An Efficient Algorithm for Detection of Soccer Ball and Players M. M. Naushad Ali, M. Abdullah-Al-Wadud and Seok-Lyong Lee * Department of Industrial and Management Engineering Hankuk University of Foreign Studies, REPUBLIC of KOREA naushad_iut@yahoo.com, wadud@hufs.ac.kr, sllee@hufs.ac.kr Abstract: In this paper we propose an effective algorithm for detection of a soccer ball and players. In the currently available methods, the ball may not be detected when it is attached to the lines in the ground. To overcome this problem, we introduce a method to separate lines from the ball and segment the ball effectively. At first we detect the ground and the edges of the original image. Then an efficient line detection method is applied, and finally players and a ball are extracted. The experimental results show that our algorithm is robust when the ball is attached to the line. Keywords: Ball detection, Line Hough transform, Image analysis, Object Recognition 1 Introduction In the field of machine vision, detection of moving objects is an interesting and difficult issue. Moving object detection is applied in different application areas such as medical image processing, human tracking, traffic system monitoring and sports video analyzing. Nowadays, sports video is one very popular research area, which involves players' movement analysis, tactics, referee's decisions, and so forth. So, automatic detection of players and a ball in the soccer video is notably important [1]. Usually, the color [2] and shape [3] are used to segment the ball. The method proposed in [4] tracks the objects present in the scene and considers the object that is tracked for the longest time as the ball. However, these simple features may fail to detect the ball when similar objects such as socks of the player, and the ground line appear in the scene. A trajectory-based algorithm is proposed in [5] to estimate the ball candidates in each frame based on the size. A circle detection algorithm based on circle Hough Transform is proposed in [6] to detect the circular shape ball in the image frames. In [7], color, shape and size are evaluated to detect the ball. In [8], density based approach is used to separate the players and the lines. However, as we focus on frames involving long-shots, the ball is * Corresponding Author 39