Pattern recognition by active visual information processing in birds

Methodological design issues in pattern recognition are of particular interests to many scientific disciplines ranging from life science to engineering. Practical applications of pattern or object recognition methods are numerous but still encountering many problems including the inherent difficulty in computerized feature extraction and classification. This paper proposes a strategy for object recognition resembling the active template matching strategy in chickens. Experimental results on several databases suggest that using the active vision processing can improve classification rates implemented with various classifiers.