Projective reconstruction for robot vision system

Intelligent robot control systems usually require the knowledge about the surrounding environment in order to achieve difficult, high level tasks, such as path planning or collision avoidance. Usually visual information provided by uncalibrated cameras is used to reconstruct the scene. This paper proposes two new resection-intersection type methods that can be used to recover 3D projective structure and camera matrices from multiple images taken about the scene. The only inputs of the algorithm are 2D feature sets (points, lines) extracted from raw camera images