Implementação em hardware de um sistema para extração de objetos de um fundo não homogêneo em tempo real.

Abstract

It is presented in this thesis the implementation and definition of a hardware system for extraction of objects from images with static and non-homogeneous backgrounds in real time, robust with respect to minor global and local brightness variations. The object extraction system was implemented on the GPIP01 (General Purpose Image Processor) platform designed by Andre Printes. Little modifications on the platform were done for the implementation of the system. The object extraction system was submitted to actual environment situations to test the performance of the extraction method implemented in hardware. The results of the tests and evaluation of the system demonstrate its capability to extract object from non-homogeneous backgrounds, to deal with differences between background objects and foregrounds objects with the same color and its tolerance to brightness variations and shadows in the background.