期刊名称：International Journal of Multimedia and Ubiquitous Engineering
印刷版ISSN：1975-0080
出版年度：2015
卷号：10
期号：12
页码：217-224
DOI：10.14257/ijmue.2015.10.12.22
出版社：SERSC
摘要：Human body detection in video surveillance is an important research point, which has been devoted for many years by researchers. But it can't detect human body with the feature of HoG and classifier of SVM. This paper proposes an improvement measure to the traditional algorithm of human body detection in order to detect human body in video surveillance in real time. Through the Integral images to speed up the computation of features of HoG and HoD under the condition of Kinect, It can ' t detect human body in video surveillance in real time. As human body detection using HoG features can't meet the real-time requirement in the current time. This paper abandon the time-consuming processing of computing features of HoG with three line interpolation and Gauss filter in terms of traditional algorithm of human body detection, and This paper introduces the notion of Integral images to speed up the computation of the features of HoG and HoD with the help of Kinect. As a result, Our improved algorithm can get a better time performance than ever, which can get the average detection rate of 30~40 FPS, current performance getting promotion greatly than ever, under the condition of CPU 3.10Ghz, RAM 4 GB, 640*480 detection window.
关键词：Computer Vision; Video Surveillance; Detection of Human Body; Kinect; ; HoG; HoD
