Abstract:Aiming at the need for the automated processing of the automobile steering knuckle,the pose estimation method of the steering knuckle was studied.The overall plan for steering knuckle inspection was constructed and the cameras distortion was calibrated,and the point cloud was generated according to the calibration result;in view of the center hole size of the steering knuckle was biased and the point cloud could not be directly used to solve the problem of indirect positioning of 3D coordinates,the 2D image processing was used to determine the pixel coordinates and the depth camera index depth values to determine the 3D coordinates of the grasping point;for different postures estimation problem of the knuckle,by using the SAC-IA algorithm,detected steering knuckle point cloud and template steering knuckle point cloud were initially registered,and the NDT algorithm was used for fine registration;the software was developed.The results show that the matching score is 27 mm2 by using the proposed algorithm,which verifies the accuracy of knuckle attitude estimation.