Vision-based MAV Navigation in Underground Mine Using Convolutional Neural Network

摘要: This article presents a Convolutional Neural Network (CNN) method to enable autonomous navigation of low-cost Micro Aerial Vehicle (MAV) platforms along dark underground mine environments. The proposed CNN component provides online heading rate commands for the MAV by utilising image stream from on-board camera, thus allowing platform follow collision-free path tunnel axis. A novel part developed consists generation data-set used training CNN. More specifically, inspired single haze removal algorithms, various data-sets collected real environments have been processed offline provide an estimation depth information scene, where ground truth is not available. calculated map extract open space in tunnel, expressed through area centroid and finally provided considers as floating object, accurate pose required. Finally, capability has successfully experimentally evaluated field trials Sweden.