An automatic method for the measurement of coarse particle movement in a mountain riverbed

Abstract

In this paper, we propose an automatic approach that is able to recognize the corresponding particles in multi-temporal point clouds and to determine 3D displacement vectors and the rotation parameters between them. The core of this method is a moving window approach combined with the Iterative Closest Point (ICP) algorithm. The particle-wise ICP algorithm is supplemented by initial transformation parameters obtained by key point matching. The routine is designed to be applied to natural objects, which are characterized by their complex geometry. The method’s performance is verified using terrestrial laser scanning data sets representing a mountain riverbed. The experiments performed show that the method enables us to recognize corresponding particles with an effectiveness of 85%. The mean absolute distances between the tile of a point cloud and the particle after alignment are used as the accuracy of the alignment. The median of these values is equal to 2 mm.