Panca Mudjirahardjo, Hadi Suyono
The first task in motion detection and recognition is moving object extraction. After extract it, we perform representation model, or feature extraction prior to evaluate it with a classifier process. Failure in moving object extraction will contribute an error in detection and recognition process. This extraction is performed by processing of several frames to distinguish foreground from the background. In this paper we developed a simple extraction method which was suitable for real time application. The first stage, we performed moving object detection by using frames subtraction. This yielded boundary label in the motion area. This boundary label couldn't extract the whole moving object. For every row of image, we calculate the boundary density function. The probability of pixels in this density is used to extract them as foreground. The experimental result, our system can extract the moving object in computation time of 71 - 80 ms, video rate of 15 frames per second (fps) and image frame size of 640×480 pixels, which is suitable for real time application. © 2018 IEEE.
Department of Electrical Engineering, Faculty of Engineering, Brawijaya University, Malang, Indonesia
