Multi-agent event detection system using k-nearest neighbor classifier
Nor Surayahani Suriani, Aini Hussain, Mohd Asyraf Zulkifley · 2014
This paper affords a method for the automatic multi-agent event recognition (M-AER) of snatch theft event. M-AER is known to be difficult since parameters such as motion, background, appearance, illumination etc. are constantly changing. In this work, motion vector flow (MVF) and directional motion histogram (DMH) that capture the interaction between two persons in a video sequence are being proposed as input of the M-AER. Assuring result of 90% accuracy has been achieved for the snatch theft detection.