RotaryEncorder

LandmarkDatabase

Vision-based tracking

　In MR-PreViz, rough camera path is known before MR-PreViz shooting. We shoot and target backgrounds and construct Landmark Database; LMDB that store 3D positions of feature points and viewpoint dependent information. (Rehearsal phase) In MR-PreViz shooting, CGI characters are superimposed geometrically correctly by matching landmarks in LMDB and feature points in target background. (Shooting phase) This method's advantages are performance and robustness in outdoor and indoor use because of matching landmarks by SIFT feature.