スケール変化に対応した人物動作判別

Abstract

This paper proposes a scale invariant identification scheme of human behavior, so that distinguishing abnormal behavior to normal behavior. This scheme makes use of CHLAC(Cubic Higher order local auto correlation), which was proposed by Otsu[5], as the feature quantities to describe moving behavior of human. In order to cope with change of human body scale in input moving images, our proposed scheme adopts an average processing for scale variation of human in feature quantities calculation. Computer simulation using moving images of human is performed with the proposed scheme. The experimental results showed that the proposed scheme has a better performance than these conventional schemes in detection of abnormal human behaviors.