Human Body Detection And Sepration In Video Stream

The detection and separation of human body in video stream is becoming a hotspottopic in the domain of computer vision reseach. This strong interest is driven by a widespectrum of promising applications in areas such as smart surveillance, image storageand retrieval, etc. This paper has done some research on detection and separation ofmultiply human bodys individual. The mainly research work is to capture completeforeground, extract human body features, and recognize normal human body by patternrecognition from single video image.In human body detection part, before detecting moving objects, we need“Background Modeling” to the visual scenes. The binary maps are captured from thebinaryzation to Background Subtraction process. Unfortunately the binary maps maybehave shadows while the objects are moving, so we must move these shadows from them,and we can get complete contours behind median filter and canny algorithm, then weextract human body features. As far as the human body feature exaction is concerned,traditional algorithm is of much computation complexity.A pixel-based featureextraction algorithm of human shape for classification is proposed in this paper, whichcan effectively handle occlusion problem. In recognition part, this paper is proposed twokinds of classification problem based on human body and nonhuman body, which areidentified by SVM.In particular the separation of human body, firstly we extract the human bodySURF feature points from video sequences, and separate these feature points fromdifferent human bodys, then make use of ZNCC to match SURF feature points in thetwo continuous images. Lastly we only need diffuse these feature points.The experiments have been demonstrated through a series of tests with videosequences gathered from both indoor and outdoor with help of VisionLab platform,showing great practicability.