Font Size: a A A

Face Detection In Complex Background

Posted on:2012-11-17Degree:MasterType:Thesis
Country:ChinaCandidate:H LiaoFull Text:PDF
GTID:2178330335459816Subject:Communication and Information System
Abstract/Summary:PDF Full Text Request
Face detection in complex background is a difficult problem in the fields of pattern classification and computer vision, and plays an important role in face recognition, video retrieval, intelligent monitoring and other fields. Recently, with the wide application of video conference, surveillance monitoring and other technologies, face detection in complex background has becoming a more challenging and hotter research subject. Some research work focusing on the subject of face detection in complex background is discussed in this thesis.The main work includes:1. In order to solve the problem of non-skin interference and disconnection of skin region, this paper proposed a discrimination method of candidate region based on skin similarity integration. Experimental results show better robustness to illumination and pose variation. Meanwhile the process speed of this method is fast, the average preprocess time is 3ms for the image size 300x400 using a computer of CPU1.76GHz.2. As there are many face-alike regions in complex background images, compared to simple background images of the same size, its processing speed is slower, and the false alarms are higher. In order to solve this problem, a three-stage face detection system combining of skin preprocess, RealBoost detection and SVM post-process is proposed. Experimental results show the effectiveness and accuracy of the system.3. In the face tracking application, an improved algorithm of corners based optical flow tracking is proposed. In order to solve the problem of interference brought by the incorrect detected corner around the face in the returned face rectangle region from the detection step, this paper proposed a face extraction method based on the maximum histogram range to extract the corner from images. Experimental results show that it can locate faces in video sequence with complex background accurately, meanwhile the better robustness to illumination,pose variation and occlusion disturbances, and faster detection speed are obtained at the same time.
Keywords/Search Tags:face detection, complex background, skin preprocess, RealBoost, SVM, optical flow tracking
PDF Full Text Request
Related items