Drivable region detection for autonomous robots applied to South African underground mining

Abstract

This dissertation focuses on enhancing autonomous robots' capability to identify drivable regions in underground terrains. A system model that compares the drivability analysis of underground terrains using the entropy model and statistical region merging (SRM) was developed, with a view to presenting an analysis of 2D and 3D results. The approach involves standard image-processing techniques, such as colour and texture feature extraction and region segmentation for underground image classification. A probabilistic method based on the local entropy was employed. The entropy is measured within a fixed window on each frame in order to compute features used in the segmentation process. This research compares the results obtained from the entropy method and SRM approach. Performance evaluation is carried out to provide useful qualitative and quantitative conclusions.