Crowd density estimation based on the normalized number of foreground pixels in infrared images
Guozhong Liu, Tianze Wang, Zheng Cao · 2013
Crowd density estimation in public scene surveillance is an important issue of public security. Compared with the visible light images, infrared images have the inherent characteristics, such as the low contrast, low signal-to-noise ratio, have made it a huge challenge for human detection and reliable crowd density estimation. In this paper, a human detection algorithm in infrared image sequences based on image subtraction , histogram analysis and morphology processing is proposed to remove background effects. The number of people is estimated quantitatively by the pixel normalized statistics method, in which the foreground pixels are counted with different weighted factors. Experimental results show that this method is simple, effective and can improve the accuracy of estimation.