A smart surveillance system with multiple people detection, tracking, and behavior analysis
Chia‐Jui Yang, Ting Chou, Fong-An Chang, Ssu-Yuan Chang, Jiun-In Guo · 2016
This paper proposes an intelligence surveillance system in indoor environments, which support the functions of people detection, people tracking, and behavior analysis. Strong variation of lightness by switching lights and frequent crossing of people are two major design challenges of the proposed system, which will decrease the detection accuracy. Therefore, we propose a mechanism of updating background to react to the variation of lightness. Moreover, the ORB descriptor is added to be a unique feature of identifying people. It can make people tracking more precise during crossing. The proposed system is developed on PCs and implemented on embedded systems. On PCs with Intel [email protected] CPU running with a single core, we can reach the performance about D1 video at 22 fps. On an embedded system with ARM [email protected], we can achieve the performance about D1 video at 8 fps.