Developing stereovision and 3D modelling for LHD automation

This paper presents current work on the development and use of stereovision for underground mining. Information on the calibration of the cameras and the feature-matching algorithm are presented. This system uses an intensity-based approach that evaluated correspondences along the epipolar line. Once features have been identified and their distance from the cameras computed, a three-dimensional map of the muck pile is developed. The approach used here uses a ray-tracing algorithm. This approach provides an uncertainty measure along with the elevation. A discussion of the use of this modelling information in the control of the LHD is included. Keywords: Robotics, automation, LHD, underground mining, vision sensing, task automation, tele-operation