Fast correspondence search algorithm for stereo vision range sensors

Perception of the environment in 3D has always been an important sensory input for machine vision applications. Stereo vision range sensors have been used for several decades to obtain 3D information of the scene. Still, the major issue involved in the stereo analysis of images is the correspondence problem. In this paper we consider the correspondence problem only for point clouds. The algorithm proposed is intended to solve the problem of the correlation of two point clouds. The clouds may have any rotation and translation to one another. Also a different amount of points within the two clouds are not a problem for the proposed algorithm. To solve the problem the graph theory is used. Simulations results are presented and at the end of the paper the application for a stereo vision 3D camera are demonstrated.