- 學位論文
利用影像視覺自動校正非線性表面投影扭曲
Automatic Image Distortion Correction for Projection on Nonlinear Surfaces
摘要
由於投影技術的蓬勃發展，曲面屏幕被廣泛應用在娛樂系統、環場電影、虛擬實境等等的產業上。但是投影在曲面屏幕上容易造成投影影像扭曲變形，觀賞者會看到的非線性失真的現象。目前因為校正影像的技術有限，需耗費龐大的人力資源且校正的彈性較差；因此，本實驗將攝影機當作觀賞者、投影機投影的畫面分別定義出不同的影像空間；前者為攝影機影像空間、後者為投影影像空間，藉由數位影像處理，找出這兩個空間相互對應的網格座標。另外再利用虛擬場景定義出修正的目標點，由電腦自動辨識出特徵圖案，並且找到正確的目標點，逐步建立出曲面的校正網格。 本研究以汽車擋風玻璃作為投影屏幕，因為擋風玻璃是一個非線性彎曲的任意曲面，投影畫面會因此而扭曲變形。為了使觀賞者在駕駛時能感受到身歷其境一般的真實感，將虛擬影像場景投射到擋風玻璃上，形成一座沉浸式投影的虛擬駕駛系統。利用本研究開發的曲面校正系統，將擋風玻璃上因投影而扭曲變形的影像校正，才能使虛擬影像栩栩如生。 以上校正的程序可以被應用在任意曲面的投影，因為以電腦作為開發的基礎，並且自動完成影像的校正，因此，增加了曲面校正的效率、降低了使用人力校正時的困難，可應用在許多大型的場合，提供技術的便利性與實用性。
並列摘要
In this study, the virtual scene projected through the projector to the windshield of a car has been corrected by automatic image vision processing to create a surround effect so that the driver will be immersed in a vivid virtual reality. Curved screen projection technology is widely used in the entertainment system, surround movies, virtual reality and then industry. Projection on a curved screen always to produce nonlinear distortion correction which requires tremendous human resources to correct and do not have much flexibility of operation. Therefore, this study analyzed the relationship between screen virtual scenes and real cameras to identify the projected image space corresponding calibration marks, and use the digital image processing technology, such as Canny edge detection and Hough transform, identify each feature marks that needed to corrected to automatically. By using computer instead of a human operator, the present study can save resources as well as to improve the correction efficiency and accuracy.