| Summary: | 碩士 === 國防大學理工學院 === 資訊工程碩士班 === 100 === The technique of augmented reality (AR) is to augment virtual objects into real images. To promote further applications, the interactions between a user and virtual objects are needed. Individual can interact with virtual objects using markers or data gloves for specific applications. However, if users want to select, translate, rotate, or scale virtual objects, using markers are not immediate perception, and using data gloves are costly and unpopular. In this thesis, we proposed a vision-based hand gesture recognition scheme to control 2D-3D virtual objects of AR easily, instinctively, and popularly. The proposed method consists of two major parts: the hand gesture recognition, and the AR presentation. In the hand gesture recognition, first, the hand region is extracted. Second, the features of hands such as the positions of finger points, the numbers of fingers, and the hand orientation are computed to recognize static hand gestures. Then, we integrate the state information and moving information of hands to calculate the dynamic hand variations such as the data of position, rotation, and scaling. Finally, these information can be transferred to the AR presentation modular to control the virtual objects. In the AR presentation modular, the 2D data are converted to 3D coordinates, and the changed virtual objects are rendered correctly. The experimental results show that the proposed approach can recognize dynamic hand gesture correctly and control the virtual objects in AR instinctively and in real time. Furthermore, the proposed scheme can be used in military applications such as moving symbols in war game system, hand command training for squad actions, or remote controlling mechanical arms for bomb disposal.
|
|---|
