擴增實境(Augmented Reality, AR)是以實景影像作為背景,將虛擬 物件之數位影像用標籤(Marker)做引導,以合理之比例及方位融合出 現於實景畫面上的一種科技。由於標籤的辨識含有方向與位置,且標 籤所代表的虛擬物體無形體之限制,相較於Kinect 係以概略輪廓定 義人形物體,且使用者必須盡量正對攝影機以避免動作識別失效, Kinect 允許的身形差異較小。 本研究嘗試以穿戴標籤之方式來追蹤大肢體,再由系統去理解肢 體之動作,再進一步利用人體肢體與虛擬物體互動。如此將可使互動 者之活動角度與活動範圍更大。 實作部份將使用擴增實境編輯軟體(D’Fusion)來擷取使用者之大 肢體(Large Motor)動作,並探討標籤設計、穿戴位置及動作校正對於 幾種大肢體動作識別之效果。
In Augmented Reality (AR), image of the real scene is used as the background, and the images of the virtual objects are guided by markers and overlapped on top of the background in designated size, position and orientation to make a sensible fusion. The marker will be interpreted as position and orientation, and a 3D image of any shape. The virtual 3D object can then interact with other virtual objects as requested. On the other hand, the interaction method of Kinect requires user to face the camera as normal as possible, such that the silhouette of a human can be recognized. The nature of Kinect will restrict user in position and orientation. In this research, markers are worn on specific positions of limbs, and the large motor motions are reasoned and shown on a virtual dummy. The AR driven dummy may then represent the human user to interact in the virtual environment. It is expected that AR driven motion have larger range in space and orientation than Kinect. In practice, the AR editor D’Fusion is adopted to capture and interpreted marker position and orientation, and thus, the large motor motions. Further analyses are conducted on marker design, wearing position and data calibrations using some example motions to make the method more applicable.