A fast background subtraction method using kernel density estimation for people counting
Cao Jianzhao, Osuji Chukwunonso Victor, Odoom Manfred Gilbert, Changtao Wang · 2017
Background subtraction is the fundamental and essential step for the objects identifying from a video sequence in the vision system. Based on kernel density estimation, a fast background subtraction method is proposed and tested. In this method, the kernel density estimation background model is improved by LUT and early break method. Meanwhile, it is compared with single Gaussian model both for speed and effect. It has been tested in a 2.6GHz Intel Core machine with 25fps on 432×240 images without code optimization. Average processing speed is 18.9 and 34.6 million second per frame with 30 kernels and 60 kernels, respectively. The result shows that the improved method can be used in real-time processing and has better result than single Gaussian model.