An Improved Passive Tracking System for Automated Person of Interest (POI) Localization with SVM based Face detection
Output of every security camera is recorded by video recorders. For a crime, criminal or suspect can be located by using these video footage as evidence. Considerable amount of time and manpower is required for scanning those footages. So, it is difficult to track Person Of Interest (POI). In this research, proposed automated system for computing desired POI from available volume of data accurately.
Restricted Boltzmann Machine is used for detecting POI with facial recognition, which includes rules of deep belief and deep learning network. Video montage is created for all desired frames. POI path is tracked by incorporating location and time information. It reduces time, human error and burden of human in detecting POI.
On different data collected, perform validation. POI are identified correctly by proposed system as shown by results. In a constrained environment, around 86% of accurate results are produced by proposed system. Around 80% of accurate results are produced on data collected by cell phone. POI can be effectively tracked by proposed system on Real video tested in the university campus.