Binocular-active-vision Inspection Method Of Vehicle Profile Based On Perpendicular Vector Constraint

In recent years,with the increasing number of automobiles,the driving safety and operation maintenance of automobiles have become the focus of increasing attention.Therefore,the regular detection of vehicle performance parameters has become an important research content in the field of automobile application engineering.Automobile morphology detection is an important part of automobile inspection research,and three-dimensional reconstruction technology based on machine vision has also become an important method of object morphology detection.The automobile morphology detection method based on machine vision has the advantages of non-contact measurement,high degree of automation,low cost and so on.It can provide reliable basis for judging vehicle type,vehicle out-of-limit detection,vehicle size parameter measurement,etc.Therefore,the research on automobile morphology detection method based on machine vision is of great significance for the rapid detection of vehicle size,vehicle intelligent classification and so on.The research status of vehicle shape detection is analyzed.To unify the measurement results of the non-public field-of-view camera used in the large-scale inspection of automobile body,the binocular active vision automobile morphology inspection method based on internal and external closed-loop solution model was studied.The detection system of this method includes binocular vision left and right internal camera,laser projection plane and external camera without common field of view,and the internal closed-loop solution model and external closed-loop solution model were constructed to realize the camera calibration of each part.According to the established cylindrical target-left and right interior camera-single external camera internal double closed-loop model,the projection matrices of the left and right camera and the three-dimensional homography relationship between the cylindrical target and the left and right camera were solved.Based on the external double closed-loop model of cylindrical target-relay camera-two external cameras,projection matrices of external cameras without common field of view and three-dimensional homography between external cameras were solved.Based on the solution of the internal and external double closed-loop models,the structured light laser plane of active vision was calibrated by using difference shadow method and Hessian matrix.In the reconstruction of laser projection point with car body,firstly,the binocular epipolar line matching was carried out in the left and right interior cameras of binocular vision.After that,the laser plane calibration results were taken as the initial value,and they were preliminarily optimized by minimizing the coincidence error of laser plane coordinates under the cylindrical target.To realize the accurate reconstruction of length parameters and angle parameters at the same time in the automobile morphology detection,according to the characteristic that the space vector has both length and direction information,the vertical vector constraint composed of Cartesian ruler was constructed to optimize the laser plane again.The vertical vector constraint took the orthogonality of the vector as the angle constraint and the vector module length as the distance constraint.At the same time,both the reconstruction angle and the reconstruction length were taken into account,and the two optimization objectives were balanced.After the optimization of the reconstruction method,the reconstruction model of the automobile morphology binocular active vision inspection method was established according to the projection relation and homography relation of the calibration solution.The reconstruction experiment of the laser intersection was carried out,and the automobile morphology was reconstructed in the coordinate system of external camera.Finally,the influence of different calibrator shapes on the uniformity of projection points was analyzed,and the influence of system baseline distance and measurement distance on detection accuracy was studied through experiments.