Shape Classification Based on Geometric Features of Evolution Points via Sparse Representation

Abstract

In this paper, a novel shape descriptor for shape recognition is proposed. An evolutionary process is introduced in which a contour is reconstructed from the bounding circle of the shape. In this evolutionary process, circle points always move toward the shape in normal direction until they arrive at the shape contour. Three different descriptors are extracted from this process: the first descriptor is defined as the number of steps that every circle point should pass from circle to shape contour which is called evolution steps (ES). The second descriptor is considered as the boundary distance (BD) of the sample points at the end of the evolution process. The third descriptor is the mean of curvature of the evolution lines that are created by moving points, (MCEL). In matching stage, dynamic programming is employed to best matching between shapes. Finally, normalizing the features makes them to be invariant to scale. Sparse representation as a new framework for classification is applied in the recognition stage. The proposed descriptors are evaluated for task of shape recognition on several data sets. Experimental results demonstrate the advantaged performance of the proposed method in shape recognition.