Static Gesture Recognition System Based On OpenCV

Hand gesture recognition is a new technique which aims to explain human hand or gesture by pattern recognition.Gesture recognition system has the characteristics of convenient installation,low cost and good user experience.Therefore,gesture recognition technology has very important research significance in the application and theory.On the basis of OpenCV,this paper classifies and classifies the most common 6 static gestures.1.Collecting and preprocessing images.We collect video streams from single or multiple cameras and save them in AVI format as input images,and invoke OpenCV library function to preprocess the input image,which in order to make prepare for gesture segmentation and feature extraction.To eliminate the irrelevant information and keep the key information of the image,this paper uses Median Filter and Laplacian Operator to smooth and sharpen the image.Because of the strong correlation between R,G and B in RGB color space,the image is transformed from RGB space to YCrCb space when skin segmentation.When the image is processed by binarization-processing,the OTSU algorithm is used to find the adaptive threshold in order to separate the skin areas.Then,according to the judging condition of the gesture area,the other non gesture skin regions are eliminated,and the gesture region is extracted from the image.Next,morphological processing is used to effectively remove the "black hole" and "spike" in the gesture images,so that the extracted gesture images are more complete.2.Extracting gesture feature vectors.We select the first 4 moments of Hu moment invariants as the characteristic parameters of the gesture image,obtain the number of fingertip by the method of finding curvature and convex hull,which forms the feature vector of gesture.3.Classing and recognizing gestures and testing the system.Firstly,the standard template gesture library is built by analyzing the feature vectors of gesture samples.Then,the library is trained to generate the classifier.Later,using the classifier to recognize and classify the test set.In this paper,we use template matching method to recognize hand gesture.The results show that the recognition rate of the most commonly used 6 static hand gestures is up to 95.5%.