Feature Extraction Of Point Cloud Data Based On RANSAC

Feature extraction from point cloud data, have many important applications in the field of computer vision, pattern recognition, city, geology and other professionals. Feature points in point cloud data, as the basic texture and geometric motifs characteristic features, are not due to change in different coordinate system. Since the birth of the three-dimensional laser scanning, it makes the point cloud data have a new life, and how to extract feature points from point cloud data has become a hot research direction.The basic elements, point, line, surface, body, etc., as a fundamental part of the point cloud data, and extracting these interested elements have many research achievements. In the two-dimensional image, Hough Transform(HT) algorithm is an important method for extracting straight lines, circles, etc., HT algorithm has been shown to have good robustness. Drawing the Hough Transform method in two-dimensional image, the 3D-HT become a feasible approach to extract surface feature points from three-dimensional data. RANSAC algorithm is also a common method of feature extraction form data, and it reflects good results and robustness in extracting regular objects from point cloud data. The least squares method and its improved algorithms are also used to discrete data points in the surface fitting and feature extraction. This paper focuses on 3D-HT, the least squares method and its improved algorithm, RANSAC algorithm to extract the points of plane and sphere features, then discusses the differences between them. By contrast different three methods, RANSAC algorithm demonstrated the superiority.Lastly, RANSAC algorithm was used to extract feature points from two actual point cloud, and extraction results with Geomagic simple comparison, further validation RANSAC superiority in point cloud extraction. Least Squares;3D-HT;RANSAC...