多台深度攝影機的座標對齊校正,在國外有許多座標對齊校正研究已研究出許多種方法,大致上以棋盤格板的校正方式做為基礎,也有利用在空間中放置立體的物體來進行座標的對齊校正,然而在有些限制的環境條件下,深度攝影機所擷取到的影像在分析影像上無法獲得到要進行對齊校正用的物體,導致不利於這些方法的校正。 因此本研究採用基於空間中拋擲的球體,進行拋體軌跡上時間與空間的關係分析,令多台深度攝影機在空間座標中,也能夠確認彼此深度攝影機的相對位置,再基於本研究的多台深度攝影機對齊校正方法,校正多台深度攝影機的座標系統後,蒐集每一台深度攝影機在三維空間中的立體影像,並且利用體素化、統計去除離群點、法向量表面條滑處理等,建立一個在多台深度攝影機的場景中的前景三維模型,最後討論其在建模上的應用與其準確性。
It’s importance on multi-camera coordinate alignment. Much researchers provided lots methods based on chessboard. But in some bad environment, such as weak or bright light situation. It can get wrong calibration result. Because of chessboard cannot analysis in the worst color image, we provide a new way to solve it. In this master thesis, it provide a terrific method that using a ball and throw it into the air. Our system will analysis the ball’s trajectory in the air on each kinect camera, and alignment each camera’s trajectory. Above method is purposed to align each camera’s coordinate. After camera’s coordinate aligned, we generate a model based on this method, and analysis model’s error.