MIMO Model-Free Adaptive Control Color Background Image Extraction to Video

Abstract

The current video background image extraction methods mainly obtain the gray background image, and the gray image processing is very sensitive to the interference noise, which brings great difficulty in detecting and tracking of moving target accurately. In order to overcome the above problems, a novel color background image extraction method is proposed using the idea of model free adaptive control method. The method introduces the pseudo-Jacobian matrix of the system and combines RGB three-channel historical data of pixels to establish the color background image extraction and update. The proposed method that under different video conditions is compared with the traditional gray background image extraction methods. The results show that the method can extract the color background image intuitively, and the separated foreground is closer to the ground truth of the video target.