Please use this identifier to cite or link to this item:
Title: UAV’s recognition of its route through image recognition
Authors: Prakash Sekaran
Keywords: DRNTU::Engineering
Issue Date: 2016
Abstract: The purpose of this paper is to implement a method to allow delivery UAV/drones to recognise their path and stay on them. This is to lessen the probability any possible accidents given real time obstacles like flora and fauna or even other drones, essentially ensuring safety. It also ensures that the drones get to their destination to deliver its package and back. The proposed model is to do image recognition of the drone’s route. This can be done through the implementation of a convolutional neural network (CNN). The data comprises of images from forest trail [1]. There are 3 classes namely for the drone to turn right (TR), turn left (TL) or go straight (GS). Based on the images the CNN will be able to classify the images into their respective classes. The model created was able to correctly classify about 90% of the images in only 10 cycles. This performance is comparable to that of human recognition of the images. This model can then be used to detect real time videos or images to keep the drone in its programmed path of travel. Further improvements can still be made to make sure the classification can be as accurate as possible. This model or similar ones could also be a stepping stone for further advancement in autonomous drone delivery of packages. They application might be slightly different from a forest trail to high rise buildings. Certain tweaks have to be made. They could be coupled with specific building constructions, GPS, sensors, etc. The possibility of autonomous drone delivery can come into reality very soon and hopefully in Singapore as well.
Rights: Nanyang Technological University
Fulltext Permission: restricted
Fulltext Availability: With Fulltext
Appears in Collections:SCSE Student Reports (FYP/IA/PA/PI)

Files in This Item:
File Description SizeFormat 
FYP report.pdf
  Restricted Access
1.08 MBAdobe PDFView/Open

Page view(s)

Updated on May 11, 2021

Download(s) 50

Updated on May 11, 2021

Google ScholarTM


Items in DR-NTU are protected by copyright, with all rights reserved, unless otherwise indicated.