Ball detection in static images with Support Vector Machines for classification N. Ancona * , G. Cicirelli, E. Stella, A. Distante Istituto di Studi sui Sistemi Intelligenti per l’Automazione—CNR, Via Amendola 166/5, 70126 Bari, Italy Received in revised form 14 March 2003; accepted 18 March 2003 Abstract We present a general method for detecting balls in images at the aim of automatically detecting goals during a soccer match. The detector learns the object to detect by using a supervised learning scheme called Support Vector Machines, in which the examples are views of the object. Due to the attitude of the camera with respect to football ground, the system can be thought of as an electronic linesman which helps the referee in establishing the occurrence of a goal during a soccer match. Numerous theoretical and practical issues are addressed in the paper. The first one concerns the determination of negative examples relevant for the problem at hand and the training of a reference classifier in the case of an unbalanced number of positive and negative examples. The second one focuses on the reduction of the computational complexity of the reference classifier during the test phase, without increasing its generalization error. The third issue regards the problem of parameter selection, which is equivalent, in our context, to the problem of selecting, among the classifiers the machine implements, the one having performances similar to the reference classifier. Experimental results on real images show the performances of the proposed detection scheme. q 2003 Elsevier Science B.V. All rights reserved. Keywords: Support Vector Machines; Classification; Object detection 1. Introduction In this paper we focus on the problem of detecting the occurrence of a goal during a soccer match, by using methods and devices which does not require the modifi- cation neither of the ball nor of the goalmouth. Automatic goal detection in soccer is an open problem which is getting particular attention from referee associations, sport press and supporters. In fact, there are not rare situations in which a goal occurs, 1 but the referee and his collaborators (linesmen) are not able to detect the goal and, more important, do not award any point to neither teams correctly. Such situations occur for example when, after a shooting, the ball touches the internal side of the crossbar, bounces off the field having crossed completely the goal line and goes back, without touching the net. One of the most significative evidences of this event, named ghost goal, occurred during the World Cup final match on 1966 between England and West Germany. In that case, an English player struck a shot towards the German goal and the ball cannoned down from the crossbar, hit the ground and bounced back out into the play. English players claimed a goal, that the ball had passed completely over the line, and the referee, after consultation with his linesman, awarded the point to the England. Optical sensors, like standard TV cameras, seem to be appropriate for approaching the problem at hand for several reasons. First of all they satisfy the main constraint of the problem, because their exploitation does not require modifications of neither the ball nor the goalmouth. They can be placed externally and also very far from the field and, if equipped with appropriate zoom lens, they provide images of the goalmouth area useful for solving the goal detection problem. Moreover, they permit to have a direct evidence of the occurrence of a goal because the perceived images can be recorded on an analog or digital support for a successive analysis, for example by an external referee. Reid and Zisserman [1] proposed an uncalibrated binocular vision system for solving the problem of goal detection in soccer. Their method exploits two images of the field acquired simultaneously from two different 0262-8856/03/$ - see front matter q 2003 Elsevier Science B.V. All rights reserved. doi:10.1016/S0262-8856(03)00063-5 Image and Vision Computing 21 (2003) 675–692 www.elsevier.com/locate/imavis * Corresponding author. Tel.: þ 39-080-5481969; fax: þ 39-080- 5484311. E-mail address: ancona@ba.issia.cnr.it (N. Ancona). 1 A goal occurs in soccer when the ball completely crosses the goal line.