Método para extração de objetos de uma imagem de referência estática com estimativa das variações de iluminação

Abstract

Video segmentation is a fundamental step in many vision systems including video surveillance and traffic monitoring. Background subtraction is a method typically used to segment moving regions in video sequences taken from a static camera by comparing each new frame to a model of the scene background. In this paper, a hardware system for video segmentation is proposed from algorithm to hardware architecture level. The video segmentation algorithm is aimed at fixed-point operations and improves a Gaussian background model by applying a two-stage linear compensation procedure to remove the undesirable subtraction results from noise and illumination changes. First, the algorithm was validated in MATLAB. Then, it was prototyped on an Altera field-programmable gate array platform (DE-2). At a clock rate of 100 MHz, the architecture can process 30 frames per second, where the image resolution is 640 x 507 pixels. The capability of the system is demonstrated for several video sequences.