Um novo método de reconhecimento de formas baseado em estatísticas de distâncias euclidianas.

Abstract

Shape analysis is an essencial branch of computer vision and until now several techniques have been suggested for shape characterization. To contribute to this field, this work introduces a novel shape descriptor based on statistics of Euclidean distances. The method consists in computing Euclidean distances between all points of the shape’s contour and filling up a matrix, which is subdivided into submatrices. For every submatrix, the algorithm calculates values such as mean, standard deviation and minimum distance in each column. Lastly, a shape signature is built from these statistics.