The Application Research Of OpenCV On Real-time Gesture Recognition

Along with science technology growing fast, people’s lives are entering the era of cloud computing with a flag of more and more intelligent devices are emerging which changes the human-computer interaction(HCI) mode deeply. In computer vision field, HCI is always a research hotspot with a purpose of promoting the intelligence, convenience, security when people communicate with machines. Many developments have witness such HCI evolvement like the changes from wire transmission to wireless, from manual control to intelligent control, from scanning recognition to automatic recognition.The thesis mainly researches the movement trace recognition of hands’ gesture in the condition of monocular vision, with the purpose of realizing the mouse’s virtualized operation by gesture. By taking advantages of the flexibility of human hands, this paper focus on the hands’ movement tracing and recognition, from theory to practice. In the theory part, the first step is to discuss the computer vision field research achievement, as well as its significance; secondly, the basic theory research on the gesture recognition gives the way to support the thesis’ s mouse virtual operation application. In the practice phase, the gesture image preprocessing and recognition are performed by experiment platform. And finally realize the mouse virtual operation.The essence of real-time gesture is a combination of every frame’s gesture in the time series. As a result, the gesture trace recognition is turned into trace the gesture in every frame. In the preprocessing of gesture image, the paper uses image segmentation based on the skin color, in order to take the gesture apart from the background, followed by smoothing process. In the skin segmentation phase, color modeling method is introduced, as HSV model was chose. In smoothing step, the Gaussian filter performs the task to eliminate the noise and make the gesture clear.The gesture movement recognition is performed by Camshift algorithm in the real-time gesture recognition application research part. The DTW dynamic temporal planning algorithm is chose to perform gesture recognition. As a result, the whole method has a good stability. At last, by calling the mouse controlling API with the gesture recognition result, the mouse operation like move up and down, move left and right, single-click can be realized by human hands’ gesture and movement.