Research On Palm Real-time Location And Gesture Recognition Method Based On Feature Extraction

With the rapid development of computer vision technology,through the mouse and keyboard as the input interface of the traditional interactive mode is already unable to meet people's needs,it should be the ideal mode of human-computer interaction that users directly with the computer to communicate with the natural human-computer interaction.Gesture recognition as people-centric more natural way of human-computer interaction,can satisfy the direct interaction between user and virtual environment,having a wide range of applications the machine control,smart home,body feeling game,sign language recognition and other fields.As a result,real-time gesture recognition technology has important academic value and application prospect.Existing research still exist some shortcomings: gesture recognition training template fixed gestures recognition method is difficult to meet the real-time requirements of interaction;a single gesture recognition method based on skin color feature extraction in gesture segmentation of the region prone to background region and area of overlap,which lead to the recognition rate is not high;gesture recognition method based on the peripheral sensors require users to wear data gloves and other equipment,which is expensive and bad for promotion.In this thesis,we mainly study the three parts of gesture recognition in gesture recognition,such as skin color segmentation,gesture feature extraction and digital gesture recognition.The main work is as follows:Firstly,in the segmentation stage of gesture area,the segmentation method of two cases without human face and human face under the common camera platform is studied.When the camera captures the image without the appearance of the face through the threshold of the skin color detector directly extract the gesture area;and when the camera image acquisition face,according to the face and hand skin color consistency principle,a new method is put forward to split the integration of face color detection gesture area.This method can obtain the range of pixel values of face region by face detection,and then combined with traditional threshold method,through the double threshold method to accurately segment the background and gesture area,you can to a certain extent,optimize the segmentation effect.Secondly,in the gesture feature extraction stage,the moving target image detection method is used to realize the motion gesture area detection based on the three-frame inter-frame difference method and the frame difference method.According to the algorithm,the gesture contours,gesture convex hulls and convex defects,fingertips and interphalangeal grooves are extracted.On this basis,a real-time palm detection and positioning method based on the minimum circumscribed circle,which has achieved an accurate and effective gesture palm positioning,The method,adapting to the gesture region for translation,rotation,and rollover,has good robustness to the recognition feature extraction under different scenarios.Finally,in the digital gesture recognition stage,according to the characteristics of fingertip to finger groove and palm,the recognition decision tree model is constructed,which realizes the quick and accurate identification of commonly used digital gestures.The real-time gesture recognition system based on VS 2013 integrated development platform verifies the effectiveness of the above method.