Human Behavior Classification Using Multiple Views

We present our current results in understanding the behavior of humans moving an a plane using multiple cameras. We exploit the merits of camera registration based on homography estimation to extract position on a 2D map. We use the output of HMM and SVM classifiers to model and extract human behaviour from both the target trajectory and the target short term activity (i.e., walking, running, abrupt motion etc). The proposed approach is verified experimentally in an indoor environment. The current results with a single moving agent are presented.