Independent Component Analysis for Ball Recognition in Soccer Images

M. Leo, T. DOrazio, and A. Distante (Italy)


Object Recognition, Feature extraction, Independent Component Analysis, Neural classifier


The ball detection in soccer images is one of the application of the most general problem of object recognition, where the approach mainly used is based on classifying the pattern images after a suitable pre processing. In this paper the pre-processing step is performed projecting the initial vectorial representation of the image on different subsets of basis extracted from the Independent Component Analysis. The coefficients of the new representation in the ICA subspace are supplied as input to a neural classifier. ICA has been chosen since the basis vectors are extracted directly from the data and the parameters of the new representation are independent. This paper shows that ICA is well suited to pre-process the data in order to recognize a pattern, and in particular together with a neural classifier it performs well to detect the ball in soccer images. Besides the paper shows that it is possible to perform a feature reduction without substantial loss in classification performance.

Important Links:

Go Back