Multi-algorithms based visual surveillance system for human detection
Rishav Garg, Gyanendra K. Verma · 2016
This paper presents a robust and efficient automatic visual surveillance system to detect presence of human being in restricted zones that are off limit i.e. military installations, secure depository, border areas etc. The system is based on multi-algorithms namely HAAR wavelet (HAAR), Local Binary Pattern (LBP) and Histogram of Oriented Gradients (HOG) and capable to tackle the following challenges: 1) invariant to shape of object (human being), 2) illumination changes, 3) orientation 4) partial occlusion and 5) shadowing effect. The experiments are being performed on in-house high definition videos under above different conditions and the performance are shown in terms of TPR (true positive rate), FPR (false positive rate) and computation time. The experimental results clearly demonstrate the efficiency of our system with comparative evaluation of above three algorithms on KTH and Weizmann database.