An Algorithm for Binary Contour Objects Representation and Recognition

摘要: Representation of an object in image for machine learning applications (recognition, retrieval, identification, etc.) has to be based on a previously chosen feature. Binary shape is very popular and commendable one. It many advantages can successfully used applications, especially engineering. To achieve better characteristics, various transformations are used. Obviously, they should robust as deformations it possible. In this paper results exhaustive exploration new method presented. This transformation from Cartesian polar coordinates, but overcomes few problems, that were not solved yet. Above all, the proposed transform more occlusion noise, two most challenging problems.