Research And Implementation Of Target Recognition Based On 3D Point Cloud Data

Target recognition is an important branch in the field of computer vision,and it is a hotspot in recent years.It has broad application prospects,such as dining robots,intelligent logistics cars,unmanned aerial vehicles and so on.It has potential commercial value.In recent years,with the rapid development of 3D data acquisition technology and diversification of acquisition methods,object recognition and 3D reconstruction based on 3D data have become the research focus of many scholars and experts at home and abroad.In this paper,the research of target recognition based on 3D point cloud data is carried out under the impetus of this research upsurge.The point cloud data in this paper is obtained by Tintin of Texas Instrument.In this paper,point cloud denoising,point cloud segmentation and edge extraction are studied innovatively.In denoising,aiming at the limi tation of the pass through filtering algorithm in dealing with multiple dimensions at the same time,an improved pass through filtering algorithm is proposed in this paper.The improved algorithm realizes the simultaneous processing of multiple dimensions,and shortens the processing time of point cloud data compared with the original algorithm.The traditional point cloud segmentation algorithm can not segment the object and other objects clearly from the point cloud data in this paper,so an improved Euclidean distance point cloud segmentation algorithm is proposed.In the boundary extraction of point cloud data,the boundary extracted by the point cloud boundary extraction algorithm based on Micro-Tangent surface is inconsistent.In view of this situation,this paper proposes an improvement to the algorithm.The point cloud data of sampling point and its K neighborhood are projected on the Z=0 plane,and then the edge of the projected point cloud is detected and extracted by using the characteristics of the boundary.Through experiments,in the filtering stage,compared with the original pass through filtering algorithm,the improved filtering algorithm reduces the running time by 20 ms when processing a single point cloud file;in the point cloud segmentation stage,the improved segmentation algorithm accurately extracts the point cloud data of object objects;in the boundary extraction,the improved algorithm improves the inconsistency of the original algorithm's boundary extraction.