This paper addresses the computational aspects and visualization tools required for the automatic visual inspection of vertical deep mine shafts. The goal is to achieve 1 mm pixel resolution in shafts with depths of up to 1 km and circumferences of approximately 30 m. Hyper resolution images, containing approximately 10^11 pixels, with tiled visualization has been chosen as the solution path. The process starts with the acquisition of a large number of images, typically 50 000 - 70 000, using a special device lowered into the shaft. The work presented here starts with these images and their associated meta data. The meta data includes time, orientation and approximated depth for each image. Eight images are provided at each depth providing a 360 ° view of the shaft. The individual images are registered to each other in the appropriate sequence and orientation, to generate a hyper-resolution image for the complete shaft. A custom coordinate system is defined, which permits the visualization and tagging of the hyper-resolution image as tiles in a GIS compatible mapping framework. This framework with its client-server architecture permits the visualization of the complete shaft in any browser environment which supports JavaScript and HTML, including mobile devices. The layering feature in the mapping framework is used to present additional information registered to the visual representation, this allows for both vector and raster graphics to be added. Every location in the hyper-resolution image can be assigned a tag which is capable of linking to additional information, such as reports, specific to that location. Using the time data in conjunction with the custom coordinate system, the presented framework allows to observe critical locations on the shafts surface over the course of multiple acquisition runs. Furthermore non-rigid registration is available to determine deformations as a function of time.
展开▼