Matching Technology of Point Cloud Models Based on Combination Features

Abstract

A combination features based technology is proposed to matching point cloud models. Firstly, automatically segment the point cloud to get ruled shape data from which the statistics features are extracted. Calculate the distance and angle between component elements of statistics feature and datum plane. The distance and angle constraints are introduced to automatically search correspondence statistics features of overlap region and the optimal combination of statistics features. The optimal combination is coupled for combination feature. If combination feature can completely identify pose of model, we call this as completely matching question; otherwise, this is called as partly matching question. For the latter, adjust the degrees of freedom which are not constrained and then precisely match the models using ICP algorithm. The examples show the proposed algorithm is simple and efficient, and the matching results are stable, reliable and convergent to a global optimal solution.