Crowd Density Estimation Based on Improved Gaussian Mixture Model
Qinglong Chang · Computer and Digital Engineering · 2012
Crowd density estimation is very important to public security.To the question of crowd density estimation in video surveillance system,a crowd density estimation technique based on improved Gaussian mixture model and pixel statistics was proposed.The feature of Gaussian model was abstracted by calculating the mean and the mean of deviation in the image.Under the direction of the constant updating rate of the model,the crowd binary images were achieved by background reconstruction using Gaussian mixture model.Finally pixel statistics was used to realize the fast estimation of crowd density.Experimental results show that the new algorithm of crowd density estimation is more accurate and efficient than previous one.