The Research And Implementation Of Motion Hand Detection Based On Video

With the development of science and technology and the increasing social demand,hand gesture recognition has a wide prospect in applications such as virtual reality,intelligent device control,robot control,medical diagnosis,computer aided manufacturing and so on.It has become the focus of human-computer interaction research.Generally,hand gesture recognition can be divided into gesture detection,gesture tracking,feature extraction and gesture recognition and other steps.As gesture detection and gesture tracking are the key steps and difficult points to realize hand gesture recognition,many scholars have studied the two steps.In this paper,by the research and analysis of the two steps,an algorithm is proposed by improving relevant algorithms and transplanted to Zedboard platform whose core is Zynq-7000.The work in the thesis aims to improve the accuracy of gesture segmentation and tracking.This thesis deeply studies and analyzes the related techniques about hand gesture detection and hand tracking.Firstly,the motion detection algorithms and skin color detection models used in gesture detection are analyzed theoretically and verified by experiments.According the comparison of the experiment result,the appropriate gesture detection is designed,which verified by the experiment later.Then the algorithm of gesture tracking is analyzed theoretically and experimentally.And combining with the requirement of embedded realization,the appropriate gesture tracking method is designed and verified experimentally.Finally,the video acquisition hardware and software platform are designed to realize and test the algogrithm proposed in the thesis.The mainstream algorithms of motion detection in the gesture detection are analyzed,and corresponding experiments are designed to compare their advantages and disadvantages.Then the fixed threshold segmentation,adaptive Otsu threshold segmentation and histogram model segmentation in skin color detection algorithm are analyzed.The fixed threshold segmentation and adaptive Otsu threshold segmentation experiments are designed,and the result of experiments are analyzed and compared.Combined mixed Gaussian model with Otsu adaptive threshold model,and added some hand geometric features according to the actual demand,the gesture detection algorithm is designed,and verified by the experiment.During the research of the hand tracking model,the Kalman filter tracking,colorbased particle filter tracking and Camshift tracking are analyzed theoretically,and the corresponding experiments are carried out to compare the advantages and disadvantages.Finally,in order to overcome the complex background,combined the Kalman tracking with Camshift tracking,the hand tracking algorithm is designed,and gesture track of the hand in the image that segmented by skin color is done to reduce the algorithm overhead.Finally,the corresponding experiment is designed to verify the feasibility of the algorithm.To realize the algorithm,in the thesis,a Zedboard evaluation borad platform is built and the OV7725 image sensor video acquisition logic module and video transmission module based on AXI-4 bus are designed.And the platform is carried with embedded Linux kernel,transplanted with computer vision library OpenCV and graphics interface library Qt.By using the method of cross compiling,the application of gesture detection and tracking algorithm in embedded platform is realized,which makes the system track the gestures accurately.