Calibration Methods And Measurement Precision Evaluation Of The Active-Vision Inspection System Of Vehicle Profile

The inspection system of vehicle profile is significant for improving the passability and safe performance of vehicles. The geometry information of vehicle profile is obtained without delay by the active-vision inspection system of vehicle profile. The system can automatically identify different vehicle models, overloading vehicles and refitted vehicles. The research of the active-vision inspection system of vehicle profile is of great significance for dimension detection of vehicle profile, vehicle intelligent classification and deformation detection of in-use vehicles. The research on the inspection system of vehicle profile perfects the detection technology of vehicle performance. The research has important practical significance and practical value to reduce the accident of vehicle and protect people’s lives and property.A camera calibration method based on point-line constraint and five laser plane calibration methods are proposed in view of the present situation of vehicle profile inspection. These five methods are the laser plane calibration methods based on known 3D coordinates, the know distance object between two points, the distance object between calibration points and the origin of world coordinate system, the distance object between calibration points and a axis of world coordinate system, and the distance object between calibration points and a plane of world coordinate system.In order to improve the detection accuracy of the active-vision inspection system of vehicle profile, a camera calibration method is proposed based on point-line constraint of the inspection system of vehicle profile. The lines are extracted from images by Hough transform. Another camera calibration method is described. The camera calibration method is based on point-point constraint of the inspection system of vehicle profile. Then, the camera calibration method is normalized. The calibration accuracy of the camera calibration method based on point-line constraint is compared to the calibration accuracy of the normalized camera calibration method based on point-point constraint of the inspection system of vehicle profile. In order to realize field calibration of the vehicle profile inspection system, a laser plane calibration method is proposed based on known 3D coordinates. The laser plane calibration method adopts a height gauge and a 3D calibration board. On this basis, four laser plane calibration methods are proposed for the inspection system of vehicle profile. These four methods are the laser plane calibration methods based on know distance object between two points, the distance object between calibration points and the origin of world coordinate system, the distance object between calibration points and a axis of world coordinate system, and the distance object between calibration points and a plane of world coordinate system. First, the target is placed in the view filed of the camera. The target is located on the position where the feature point on the target is coincident with the laser plane. The feature points on the target are obtained with the distance object. Then, the location coordinates of feature points in the images and camera transformation matrix are adopted to construct the objective function of laser plane equation. At last, the equation of the laser plane is optimized with several constraints. The laser plane equation are obtained in the world coordinate system. The position and posture of the laser plane are solved by the distance constraints and 1D or 2D targets in the world coordinate system. The transformation of the laser plane equation is avoided from the camera coordinate system to the world coordinate system. Direct measurement of the object is realized in the world coordinate system. In the section of testing and validation, the reprojection accuracy experiments are carried on between the camera calibration method based on point-line constraint and the camera calibration method based on point-point constraint. Then, the images are noised to compare the stability of two camera calibration methods. Next, the five laser plane calibration methods of the active-vision inspection system of vehicle profile are verified by experiment and analysis. Finally, 3D reconstruction of the intersection lines between laser plane and vehicle body are performed.