The Method of Shape Recognition Based on V-System

This paper uses a new class of complete orthogonal system in L₂[0, 1], which consists of piecewise polynomials (called V-system), to process shape recognition. A novel boundary based shape recognition method is proposed. First the contour of an object is regarded as a geometric graph, and the graph is expanded in a V-series, which transform the graph to the spectrum space and quantify overall feature of the object. Further a shape similarity measure algorithm based on normalized V-descriptor (a new shape descriptor) is presented. Several shape recognition experiments under different conditions and circumstances for some objects selected from database MPEG-7-Part B are given. The comparative study between normalized V-descriptor and classical normalized Fourier descriptor shows that normalized V-descriptor presents better performance than normalized Fourier descriptor in shape recognition.