A Hybrid Moment Method for Identification of 3d-objects from 2d-monocromatic Images

The classification of a desired 3D object using Hu-moment invariants is based on identifying from six reference libraries an object having the minimum distance to desired object. To decrease classification error, this paper describes a modified approach to identify a 3D object based on a hybrid of Hu and Zernike moment invariants as a feature vector. Zernike moment invariants are used to find the pose of the object aircraft at first and then Hu moments are used to identify the object. The modified approach was tested by several experiments in different situations including noisy images. Finally, we show that the proposed approach achieves high identification rates compared with those of the View Information Encoded With Network Architecture (VIEWNET) and of the Hu moment invariants applying approximately the same set of objects and decision rule used in.