摘要：Crowd density estimation is very important in intelligent crowd monitoring. In this paper, a new approach of crowed density estimation is proposed. This method combines the advantage of pixel statistical feature and texture analysis, and reduces the impact of perspective distortion by dividing the region of interest. Moreover, we estimate the crowd size (the range of numbers) for high density and extremely high density crowd. The experiment results show that the proposed method has a relatively high accuracy in the whole density range.
关键词：Region of interest;Crowd density estimation;Crowd size estimation;Pixel statistical;GLCM
