Abstract
This paper proposes a fracture surface matching algorithm based on integral invariants. We compute points' volume descriptor on multi-scale, search similar feature points to build matching constraint clusters. Extracting constrained cluster feature and giving a representation, according to convex and concave correspondence of cluster surface, it obtains initial matching cluster pairs. We use spatial geometric consistency constraint's vote method and search the maximum independent set to prune non-matching pairs, so this coarse registration becomes an optimization problem. At last two fragments can be precisely aligned by ICPIF method based on the result of coarse matching. Experimental results show that the algorithm can achieve better matching and reassemble.