Semantic feature extraction technique for human bodies is essential for some application, such as reverse engineering, mannequin dimensions extraction and parametric design of mannequins. This paper present a method on 2D orthographic projection to extract the semantic feature of 3D human model. Firstly, we get the human models orthographic projection and use a new intelligent contour tracking algorithm to get the ordered silhouettes. Secondly, semantic features of the human body model are extracted from the silhouettes which is ordered by CCW counter clockwise before in orthographic projection. Compared with earlier approach, our method pointedly has avoided the complex extracted for 3D mesh surface, and both the robustness and the efficiency of the algorithm are improved.