The Research And Implementationof Continuous Contour Extraction From Gesture In Video Images

Gesture contour extraction has a very important position in the field of computer vision,for example,gesture contour is the basis of gesture extraction and gesture 3D modeling.The Continuous and background-free gesture contour is an important prerequisite for gesturerecognition or modeling.This paper has analyzedthe research status of gesture contour extraction athome and abroad,and analyzesd the defectsof traditional methods.It is found that the existing methods are sensitive to background object,light intensity and skin color.These defects lead to the extracted gesture contour discontinuous,containing the contours of background objects.In order to solve the problems of gesture contours extracted by traditional algorithms,this paper has deeply studied the points of hand localization in image and connection of contour breakpoints.In order to remove the background object contour,this paper starts from the position of hand or finger to reduce the scope of the target contour search.To extract a continuous contour,this article starts from the analysis of the point where the curve of the slope of the point and the surrounding point condition to connect the breakpoints in the best way.In this paper,a localization algorithm bas ed on finger model,a localization algorithm based on wrist-palms model,GMM(Gaussian Mixture Model)localization algorithm,and contour breakpoints connection algorithm are studied and used.What have done in this paper areas following:(1)The localization algorithm based on finger model is an innovative algorithm in this paper.The algorithm uses the geometric features(smooth contour,a good symmetry by fingertips,fingertip bend angle)of the finger to search for the fingers in the rough contour and locate the fingers.As long as there are fingers in the image,the finger can be well identified,and the algorithm does not depend on the complexity of the background or light intensity.(2)The Wrist-Palm basedpositioning algorithm uses the modle which consistsof wrist and palm to search for wrist and palm,this algorithm has a good accuracy.As long as the wrist and palm present in the image together,they canbe well searched out.(3)In order to connect discontinuous contours,a new algorithm is proposed in this paper called continuous contour extraction algorithm.The algorithm classifies breakpoints and uses different strategies to connect breakpointsaccording to different situations.To verify the effectiveness of the algorithm,OpenCV(Open Source Computer Vision Library)is used to test and verify the system.And the result is compared with the traditional contour extraction algorithms such as Sobel operator,Canny operator and Longest contour extraction algorithm.The test and verification results show that the extracted gesture contoursare continuous and does not contain contours of background objects,and the algorithm has good robustness.