Research And Application Of Point Cloud Recognition Algorithm Based On Feature Matching In Workpiece Picking

3D point cloud target recognition is the core task in the field of machine vision and intelligent detection,and it is also the key technology of automatic picking by industrial robots in recent years.Aiming at the problem that the current feature description algorithm has weak description ability and low recognition accuracy for noisy point clouds,a feature description algorithm based on local point pair features is proposed.A simplified asymptotic consensus algorithm for matching screening is proposed.Aiming at the problem that the classical ICP recognition algorithm is prone to fall into local optimum and has low efficiency,a point cloud recognition algorithm based on local point pair features and ICP is proposed.Compared with the traditional algorithm,the algorithm in this paper is significantly improved in accuracy and efficiency,and can realize the accurate identification of industrial components.The main research contents of this paper are as follows:(1)Aiming at the problems of poor robustness and large amount of computation in the classical point-pair feature description algorithm when dealing with noisy point clouds,a feature description algorithm based on local point-pair features is proposed.The algorithm constructs a feature descriptor by calculating the geometric feature parameters of the point cloud in the local range of the feature points,performs histogram statistics on the feature parameters,and determines the optimal histogram distribution of the descriptor.Descriptive comparison experiments show that the constructed descriptor has a stronger ability to describe the spatial differences of noisy point clouds and is better than similar algorithms.(2)Aiming at the problem of false matching elimination based on eigenvector Euclidean distance matching feature points,a simplified asymptotic consensus algorithm for matching screening is proposed.The algorithm first uses the rotation invariant deviation to obtain the hypothesis generation set and estimates the pose transformation parameters,and then uses the parameters to check the correctness of the matching of other feature points,and adds the matching relationship that meets the threshold condition to the hypothesis generation set to obtain the final matching result.The matching experiments show that the pose parameters calculated by this algorithm are more accurate and can effectively reduce the error of point cloud recognition.(3)Aiming at the problem that the classical ICP recognition algorithm is prone to fall into local optimum and has low efficiency,a point cloud recognition algorithm based on local point pair features and ICP is proposed.The algorithm uses local point pair features and completes rough identification of point clouds through feature matching,which provides a good initial value of rotation and translation matrices for ICP algorithm.Coarse-fine recognition experiments show that the algorithm in this paper can effectively reduce the number of recognition iterations and pose recognition errors between the model point cloud and the scene point cloud.The comparison experiment of workpiece recognition shows that the algorithm in this paper is superior to the similar algorithms in terms of recognition time and recognition accuracy.