Point Cloud Registration Based On Non-coding Markers

The registration of 3D point clouds is an indispensable task in reverse engineering.The registration and fusion of 3D point clouds from multiple perspective views is a problem that scholars and researchers have been studying for a long time.The point cloud registration based on non-coding markers is focused on in this paper.The main innovations and work of this paper are as follows:1)In order to detect markers in images quickly and accurately,an ellipse detection algorithm based on the properties of elliptical circumscribed rectangles is proposed.Firstly,the director circle of the ellipse is introduced into the positioning of the ellipse center;then,the other parameters of the ellipse are determined by the properties of the elliptical circumscribed rectangle;finally,the ellipse is clustered according to the length of major axis and minor axis by the MeanShift algorithm to extract the target ellipse class.Experiments in simulation and real images show that the algorithm has high precision and excellent detection performance.The algorithm also have good recognition performance for ellipse with partial defects;2)In order to eliminate the wrong matching caused by measurement error,markers detection error and markers pasting error,a new matching method that imitating coding markers matching is proposed.Firstly,the distance matrix is used to find all matching triangles.Then three filter criteria are constructed to extract the correct matching set.The coarse registration is completed by using the matching information.For fine registration,an improved Trimmed ICP is proposed to avoid overmatch of points.Experiments show that the algorithm does not depend on other parameters to determine overlap rate.Good registration results can be obtained at point clouds with different density;3)In order to remove point clouds in overlapping area after point cloud registration and preserve the point cloud details,fusion algorithm based on the normal vector difference is proposed.On the part where the difference between the normal vectors of matching points is small,the point cloud is smoothed and then fused.On the other part,the point cloud fusion algorithm based on normal constraint is used.Experiments show that the proposed algorithm can effectively preserve the local features of the point cloud.Finally,the global registration of point cloud is designed from integrating the contents of chapters such as markers detection,point cloud registration and fusion.