DETECTING ABNORMAL ACTIVITIES FROM INPUT VIDEOS AND REPORTING TO AUTHORITIES.

Computer Engineering

In the context of pattern recognition, abnormal event detection plays an important role. Anomaly detection is an active area of research on its own. Previously known anomaly detection techniques are usually not object based, where objects are not recognized prominently. In this system we find optimized object aware anomaly detection technique, based on certain object categories focusing on mobile objects. Algorithm used performs block based foreground segmentation to restrict our analysis to moving objects and unrelated background dynamics. Object detector is used to discard unrelated objects on connected blocks. Histograms of block-motion trajectories are extracted and cluster them to represent normal events. This framework gives a relatively low computational complexity and high detection accuracy.

This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.