NOVEL FRAMEWORK FOR VIDEO OBJECT SEGMENTATION AND TRACKING USING AUTOMATIC THRESHOLD DECISION AND DIFFUSION DISTANCE

Abstract: Video surveillance process takes video as an input, processes the video frames and performs actions accordingly. The video surveillance process consists of many phases. Video object segmentation and tracking are two crucial building blocks of smart surveillance systems. Threshold decision is a complex problem for video object segmentation with a multibackground model. Some conditions like nonrigid object movement, target appearance variations due to changes in illumination, and background mess make robust video object tracking complex. Multiple thresholds might be needed in connection with more refined algorithms. To make it completely automatic for variant conditions, First an automatic threshold decision technique that can automatically and specifically determine the threshold values for dynamic backgrounds is proposed. Second, a video object tracking framework based on a particle filter is proposed with the probability function composed of diffusion distance for measuring color histogram similarity and motion clue from video object segmentation. The proposed framework will track multiple moving objects under drastic changes in illumination and background mess.