关键词: 人群人数统计, 分割特征, 纹理特征, 多元线性回归, 高斯过程回归

Abstract: To solve the problems of the separation of segmentation characteristics and texture features in the crowd statistic, together with improving the accuracy in regression model, this paper proposes a new kind of method of crowd statistic based on PCA and multivariate statistical regression. The research takes the measure of PCA in order to reduce the dimension of the crowd prospect segmentation features and texture features which are extracted. This paper establishes a multiple linear regression model so as to determine the trend of the relationship between characteristic quantity and the number of crowd. The research modifies the Gaussian process regression model according to the trend. The experimental result shows that this method is more suitable for the statistic of large-scale crowd.

Key words: crowd counting, segmentation feature, texture feature, multiple linear regression, Gaussian process regression