Movement detection using an order statistics algorithm

Abstract

ABSTRACT In this paper, we present a novel algorithm to motion detection in video sequences. The proposed algorithm is based in the use of the median of the absolute deviations from the me dian (MAD) as a measure of statistical dispersion of pixels in a video sequence to provide the robustness needed to detect motion in a frame of video sequence. By using the MAD, the proposed algorithm is able to detect small or big objects, the size of the detected objects depend of the size of kernel used in the analysis of the video sequence. Experimental results in the human motion detection are presented showing that the proposed algorithm can be used in security applications. Keywords: Median of the absolute deviations from the median, motion detection, video sequences 1. INTRODUCTION Video surveillance has been a popular security tool for years [1]. They are used for safety and security in public environments and in the private sector. The science and technological applications of advanced video surveillance systems have progressed tremendously in recent years due to widened research in areas including transport networks, elder care, traffic monitoring, traffic flow analysis, endangered species conservation, home nursing, human activity understanding, observation of people and vehicles within a busy environment, etc. [2]. The design of an automatic video surveillance system requir es several critical functionalities such as, motion detection and tracking, classification, behavior monitoring, activity analysis, and identification [2-6]. Motion detection is one of the current research areas in the design of video surveillanc e systems. It provides segmentation of the video streams into foreground and background components in order to extract the desired moving objects and it is a critical preprocess for several computer vision applications, including object-based video encoding, human motion analysis, and human–machine interactions [2]. Detecting humans in frames of video sequences is a useful application of computer vision. Loose and textured clothing, occlusion and scene clutter make the human detection a difficult problem due bottom-up segmentation and grouping do not always work [7]. Several methods to motion detection can be implemented in video surveillance systems [1-7]. The common and easy way to develop such a system is to compute the absolute value of the difference between the current frame and a reference frame. After this, a threshold is imposed in the difference matrix. The pixels whose value are higher than a threshold are supposed to be part of an object moving into the scene, and those that are smaller belong to the background [4]. The value of the threshold to be chosen has a critical part in th is kind of systems, this value must be applied to the whole image. If the value of the mentioned threshold is small, the false alarm probability can increase, mainly due to the noise picked up by the camera, and the noise gene rated by a non-static background, i. e. the natural noise in an image such as a branch tree moving in a windy day. Otherwise, if the value of the threshold is high, the miss probability increases because the absolute difference between the background gray level and the object gray level is likely to be lower than the