Research on 3D object recognition from wire-frame based on edge moment

This paper presents a method based on sub-area edge moment for the recognition of three-dimensional objects from wire-frame. In the recognition of 3D objects, 3D object wire-frame is divided into several fan-shaped sub-areas with the centroid as the center. By using improved edge moments calculation method, the low-order edge moments for each sub-area were computed. Then low-order edge moments for each sub-area were combined into a new set of edge moments feature vector to be the feature of a 3D object. These feature vector is presented to neural network for recognition. The theoretical and experimental analyses prove that using the sub-area edge moment as feature for 3D object recognition can achieve better recognition performance than using the edge moment.