Abstract

It's very important to realize global localization in order to operate the mining devices underground without product accidents and improve the efficiency in the research of Digital Mine. A monocular vision-based global localization way for unmanned underground mining vehicle is presented. This algorithm is established on the parallel perspective mapping model constrained by the sight direction of the vehicle-mounted camera. A world coordinate system of the laneway and the landmark sequence-world coordinate mapping is required to be established. The system gathers environmental image through the vehicle-mounted CCD camera, after a serial of pre-processes of image. The system gets sequence information of landmarks via recognizing the 1D barcode and searches the unique world coordinate by sequence. Using the localization algorithm, the system computes the position of the camera and finally finds mining vehicle's real-time position in world coordinate system. The experimental results indicate that this method is low-cost, has strong environment adaption and high localization accuracy and is suitable for the requirement of underground mining work.

Paper Details