Abstract—People counting is an important part of video surveillance. In recent years, significant progress has been made in the field using the method of feature regression. In this context, feature extraction and using a machine learning algorithm to establish the relationship of extracted feature and the number of people are two basic steps. To extract the feature of crowd, methods in the literature either using the statistics values of the foreground pixels, or using the number of corners. In this paper, in order to obtain a better description of crowd, both of the two kinds of features are obtained respectively by using the FAST algorithm and VIBE algorithm, and a processing of normalization is done to solve the problem of perspective distortion. Then, the correspondence between these features and the number of people is studied by SVR. In addition, in order to avoid the improper selection of parameters of SVR, the PSO algorithm is used to select the relevant parameters in SVR. The method has been tested on the PETS2009 datasets and the self-shooting datasets, and the experimental results show the effectiveness of the method. And, the method has been extensively compared with the algorithm by Albiol et al, which provided the highest performance at the PETS 2009 contest on people counting. The results confirm that the proposed method improves the accuracy and robustness.
 

Index Terms—FAST, VIBE, PSO, SVR, People counting

Cite: Jiaojiao Yuan, Hong Bao, Haitao Lou, and Cheng Xu, "A Method for People Counting Using Low-level Features Based on SVR with PSO Optimization," Journal of Communications, vol. 12, no. 11, pp.  617-622, 2017. Doi: 10.12720/jcm.12.11.617-622.