Research On Dynamic Gesture Recognition And Its Application In HUD

As the traffic is becoming more and more convenient,demand of people becomes diversified.At the same time,many car companies began to focus on researching which provide with safe driving methods,such as HUD on the car.The HUD reflect the information on the windshield by instrument which using the reflection theorem in the optics.The way of control the instrument normally including touching with hands,voice or gesture.In order to keep the drivers' attention on the road condition,so the way of control HUD is important which must be a flexible and non-touch mode.Aiming at this problem,in this paper,it puts forward a way of non-touch gesture using different kinds of gesture while control the HUD which is flexible and convenient,extracted by hand area algorithm skin probability map based on contour curvature,and by the least squares fitting motion region centroid estimation algorithms which are based on the recognition of the moving gesture.The process of gesture control containing several steps,for first it uses camera to capture real-time gesture image sequences,and then uses computer vision algorithm of gesture segmentation,feature extraction,gesture recognition,and finally the system is transplanted to the Android platform,and realize the dynamic gesture control HUD.In this article,according to the technical analysis and research of dynamic gesture recognition,the work the work is mainly divided into the following several aspects:Firstly,this paper introduces in detail the image acquisition equipment and image preprocessing methods,including denoising,smoothing,morphological processing,binaryzation and color space conversion operations.Secondly,his paper focus on the details of the algorithm of skin probability map based on contour curvature,using the principle of division of skin region clusters of the skin and non skin regions,and calculate the probability of interference,then use difference between hand and facial contour curvature characteristics to remove the face in skin region,and get the hand area.Thirdly,the paper emphases the details of the least squares fitting centroid motion estimation algorithm based on the centroid algorithm to obtain from the hand's area after segmentation,and the centroid trajectory formation were fitted by the least square method,linear motion trajectory,the classification and recognition.Forthly,the system is transplanted to the Android platform to control the HUD system,to realize interactive function.In this paper,we realize the real-time identification of dynamic gestures,and apply it on the HUD platform of Android.The experimental results show that the algorithm uses color probability map based on contour curvature can effectively extract the hand area,and based on least squares fitting motion region centroid estimation algorithm can achieve a good recognition rate.The system has a good application prospect in the interaction of vehicle device.