Motion Segmentation Based on a New Genetic Algorithm

Video surveillance has been used in many monitoring security sensitive areas such as banks, department stores, highways, crowded public places and borders. Detecting moving regions such as vehicles and people is the first basic step of almost every vision system, because it provides a focus of attention and simplifies the processing on subsequent analysis steps. It is one of the most difficult tasks, the motion segmentation accuracy determines the eventual success or failure of computerized analysis procedures. For this reason considerable care should be taken to improve the probability of rugged segmentation. Many approaches exist for detecting moving object in a sequence of images. Commonly used techniques for motion detection are background subtraction, temporal differencing and optical flow. Background subtraction[1,2,3,4,5,6] attempts to detect moving regions in an image by subtracting current image from a reference background image in a pixel-by-pixel fashion. Temporal differencing[7,8] makes use of pixel intensity difference between two or three consecutive frames in an image sequence to extract moving regions. Motion segmentation based on optical flow[9,10] uses characteristics of flow vectors of moving objects over time to detect change regions in an image sequence. In these methods, background subtraction provides the most complete feature data, but is extremely sensitive to dynamic scene changes due to lighting and extraneous events. Temporal differencing is very adaptive to dynamic environments, but generally has a poor performance in extracting all relevant feature pixels. Optical flow can be used to detect independently moving objects in the presence of camera motion. However, most optical flow computation methods are computationally complex, and cannot be applied to full-frame video streams in real-time without specialized hardware. Recently, genetic algorithms based motion segmentation have been proposed[11], It proposes a new video sequence segmentation method based on the genetic algorithm (GA) that can improve computational efficiency. The computation is distributed into chromosomes that evolve using distributed genetic algorithms (DGAs). But this method needs the segmentation in spatial and temperal field seperately. Then combines them together, which is time consuming. Inspired by the paper [11], we introduce a new motion segmentation method which combines the spatial segmentation and temperal segmentation together. First we construct the image model based on the Markov Random Fields (MRF) for each frame. Then the segmentation is represented by the minimization of a posterior energy function. We use the genetic algorithm(GA) to find the solutions. The background differecing and evolution probability are combined to find the unstable individuals. The advantage of this method is that it decrease the number of the evolution individuals and decrease the computation time consuming.