Human Behavior Tracking with Vision Heatmap Toward Smart Building System

Abstract

Vision system has broad applications in diverse areas. One of such areas is the increasing interest of systematic monitoring of human movements in certain space, in restricted areas or even in open space such as on the shop floor or even on a certain path. Non-vision methods suffer from many inherent weaknesses, but vision methods demand more resources. The focus of this research is to produce a heatmap system that shows human traffic behavior using camera vision with real-time image processing on a low-cost miniPC or laptop in a typical smart building, where relevant parameters are closely monitored. A static background image is used as the base image and is compared with current video frames to give information about changes that happen in the specified area. These changes are processed further to obtain the path of human movement. Changes that happen in the path will change the RGB value of the path. The change of RGB values translates into color gradation. The color gradation result is added to the static reference background image to produce a real-time heatmap with changing cumulative color gradation. Busy areas will produce a brighter heatmap. The heatmap system is processed successfully both on a miniPC (Raspberry Pi 2 Model B, Cubie, Odroid) and on a laptop. The system also would add vision analytic values to a simple surveillance or CCTV system with low cost.