Feature extraction for automatic ball recognition: comparison between wavelet and ICA preprocessing

Abstract

The ball detection in soccer images is one of the applications of the most general problem of object recognition, where the approach mainly used is based on classifying the pattern images after a suitable preprocessing. In this paper we have compared two different preprocessing techniques: the initial vectorial representation of the image has been projected both on the Haar basis and on the basis extracted from the independent component analysis. The coefficients of the new representation in the ICA and Wavelet subspaces are supplied as input to a neural classifier. ICA and wavelet representations have been chosen since they are well suited to increase the inter class differences and decrease the intra class ones. The experimental results on real soccer images show that the classification performances applying the ICA and wavelet preprocessing techniques are quite the same and that combining ICA and wavelet the percentage of pattern recognition can be further increased.