Year: 2018
Type: conference-paper
Abstract: In recent years, the UAV technology is unceasingly emerging as a revolutionary reform among the research community. In this paper, we propose a method that facilitates UAVs with a monocular camera to navigate autonomously in previously unknown and GPS-denied indoor corridor arenas. The proposed system uses a state-of-the-art Convolutional Neural Network (CNN) model to achieve the task. We propose a novel approach, which uses the video feed extracted from the front camera of the UAV and passes it through a deep neural network model to decide on the next course of maneuver. The entire process is treated as a classification task where the deep neural network model is responsible for classifying the image as left, right or center of the corridor. The training is performed over a dataset of images, collected from various indoor corridor environments. Apart from utilizing the front facing camera, the model is not dependent on any other sensor. We demonstrate the efficacy of the proposed system in real-time indoor corridor scenarios.
Source: Procedia Computer Science
Authors: Ram Prasad Padhy, Sachin Verma, Shahzad Ahmad, Suman Kumar Choudhury, Pankaj Kumar Sa
Institution: National Institute of Technology Rourkela
Language: English
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
Open Access status: diamond
Awards: G4004428959, G4506975916
OpenAlex ID: W2883171824