| Today,a great number of security accidents are caused by the excessive flow of people.Once such accident occurs,the consequences would be severe.To prevent such accidents,we can count the number of person by video monitoring and limit the flow of people in time.This thesis focuses on the pedestrian detection and tracking at the vertical angle,and ultimately achieves the purpose of people counting.The main contents in this thesis are listed as follows:1)A pedestrian head detection algorithm is proposed based on multi-feature.Four head features are defined on the basis of depth images.According to the features,the detection of pedestrian head is achieved.The experiments’ results show that the proposed algorithm can effectively solve the head detection problem of complex scenes,such as hairstyle diversity and head part-occlusion.2)Based on concentric circles,a pedestrian head detection algorithm is proposed.On the basis of depth images,we do deep analysis on the head regions in depth images and find that the regions can be matched and segmented by concentric circles.And then using the mean-angel ray method to extract sequence pairs of concentric circles and to compute the distance of these pairs,thus to achieve the detection of pedestrian head.3)Based on the head feature points and the concentric circles,a fast pedestrian head detection algorithm is proposed.The computation complexity of the algorithm in 2)is high,which needs a lot of iterations for detecting head region.In order to improve the detection speed,a head feature point approach is proposed to filter most of pixels that belongs to non-head regions.By combining the proposed approach with concentric circles,the head region in depth images can be quickly detected.4)A statistic algorithm of the flow of pedestrians based on Kalman filter is proposed.Kalman filter is used to realize the accurate tracking of pedestrians.Then the trajectory of the head which is generated by using the Kalman filter technique is analyzed,and the judged strategies of the trajectory is summarized,at last the statistics of human traffic is implemented. |
