Print Email Facebook Twitter Vision-based Terrain Segmentation and Roughness Estimation Title Vision-based Terrain Segmentation and Roughness Estimation: Application on the CENTAURO Robot Author Suryamurthy, Vivekanandan (TU Delft Mechanical, Maritime and Materials Engineering) Contributor Wisse, Martijn (mentor) Kanoulas, Dimitrios (mentor) Tsagarakis, Nikos (mentor) Degree granting institution Delft University of Technology Programme Mechanical Engineering Date 2018-08-31 Abstract Intelligent terrain perception for search-and-rescue robotic applications, requires a high-level understanding of both the terrain type and its chief physical characteristics. Roughness is one such important terrain property, since it could play a key role in robot control/planning strategies, while navigatingin an unknown environment. In this paper, we present a single deep neural network architecture that predicts the pixel-wise terrain labels (i.e., sand, stone, wood, metal, road/sidewalk, and grass) and regresses their roughness from an input RGB image. Our approach, inspired by human analogy, leverages the basic image feature space from a pre-trained network (SegNet) to estimate the roughness. We experimentally validate our approach in real-world images, using RGB cameras. Moreover, we implement the algorithm on our four-legged centaur-like robot CENTAURO and demonstrate the use of our method inassuring the stability of the robot in real-world scenarios, where the robot is traversing terrains of varying roughness. Subject Deep LearningConvolutional Neural NetworksTerrain segmentationTerrain roughness estimation To reference this document use: http://resolver.tudelft.nl/uuid:865edf61-77e0-4bec-b253-cfe330053581 Embargo date 2018-12-31 Part of collection Student theses Document type master thesis Rights © 2018 Vivekanandan Suryamurthy Files PDF vivek_report.pdf 7.49 MB Close viewer /islandora/object/uuid:865edf61-77e0-4bec-b253-cfe330053581/datastream/OBJ/view