Human Motion Recognition System Based On Characteristics Of Movement And Shape

With the rapid development of computer vision technology,human motion recognition technology has occupied an important position in this field.It has important practical value and research value in security protection,advanced human-computer interaction,video search analysis,and sports analysis.Due to the characteristics of human motion such as non-rigid bodies,illumination changes,complex backgrounds,the human motion recognition becoms enormous challenges.The basic human motion recognition system consists of three parts: the moving target detection,feature extraction and human motion recognition.(1)In the moving target detection,Explain and compare the basic methods of target detection: the advantages and disadvantages of the optical flow method,the inter-frame difference method,and the background difference method,and the applicable range.The background difference method was used to detect moving targets.Then Gaussian background modeling was analyzed and discussed.An improved Gaussian modeling method was proposed to adaptively select Gaussian distribution numbers.Experimental results show that the model can accurately extract moving targets.(2)In the feature extraction,based on the research of a large number of papers,this paper extracts three kinds of motion characteristics,including the target's Hu moment,velocity and centroid,two kinds of shape characteristics,including tightness,and rectangular shape,width and height.These characteristics were tested in the Weizmann video database.The results show that these features can effectively describe human movements.(3)In the recognition of human motion,Firstly,the principle of support vector machine,the optimal hyperplane,the types of kernel functions and multi-class support vector machine algorithms are introduced.In this paper,the support vector machine(SVM)is used as the classifier for classification recognition.The SVM classifier is trained with the training set,and the radial basis kernel function in the LibSVM library is used as the parameter of the SVM.Finally,the test set is input into the trained classifier for action recognition.This thesis uses the Visual Studio 2013 platform,and the system based on the OpenCV computer vision library and LibSVM library.The accuracy and real-time of the system are tested using multiple-segment videos in the Weizmann video database.Experimental results show that the system can accurately describe human movements.