Human activity recognition for automatic visual surveillance of wide areas
Marco Leo, Tiziana D’Orazio, Paolo Spagnolo · 2004
The problem of automatic recognition of human activities is among the most important and challenging open areas of research in Computer Vision. This paper presents a methodology to automatically recognize the human activities embedded in video sequences acquired in outdoor environments by a single large view camera. The activity recognition process is performed in three steps: first of all the binary shape of moving people are segmented, then the human body posture is estimated frame by frame and finally, for each activity to be recognized, a temporal model of the detected postures is generated by Discrete Hidden Markov Models. The system has been tested on image sequences acquired in a real archaeological site meanwhile actors perform both legal and illegal actions. Four kinds of activities have been automatically classified with high percentage of correct decisions. Time performance tests are very encouraging for using the proposed method in real time applications.