Criação de filtros de imagens através da utilização de programação genética

Abstract

The techniques of computer vision have been used more and more by the industries in order to aid the automation of their processes; however, the implementation of computer vision techniques has several difficulties according with the application. Such techniques are limited to the computational cost along with its complexity. Problems of elements identification in industrial sceneries are examples of an application that can generate a larger complexity in the process automation. Researches are accomplished using several techniques for solving those difficulties. This work approaches the field of computer vision through the use of artificial intelligence techniques. Here, methods are evaluated for creation and replication of binary images filters through the use of the genetic programming with the objective of elements identification in an industrial scenery. Two methods that possess different approaches are evaluated; one uses operations between pixels and other mathematical morphology for objects detection. The results are presented qualitatively as well as quantitatively through comparative images of the evaluated methods and statistical measures, respectively. Through the GPLab toolbox together with Matlab, the automatic creation of filters for objects identification was possible, so the detection and deletion of elements in images can be used by other support systems to automatic operation in an industrial environment. The results established a efficient way of filter creation with the use of the genetic programming.