Human Identification using Histogram of Oriented Gradients (HOG) and Non-Maximum Suppression (NMS) for ATM Video Surveillance

Today video surveillance is the scorching topic in the research field of Computer Vision. Around 400 million surveillance cameras are used in various sectors which merely act as blind and record videos for post-incident analysis. With the increase in crimes around the world, video surveillance plays a key role in identifying anomaly activities in a day to day application. One such application is a robbery in Automated Teller Machine (ATM). Identifying and tracking unlawful human activities is a challenging task in the ATM video surveillance system. Thus, an effective Human detection method using computer vision and image processing is proposed to create phenomenal results. The proposed approach, incorporate the most acclaimed Histogram of Oriented Gradient (HOG)-Support Vector Machine (SVM) with a combination of Adaptive Background Generation Model and Non-Maximum Suppression (NMS) algorithm to detect human with appropriate results in a video sequence. The experimentation result of the proposed method is applied on ATM Video Surveillance, real-time dataset. The performance analysis is evaluated by considering the average value of metrics such as the number of an input frames, ground truth elements and frames with human identified which acquired 97% of accuracy.