基于共生与非共生梯度直方图的人体检测-信号与信息处理专业论文.docxVIP

ABSTRACT ln 由is paper，we mainly consider the problem of human detection in the 阳tic images. 叽le goal of human detection ?s to determine the presence of humans and the positions of the humans. Human detection is crucial to image/video巾，ased intelIigent surveillance and thus is important for social security. 电 Two key points of the human detection are fcωωreex衍action and the designing of 伽e classifier.ln this paper，we focus on the first one，i.e. fc饵ture ex归功ction. One ofthe 4 most successful features for human detection is Histograms of Oriented Gradients (HOG)，which was propos创 by Dalal 创.al.in 2005. For improving the 叩eed ofthe human detection and the efficiency of the classifi町 Wang et.al.伊。posedω ∞mpute HOG features by ωientation decomposition and convolution mask at International Conferenω on Computer Vision (ICCV) in 2009. For enhancing the 肌pressiveness and discernment of HOG， Watanabe et. al. proposed Co-occurrence Histograms of Oriented Gradients (Co-HOG) in 2009. In this paper，based on the works above，we proposed three novel methods for improving the performance ofhuman detection which are listed as follows: To improve the performance of the method proposed by Wang et. 此， we proposed to employ mean filter for smωthing the gradients. Compared ω Wangs filter which is lin创.r to distance，the proposed method results in better detection performance. To further improve the per岛Irmanωof the method proposed by Wang et. al， proposedωenlarge the size of filter from 7 X 7 to 9 X 9. Existing Co-HOG uses merely the information ofthe direction ofthe gradient and thus the information ofthe 阴dient magnitude is neglected. 丁0 make full 问 use of the information of the gradients，we propose to use not only the 句 information ofthe direction but also the magnitude ofthe gradients. 丁0 sum up，we have implemented several classical human detection methods and proposed several novel methods to improve the per岛rmance of the human detection systems. KEY WORD