A 3D polygonal line chains matching method for face recognition

Abstract

In this paper, a novel 3D polygonal line chains matching method is proposed. Different from traditional method that use points and meshes to represent and match 3D shapes, our method represents 3D surfaces using 3D polygonal line chains generated from ridge and valley curves. Then a 3D polygonal line segment Hausdorff distance measure is developed to compute the similarity between two 3D surfaces. This representation, along with the distance metric, can effectively harness structural and spatial information on a 3D surface. The added information can provide more and better discrimination power for object recognition. It strengthens and improves the matching process of similar 3D objects such as 3D faces. Experiments on FRGC v2 database leads to a rank one recognition rate of 96.1%.