Abstract: In this paper, apartial shape matching algorithm is proposed for high noise relic fragments based on surface feature optimization.We compute volume integral invariant of the fracture points under multi-scale and search similar feature points to build matching constraint clusters.Initial matching cluster pairs are obtained by extracting and representing feature for constrained cluster based on convex and concave correspondence of cluster surface.Then the rough matching problem can be converted to an optimization problem by applying the method of consistent constraint vote in geometric space and searching the maximum independent set to prune non-matching pairs.Finally two fragments can be precisely aligned based on the result of rough matching by using iterative closest points using invariant features.Experimental results show that the algorithm can achieve better matching and stitching.