Abstract: This paper proposed a novel feature descriptor based on interior dihedral angle histogram for non-rigid 3D shape retrieval to describe the concave-convex and local geometric features of 3D models. We first defined the interior dihedral angle histogram and analyzed its statistical properties. The specific steps were given to extract the histogram feature. Then we employed the genetic algorithm to optimize the weights of various features and designed a non-rigid 3D shape retrieval algorithm based on the combined feature. The experimental results on the non-rigid dataset published by SHREC show that the interior dihedral angle histogram is more discriminative and efficient. In addition the combined feature further improves the retrieval precision.