Localization of Unmanned Aerial Vehicles in Corridor Environments using\n Deep Learning

摘要

Vision-based pose estimation of Unmanned Aerial Vehicles (UAV) in unknown\nenvironments is a rapidly growing research area in the field of robot vision.\nThe task becomes more complex when the only available sensor is a static single\ncamera (monocular vision). In this regard, we propose a monocular vision\nassisted localization algorithm, that will help a UAV to navigate safely in\nindoor corridor environments. Always, the aim is to navigate the UAV through a\ncorridor in the forward direction by keeping it at the center with no\norientation either to the left or right side. The algorithm makes use of the\nRGB image, captured from the UAV front camera, and passes it through a trained\ndeep neural network (DNN) to predict the position of the UAV as either on the\nleft or center or right side of the corridor. Depending upon the divergence of\nthe UAV with respect to the central bisector line (CBL) of the corridor, a\nsuitable command is generated to bring the UAV to the center. When the UAV is\nat the center of the corridor, a new image is passed through another trained\nDNN to predict the orientation of the UAV with respect to the CBL of the\ncorridor. If the UAV is either left or right tilted, an appropriate command is\ngenerated to rectify the orientation. We also propose a new corridor dataset,\nnamed NITRCorrV1, which contains images as captured by the UAV front camera\nwhen the UAV is at all possible locations of a variety of corridors. An\nexhaustive set of experiments in different corridors reveal the efficacy of the\nproposed algorithm.\n