A fast three-dimensional object recognition based on modulation analysis

Abstract

A novel method of the 3-D object recognition based on modulation analysis is proposed. Two orthogonal gratings with a small interval are vertically projected on an object surface and the recognized object is placed between the imaging planes of the two gratings. Then the image of the object surface modulated by the orthogonal gratings can be obtained by a CCD camera from the same direction, and the object height information is encoded in the image. The image is processed by the operations consisting of performing the Fourier transform and spatial spectrum analysis. The fundamental frequency can be obtained in two different directions. Then we calculated the recognition parameter RP (the normalized energy of each fundamental frequency). According to the RP values, we can determine the object position in the 2-D orthogonal co-ordinate system. The RP values will vary for different objects, so the different objects will be shown in different positions of the co-ordinate system. The expressions of the recognition parameter in the space and spectrum region have been given in this work. There are some advantages of our method. Firstly, it can recognize the 3-D object by the fundamental frequency energy analysis in two directions of the frequency domain, which cannot be almost discerned from their 2-D intensity images. Secondly, it has the advantages of shift-invariance and rotation-invariance, and can avoid the influence of shadow on 3-D object recognition. Finally the method has noise resistance ability. Both the computer simulation and the experimental results are presented in support of the proposed idea.