| Summary: | 碩士 === 國立清華大學 === 資訊工程學系 === 102 === In the class of perceptive user interfaces (PUIs), gesture recognition with fingertips tracking is one of the natural ways of making the computer aware of what the user is doing. The aim of this thesis is to propose a method for detecting fingertip locations in stereo images and tracking fingertip trajectories across image frames. For demonstrating the quality performance, the applications of writing-in-the-air with stretched index finger is given.
The proposed method is working on both color images and the corresponding disparity maps using a designated stereo matching algorithm aware of the color and depth features of hands simultaneously. Within matching, the color and structure information are combined first to be the matching costs of left and right image taken from twin-lens cameras, and the concept of full-image aggregation is used to obtain the disparity maps. Then, the hand region is segmented and the palm center is located with the help of depth information. For tracking the position of user’s fingertip, two detection procedures are employed for two cases of gestures (pointing to the camera or pointing to the side). Finally, angular restriction and position refinement of fingertips are applied to enhance the accuracy. The experimental results demonstrate that our method performs well for both indoor and outdoor tracking at a speed of about 11.29 frames per second for the 640×480 images. Besides, its capabilities are further demonstrated in the application of writing-in-the-air.
|
|---|
