Objects matching using optimization techniques [Emparelhamento de objectos utilizando técnicas de optimização]

Abstract

In this work, the objects matching problem (composed by data points) is solved as a global optimization problem. The proposed solution is compared with a previously developed one, based on a local approach. With this comparison, the significant improvement of the results is confirmed. In this paper, two previously developed methodologies for obtaining the quantification of the affinities between the data points are also presented. One of these methodologies is based on the modal analysis of the shapes of the objects, while the other one uses the finite elements method and the modal analysis. Usually, the existing solutions for matching problems allow only matches of type "one with one". However, there are many applications in which this type of matching is not adequate as it can imply loss of information. To overcome this problem, an applicable solution was developed for contour objects. This solution allows the adequate matching of the excess points, getting correspondences of type "one with several" or vice versa.