An Automatic Spherical Targets Detection Method with Multiple Geometrical Constraints

Abstract

Spherical targets are used extensively in the registration and coordinate transformation of the railway point cloud. Thus, it is necessary to accurately detect the spherical targets from the railway point cloud. This paper proposes an automatic spherical targets detection method with multiple geometrical constraints. In this method, possible spherical points are extracted by the improved three points filter method. And possible spherical points are refined according to neighborhood height difference and curvature. Then, the refined possible spherical points are spatially clustered by the Euclidean clustering method and the potential target point clouds can be extracted by constructing the spherical neighborhood according to the cluster centroid. Finally, the ratio constrained random sample consensus (RC-RANSAC) method is proposed in this paper, based on the RANSAC method, to detect the spherical targets in the potential target point clouds. The point cloud scanned from the high-speed railway is taken as experimental data. The spherical targets in the point cloud are detected by this method. The experimental results show that the proposed method can detect the spherical target with and without the background in radial direction.