Please use this identifier to cite or link to this item:
http://hdl.handle.net/10603/355943
Title: | Improved YOLO Deep Learning Network Models for Better Performance in Drone Detection |
Researcher: | Kavitha, T |
Guide(s): | Lakshmi, K |
Keywords: | Computer Science Computer Science Software Engineering Engineering and Technology |
University: | Periyar Maniammai University |
Completed Date: | 2021 |
Abstract: | In recent years, the use of Unmanned Aerial Vehicles (UAVs) and drones has increased tremendously due to easy availability and affordability. Drones are used in different domains for various applications like delivery of goods, monitoring crowds, and securing an area against intruders. Nevertheless, they have the potential to create serious security concerns when used to trespass secure areas and capture information. Consequently, there is a pressing need to protect privacy and eliminate possible drone-related security threats and breaches in sensitive areas. Intense research is being carried out in the domain of UAV and multiple solutions related to privacy and security issues have been proposed by many eminent scholars and researchers over the years. A Deep Learning (DL) based computer vision is one such solution that can be used to detect trespassing or intruding drones effectively. The present work reviews the key state-of-the-art deep learning techniques used for drone detection and proposes a robust drone detection model based on You Only Look Once (YOLO) object detection algorithm. The performance of the YOLOv3 based drone detection system is evaluated using data sets of different sizes. The results show that the new system performs efficiently and produces better results when compared to the existing drone detection systems. newlineDL has captured the attention of both academicians and researchers owing to its endless possibilities and valuable applications in multiple domains. Researchers of almost every domain actively seek to obtain a deep learning based solution to optimise existing processes and resolve complex problems. Implementing DL for a specific application is a challenging task for researchers as training a DL network requires huge amounts of training data and very high computational power is needed to process this exceptionally large data. Even a small-scale application using the DL technique demands several days of training the model on high-end and powerful GPU or TPU clusters. |
Pagination: | |
URI: | http://hdl.handle.net/10603/355943 |
Appears in Departments: | Department of Computer Science and Engineering |
Files in This Item:
File | Description | Size | Format | |
---|---|---|---|---|
10 chapter 1.pdf | Attached File | 66.8 kB | Adobe PDF | View/Open |
11 chapter 2.pdf | 627.04 kB | Adobe PDF | View/Open | |
12 chapter 3.pdf | 738.68 kB | Adobe PDF | View/Open | |
13 chapter 4.pdf | 406.54 kB | Adobe PDF | View/Open | |
14 chapter 5.pdf | 54.44 kB | Adobe PDF | View/Open | |
15 references.pdf | 132.74 kB | Adobe PDF | View/Open | |
16 list of publications.pdf | 21.77 kB | Adobe PDF | View/Open | |
17 curriculum vitae.pdf | 129.49 kB | Adobe PDF | View/Open | |
18 plagiarism report.pdf | 37.02 kB | Adobe PDF | View/Open | |
1 title page.pdf | 46.56 kB | Adobe PDF | View/Open | |
2 certificate.pdf | 115.82 kB | Adobe PDF | View/Open | |
3 declaration.pdf | 157.6 kB | Adobe PDF | View/Open | |
4 acknowledgement.pdf | 26.64 kB | Adobe PDF | View/Open | |
5 contents.pdf | 42.82 kB | Adobe PDF | View/Open | |
6 list of figures.pdf | 22.04 kB | Adobe PDF | View/Open | |
7 list of tables.pdf | 14.66 kB | Adobe PDF | View/Open | |
80_recommendation.pdf | 90.81 kB | Adobe PDF | View/Open | |
8 list of abbreviations.pdf | 19.1 kB | Adobe PDF | View/Open | |
9 abstract.pdf | 27.77 kB | Adobe PDF | View/Open |
Items in Shodhganga are licensed under Creative Commons Licence Attribution-NonCommercial-ShareAlike 4.0 International (CC BY-NC-SA 4.0).
Altmetric Badge: