An implementation of a camera technique to obtain three-dimensional (3D) vision information for simple assembly tasks

摘要

Camera calibration is the problem of determining the elements that govern the relationship of transformation between the 2-D image that a camera perceives and the 3-D information of the imaged object. Camera calibration is important because it allows the determination of the relative rotation and translation between the lens centered coordinate system and an arbitrarily defined world coordinate system. For applications where visual information is required by the robot to perform tasks, the cameras can be calibrated with respect to the robot base coordinate system. After the camera is properly calibrated, its 2-D image coordinates can be translated into real world locations for the objects in its field of view. Thus, the 3D location of objects that are in the field of view of the cameras can be determined. This allows for the manipulation of these objects using the robot. Two different calibration schemes were implemented. The first is based on the work of Yakimovsky and Cunningham, at JPL. The second is based on the work of Roger Tsai, at IBM. In this report the implementation details of the techniques by Roger Tsai are presented.