MSF-GhostNet: Computationally Efficient YOLO for Detecting Drones in Low-Light Conditions
Uncrewed aerial vehicles (UAVs) are popular in various applications due to their mobility, size, and user-friendliness. However, identifying malicious UAVs presents challenges that need to be encountered in general image-based object detection. These challenges arise because UAVs can fly at differen...
Saved in:
| Main Authors: | , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
IEEE
2025-01-01
|
| Series: | IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing |
| Subjects: | |
| Online Access: | https://ieeexplore.ieee.org/document/10818706/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1832590414018772992 |
|---|---|
| author | Maham Misbah Misha Urooj Khan Zeeshan Kaleem Ali Muqaibel Muhamad Zeshan Alam Ran Liu Chau Yuen |
| author_facet | Maham Misbah Misha Urooj Khan Zeeshan Kaleem Ali Muqaibel Muhamad Zeshan Alam Ran Liu Chau Yuen |
| author_sort | Maham Misbah |
| collection | DOAJ |
| description | Uncrewed aerial vehicles (UAVs) are popular in various applications due to their mobility, size, and user-friendliness. However, identifying malicious UAVs presents challenges that need to be encountered in general image-based object detection. These challenges arise because UAVs can fly at different altitudes, making it challenging to distinguish them from other flying objects and identify their size. In addition, the speed of UAVs also adds to the difficulty of capturing their clear images, which can lead to blurring, particularly in complex backgrounds. To address these challenges, we present an improved YOLOv5 architecture named multiscale feature map GhostNet (MSF-GhostNet) by introducing GhostConv and C3Ghost modules to reduce the redundant operations in the head and neck. We also proposed three feature map combinations to evaluate the performance of multiscale and multitarget flying objects, including drones, birds, planes, and helicopters. This approach significantly reduces the waste of computing resources when detecting small-sized flying objects. We also integrated autoanchor and batch size mechanisms to ensure efficient model training and avoid overfitting. Our proposed model showed 1.25% fewer false positives than the state-of-the-art GhostNet-YOLOv5 model. The proposed MSF-GhostNet outperformed GhostNet-YOLOv5 with higher precision, recall, and F1 scores (1.3%, 5.3%, and 3.7%, respectively) and reduced model parameters and model size by 3.1% and 4.1%, respectively. The proposed solution also outperformed several other state-of-the-art algorithms exists in the literature. |
| format | Article |
| id | doaj-art-a83ed012c1af44af8c8434c4288c0ce5 |
| institution | Kabale University |
| issn | 1939-1404 2151-1535 |
| language | English |
| publishDate | 2025-01-01 |
| publisher | IEEE |
| record_format | Article |
| series | IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing |
| spelling | doaj-art-a83ed012c1af44af8c8434c4288c0ce52025-01-24T00:00:59ZengIEEEIEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing1939-14042151-15352025-01-01183840385110.1109/JSTARS.2024.352437910818706MSF-GhostNet: Computationally Efficient YOLO for Detecting Drones in Low-Light ConditionsMaham Misbah0https://orcid.org/0009-0005-6277-9974Misha Urooj Khan1https://orcid.org/0000-0002-5852-0469Zeeshan Kaleem2https://orcid.org/0000-0002-7163-0443Ali Muqaibel3https://orcid.org/0000-0001-7865-1987Muhamad Zeshan Alam4https://orcid.org/0000-0002-0114-8248Ran Liu5https://orcid.org/0000-0002-6343-4645Chau Yuen6https://orcid.org/0000-0002-9307-2120Department of Electrical and Communication Engineering, United Arab Emirates University, Al Ain, United Arab EmiratesDepartment of Electrical and Computer Engineering, COMSATS University Islamabad, Wah Campus, Wah Cantt, Rawalpindi, PakistanDepartment of Computer Engineering, and Interdisciplinary Research Center for Smart Mobility and Logistics, King Fahd University of Petroleum and Minerals, Dhahran, Saudi ArabiaElectrical Engineering Department and Center for Communication Systems and Sensing, King Fahd University for Petroleum and Minerals, Dhahran, Saudi ArabiaFaculty of Computer Science, University of New Brunswick, Fredericton NB, CanadaSchool of Electrical and Electronics Engineering, Nanyang Technological University, SingaporeSchool of Electrical and Electronics Engineering, Nanyang Technological University, SingaporeUncrewed aerial vehicles (UAVs) are popular in various applications due to their mobility, size, and user-friendliness. However, identifying malicious UAVs presents challenges that need to be encountered in general image-based object detection. These challenges arise because UAVs can fly at different altitudes, making it challenging to distinguish them from other flying objects and identify their size. In addition, the speed of UAVs also adds to the difficulty of capturing their clear images, which can lead to blurring, particularly in complex backgrounds. To address these challenges, we present an improved YOLOv5 architecture named multiscale feature map GhostNet (MSF-GhostNet) by introducing GhostConv and C3Ghost modules to reduce the redundant operations in the head and neck. We also proposed three feature map combinations to evaluate the performance of multiscale and multitarget flying objects, including drones, birds, planes, and helicopters. This approach significantly reduces the waste of computing resources when detecting small-sized flying objects. We also integrated autoanchor and batch size mechanisms to ensure efficient model training and avoid overfitting. Our proposed model showed 1.25% fewer false positives than the state-of-the-art GhostNet-YOLOv5 model. The proposed MSF-GhostNet outperformed GhostNet-YOLOv5 with higher precision, recall, and F1 scores (1.3%, 5.3%, and 3.7%, respectively) and reduced model parameters and model size by 3.1% and 4.1%, respectively. The proposed solution also outperformed several other state-of-the-art algorithms exists in the literature.https://ieeexplore.ieee.org/document/10818706/Computational complexityGhostNetinfrared (IR) imagesuncrewed aerial vehicle (UAV) detectionYOLOv5 |
| spellingShingle | Maham Misbah Misha Urooj Khan Zeeshan Kaleem Ali Muqaibel Muhamad Zeshan Alam Ran Liu Chau Yuen MSF-GhostNet: Computationally Efficient YOLO for Detecting Drones in Low-Light Conditions IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing Computational complexity GhostNet infrared (IR) images uncrewed aerial vehicle (UAV) detection YOLOv5 |
| title | MSF-GhostNet: Computationally Efficient YOLO for Detecting Drones in Low-Light Conditions |
| title_full | MSF-GhostNet: Computationally Efficient YOLO for Detecting Drones in Low-Light Conditions |
| title_fullStr | MSF-GhostNet: Computationally Efficient YOLO for Detecting Drones in Low-Light Conditions |
| title_full_unstemmed | MSF-GhostNet: Computationally Efficient YOLO for Detecting Drones in Low-Light Conditions |
| title_short | MSF-GhostNet: Computationally Efficient YOLO for Detecting Drones in Low-Light Conditions |
| title_sort | msf ghostnet computationally efficient yolo for detecting drones in low light conditions |
| topic | Computational complexity GhostNet infrared (IR) images uncrewed aerial vehicle (UAV) detection YOLOv5 |
| url | https://ieeexplore.ieee.org/document/10818706/ |
| work_keys_str_mv | AT mahammisbah msfghostnetcomputationallyefficientyolofordetectingdronesinlowlightconditions AT mishauroojkhan msfghostnetcomputationallyefficientyolofordetectingdronesinlowlightconditions AT zeeshankaleem msfghostnetcomputationallyefficientyolofordetectingdronesinlowlightconditions AT alimuqaibel msfghostnetcomputationallyefficientyolofordetectingdronesinlowlightconditions AT muhamadzeshanalam msfghostnetcomputationallyefficientyolofordetectingdronesinlowlightconditions AT ranliu msfghostnetcomputationallyefficientyolofordetectingdronesinlowlightconditions AT chauyuen msfghostnetcomputationallyefficientyolofordetectingdronesinlowlightconditions |