Automated visual surveillance using kernel tricks
Tarem Ahmed, Xianglin Wei, Supriyo Ahmed, Al‐Sakib Khan Pathan · 2013
Extensive network of multimodal surveillance and security sensors prevalent in many places. Task of simultaneously monitoring multiple images tedious and monotonous for a human. Existing algorithms involve high complexities, need significant memory and storage resources, and typically involve custom equipment. We present three algorithms built using kernel machines to perform automated, real-time intruder detection in surveillance systems. Proposed algorithms are adaptive and portable, with computational, storage and memory complexities independent of time, making them naturally suited to online use.