engleski

Features analysis and Fuzzy-SVM classification for tracking players in water polo

This paper presents a novel approach for detection and tracking humans in water. Uniqueness of the tracked objects has been defined after analysis of standard color models. Based on the analysis results, YCbCr is proposed as the best color model for targeted application. Furthermore, relation of Cb and Cr components for different categories of targeted objects (object parts) were analyzed and used as features that can be used by classifier. Fuzzy-SVM classifier is proposed as the best solution for particular domain of problems. Unlike other Fuzzy-SVM methods, presented method is focused on fuzzy logic and applies binary SVM only in special situations when classification of input data is uncertain. In order to test and evaluate hypothesis, proposed method was compared to standard classification methods. Experimental results demonstrated validity and efficiency of the proposed approach.

Fuzzy SVM ; YCbCr ; classification ; water polo ; tracking player ; feature extraction

nije evidentirano