Enhanced people counting system based head-shoulder detection in dense crowd scenario

Abstract

Counting precisely the number of people in a crowd is one of the most attractive issues for video analytics application. In this paper, an integrated method using Histogram of Oriented Gradient (HOG) and Completed Local Binary Pattern (CLBP) is proposed to detect a head-shoulder region of people within image or video sequence. Head-shoulder region is used as features to detect people against the false positive and false negative issue. HOG and CLBP are used to extract the edge contour and texture features of head-shoulder region, respectively. The two features are fused together to generate a combined feature vector. Support Vector Machine (SVM) is used to execute classification of the fusion features to classify people from a mixture of objects. The results show that the detection rate of the proposed method HOG-CLBP, on Recall value and Accuracy, achieves better performance compared to the current method for dense crowd scenario.