Planar-Target Calibration And Reconstruction Method Of Feature Points On Vehicle Body In Inspection System Of Vehicle Profile

Automotive industry developed rapidly in recent years,machine vision has gradually become a key technology to improve measuring efficiency and ensure detection accuracy.With the machine vision,the inspection system of vehicle profile could provide useful subject in several areas,such as automatic identification of overload vehicles,vehicle classification,detection of the vehicle parameters and reconstruction of the vehicle information,etc.Therefore,the research on the calibration of planar target and reconstruction of the feature points in inspection system of vehicle profile has important practical significance to automatically detect the vehicle,improve the vehicle performance and promote the development of detecting techniques.According to the previous research for the calibration of the laser plane and reconstruction of the feature points in inspection system of vehicle profile,we present an optimizing calibration method of the laser plane based on the planar target.In order to improve the accuracy of the extraction of the laser,the paper adopts the RANSAC method to fit the projected light.On the basis of projecting relation between the 2D in the image and 3D in the space,the equation of dynamic laser plane is obtained by the calibration,thus the 3D information of the feature points on the vehicle could be reconstructed through Open CV and Open GL.In order to improve the calibration precision of the laser plane and eliminate outliers which are affected by light and other factors,we use the RANSAC method with the point set of the laser centers which are obtained by the Hessian matrix.The extracted strip only contains interior points,so this method has high accuracy.The average distance from the laser centers to the extracted strip is selected as the assessment criteria.The experiment discusses extracting accuracy between the RANSAC method and the least square method.For the purpose of reconstructing 3D feature points of vehicle,firstly,we present an initial value calibration method of the laser plane for the detecting system.We establish a calibration model of the laser plane and analyze the generative process of the model parameters during the calibration.The internal and external parameters of the camera are calibrated through the planar target,the coordinates of projecting plane and target plane under the camera coordinate system.The initial coordinate of the laser plane can be solved through the Plücker matrix with intersection lines and singular value decomposition method.Then the maximum likehood estimate method is applied to optimize the laser plane.The optimized object is the reprojected minimum error of the parameterized projecting plane.The optimized objective function is established.The optimized coordinates of the laser plane is solved through through Levenberg-Marquard method.The optimized method is evaluated by the experiments of the reprojected line of the laser plane.Finally,according to the screw lead of the linear motion system with the laser projector,the coordinates of the laser plane in any movement position can be solved.Based on the projected relation among 2D coordinates of the laser points in the image and 3D coordinates in the space,and the condition of the laser points on the laser plane which has been calibrated before,3D coodinates of the feature points on the vehicle can be reconstructed.Several fuctions are realized through Open CV and Open GL,such as collecting images from the camera,extraction of the laser line from the differential image,indetification of the laser centers,reconstruction of the 3D coodinates of the feature points on the vehicle and display of the 3D point cloud of the vehicle,etc.Experiments are conducted with the reconstruction of the 3D coordinates of the feature points on the vehicle.The evaluated object is reconstructed distance error of the projected feature points on the ruler.Reconstruction method is verified by the experiment.