Reconhecimento de objetos em imagens digitais utilizando otimizaÃÃo por enxame de partÃculas / Object recognition in digital images using particle swarm optimization

AUTOR(ES)
DATA DE PUBLICAÇÃO

2009

RESUMO

The automatic recognition of objects aims to extract the visual information by computional devices. One way of addressing the problem of object recognition is through the technique of Template Matching viewed as a problem of global optimization. The objective is to determine the position of an object in an image, its rotation angle and the scale factor used. The Particle Swarm Optimization (PSO) method is fairly widespread in the literature and widely used in global optimization problems, and was chosen as part of the solution. The measures of similarity are important to the operation of the method, so studies were performed to select a good function. Also, experiments were performed to determine a set of parameters suitable for the optimization method, with the aim of improving the quality of solutions provided. To determine its effectiveness, the method was tested with color and gray-level images, with translation, rotation, scale change and partial occlusion of objects, as well the presence of noise and change of brightness in the images. Moreover, a comparison was made with another evolutionary approach to object recognition, to verify the performance of the proposed method. The results showed that the PSO was able to find the predefined objects with high performance, suggesting that theproposed method is very promising, especially with respect to robustness and simplicity. So this may be an option to be used in real-world object recognition applications of computer vision

ASSUNTO(S)

computer vision otimizaÃÃo por enxame de partÃculas image processing - digital techniques processamento de imagens - tÃcnicas digitais visÃo por computador sistemas de reconhecimento de padrÃes particle swarm optimization pattern recognition systems engenharia eletrica

ACESSO AO ARTIGO

Documentos Relacionados