Identificação de fronteiras de regiões fechadas mediante o estudo da curvatura e análise de wavelets.

Abstract

This work presents a methodology which can be used to identify falsifications, adulterations and factory deformities. The proposed method consists on shape identification, employing the curvature analysis of parametrical path in plane, together with the one-dimensional wavelets discrete transform, seeking, in order, to rebuild the original image. The schedule followed was basically the following: initially, the two-dimensional images were selected and processed in a graphic editor software, finding points on the boundaries of each styles, with the aim to generate the curvature function in a numerical form. The next step consisted in applying the one-dimensional wavelets discrete transform, with the intention to find sub-signals which represent the main characteristics of the curvature function of each image. Finally, the models were identified and the original image was rebuilder. The combination of the notions of Differential Geometry, Wavelets Discrete Transforms and Principal Components Analysis indicated to be favorable to style identification, achieving the proposed target to one-dimensionalyze the styles and identify the shapes. This method can be applied in industrial areas that use boundaries characterization, which is the case of fashion designs, aerodynamic profiles and computer graphics, mentioning only a few possible situations.