- 清水彰一, 藤吉弘亘
- 計測自動制御学会システムインテグレーション部門, p. 179, 2004
Download: PDF (Japanese)
We present a method for calculating 3D positions of objects using two unsynchronized cameras (30 fps) mounted on a mobile robot. If robotns motion and orientation are known, a viewing ray of each camera passing through the point of the object on the image is calculated. Then, a curved surface passing through the viewing rays of previous three frames is generated, and the 3D position is calculated as an intersection point between the ray and the curved surface. 3D position of last frame is calculated by linear prediction using previous two frames, and the 3D position is calculated once more using constraint by viewing ray information in 3D space obtained from the last image. Since the 3D position is calculated at every shutter timing of each cameras, the 3D position can be obtained at 60 fps using only two unsynchronized cameras.