Statistics of Pedestrian Flows Based on Feature Matching in Video Sequence
Li Zha · Video Engineering · 2011
The information of the pedestrian flows has great significance in many areas,such as traffic control,public places,business analysis.According to the shortages of conventional methods,a statistics of pedestrian flows method based on feature matching is presented in this paper.This method consists of two stages.The first stage is collecting some representative samples which consist of feature vector and the corresponding pedestrian number,and then classify all samples according to pedestrian number.The second stage is to count the pedestrian: the feature vector of a motion video is extracted,which is then compared with each class of samples based on Mahalanobis distance feature matching to estimate the optimal number of pedestrians.The experimental results show that the method has high detection probability and satisfies the requirement of real-time application.