| This thesis deals with the problem of applying computer vision techniques in an underground mine environment. In particular, the problem of the localization and identification of oversized rock fragments prior to secondary breakage, following the initial drilling and blasting operation is addressed. The strategy employed is based on the methodologies developed for reconstruction and interpretation of range image data. Discrete rock pile images acquired using the NRCC/McGill laser rangefinder were used in this study. The main contribution of this thesis is the complete study of the paradigm which involves: range data acquisition, surface reconstruction, segmentation, and fitting of parametric shape models. The final representation obtained from the model, describing the spatial and geometric properties of each rock fragment and can be used to control an automated rock-breaking mechanism. To support the strategy developed, a number of experimental results at different processing stages are presented. |
