Issuu

International Research Journal of Engineering and Technology (IRJET)

e-ISSN: 2395-0056

Volume: 10 Issue: 04 | Apr 2023

p-ISSN: 2395-0072

www.irjet.net

A DEEP LEARNING APPROACH TO CLASSIFY DRONES AND BIRDS Prof. Bharath Bharadwaj B S1,Apeksha S2, Bindu NP3, S Shree Vidya Spoorthi 4, Udaya S5 1Assistant Professor Dept. of Computer Science & Engineering Maharaja Institute of Technology ,Thandavapura 2,3,4,5Students, Dept. of Computer Science & Engineering Maharaja Institute of Technology ,Thandavapura

---------------------------------------------------------------------***---------------------------------------------------------------------

Abstract - Drones are gaining popularity not just for

However, using visible imagery also presents challenges like crowded backgrounds and confusing drones with birds, which requires a suitable method to solve. The YOLO Deep Learning Network is the ideal solution for this problem due to its higher accuracy, speed, and ability to analyze input images accurately. The latest version of YOLOv4 Deep Convolutional Neural Networks has proven to have the best speed and accuracy in detecting objects, making it a suitable method for UAV detection and recognition using visible imagery..

recreational use, but also for various engineering, disaster management, logistics, and airport security applications. However, their potential use in malicious activities has raised concerns about physical infrastructure security, safety, and surveillance at airports. There have been several reports in recent years of unauthorized drone use causing disruptions in airline operations. To address this problem, a new deep learning-based method has been proposed in this study. This approach is efficient in detecting and recognizing two types of drones and birds, and it outperforms existing detection systems in the literature. The physical and behavioral similarities between drones and birds often lead to confusion, but the proposed method can not only detect the presence or absence of drones but also distinguish between different types of drones and differentiate them from birds.

1.1 Overview

Key Words: drone; UAV; deep learning; convolutional neural network CNN; drone image dataset; drone detection; drone recognition.

Convolutional neural networks (CNNs) are widely recognized as the most effective deep neural networks for object recognition. They excel in feature extraction, which has made them the focus of extensive research and development for this purpose. Compared to conventional object recognition methods, CNNs are preferred because they are capable of extracting a greater number of features, making them highly effective for this task.

1. INTRODUCTION

1.2 Problem Statement

As drone manufacturing technologies continue to advance, their usage in military, commercial, and security settings is increasing. These unmanned aerial vehicles (UAVs) have become a popular choice for applications like airport security, facility protection, and integration into surveillance systems due to their effectiveness. However, drones can also pose a serious threat in security settings, making it important to develop efficient approaches to detect and identify them. This is especially crucial in areas like airport security and military systems, where the intrusion of drones could have dire consequences.

There is growing apprehension about the security, safety, and surveillance of physical infrastructure at airports, as they can be exploited for malevolent purposes. Several instances of unauthorized use of drones at airports have been reported, resulting in disruptions to airline operations and difficulty in locating the drones or birds.

2. EXISTING SYSTEM The ability to detect radio signals, such as telemetry and video feeds, enables the identification of both the drone and the operator if the drone is being controlled remotely. However, drones that fly autonomously along preprogrammed paths using GPS or compass and timer cannot be detected by these systems. The use of video cameras for visual detection is limited in range and only provides directional information, with little to no indication of the drone's distance. Detection in airport environments is particularly challenging due to the reflection of radio signals and difficulties in identifying small drones near the ground, trees, or buildings using radar. Additionally, sensors must be able to distinguish

The detection, recognition, and identification of drones are essential to ensure public safety and mitigate the potential threats they pose. Detection involves observing the target and identifying any suspicious activity, while recognition involves categorizing the target. Identification refers to accurately diagnosing the type of target. While different sensors can be used to detect and recognize drones, visible imagery is preferred due to its high resolution, low cost, and compatibility with various drones.

Impact Factor value: 8.226

ISO 9001:2008 Certified Journal

Page 1754