Object Recognition from a 2D lmage using a Genetic Algorithm

Abstract

A new approach for recognizing an object from a 2D image is proposed in this paper. In the proposed method, coarse 3D model for an object is prepared in advance. We employ the skeleton model as the 3D model to treat many objects which belong to the same category. A skeleton model can be represented by some straight lines. In the proposed method, first, the input image is transformed to the binary image and the binary image is thinned. Then the skeleton model is matched to the thinned image using a Genetic Algorithm(GA). In this matching process, the skeleton model is rotated, magnified or reduced in 3D space and the model is projected onto 2D space to match with the thinned image. In this way, the 2D images observed from various points of view can be obtained. If the object is recognized, the 3D shape of the object can be estimated. We treat the upper half of the human body as an example of an object in this paper, and computer simulation results indicate the validity of the proposed method.