.Finally,simulation analyses are conducted on the calculation model of the kingpin and wheel axis to verify the accuracy of measuring the kingpin and wheel axis.And this study explores the factors that influence the accuracy of the solution model for the kingpin and wheel axis.The simulation results show that this method has better detection performance for the kingpin and wheel axis,while also conforming to national standards in terms of accuracy.(3)The current method for detecting wheel alignment parameters determines the reference plane based on the relative position of the wheels,which differs from the requirement of national standards that mandate the vehicle body as the reference.The Expectation Maximization method from coarse to fine is studied to solve the problem of the symmetrical plane of the vehicle body.Firstly,this method constructs an EM algorithm based on coarse-to-fine iteration and utilizes the Kd＿Tree algorithm to construct a set of candidate symmetric points in the M neighborhood of the vehicle body points.The purpose is to accelerate the parallel calculation of the vehicle-body’s symmetry plane.Then,a vehicle point cloud dataset with standard symmetry planes is established.A method for evaluating the benchmark measurement of wheel alignment is designed using the projection angle of the symmetrical plane of the vehicle-body.Finally,the method is compared with other methods using the vehicle dataset and evaluation indicators.Additionally,the impact of measurement distance and missing vehicle-roof point clouds are investigated in the accuracy of solving the symmetry plane of the vehicle-body.(4)A global calibration method is proposed for wheel alignment parameter detection system based on quaternion,surface light field,and textureless targets.The method aims to address the issue of low calibration accuracy in the camera system particularly for cases involving multiple RGB＿D cameras with longer distances and larger angles.This method uses surface lasers and a textureless plane to calibrate multiple remote cameras in the visual inspection system.The 3D laser planes with a long projection distance serve as a bridge between two cameras and the quaternion is used to solve the pose between the two cameras.The laser projection plane emitted by the surface laser equipment is not easily divergent,and the use of non-textured target at close range to obtain the intersection points of the laser solves the problem of low point cloud accuracy in long-range RGB-D camera acquisition.Finally,the unification of multiple camera coordinate systems is completed in the detection system.In addition,the above researches are used to conduct verification device evaluation tests and vehicle tests.The visual detection system for the measurement of wheel alignment parameters which based on the vehicle-body reference plane has high accuracy and meets the requirements of national standards for the measurement accuracy of wheel alignment parameters.The above researches achieve non-contact measurement of wheel alignment parameters and solve the error problem caused by traditional wheel alignment parameter detection methods using the relative position between wheels as the measurement benchmark.These researches play a positive role in promoting the technological development of the vehicle inspection field.