We found a match
Your institution may have access to this item. Find your institution then sign in to continue.
- Title
Robust UAV detection based on saliency cues and magnified features on thermal images.
- Authors
Mebtouche, Naoual El-Djouher; Baha, Nadia
- Abstract
Recent advances in the development of unmanned aerial vehicles (UAV), also known as drones, raised serious security concerns in critical locations. It is, therefore, necessary to design robust systems that can detect drones in any situation. In this paper, we present a novel approach for robust and accurate drone detection in different lighting conditions from thermal images based on deep learning. The main contributions of this paper consist of: First, the introduction of a thermal saliency map as a new feature to adapt deep learning models to learn features from thermal sources. Second, the introduction of a novel module called Magnifying Small Objects (MSO) as a guide for the deep Neural network to improve the detection and localization of small drones, for robust drone detection. Third, a new data augmentation strategy is proposed to generate novel drone images from different sources. Comprehensive experiments are conducted on University of Southern California (USC) thermal UAV dataset and Drone Detection Dataset. A comparative evaluation of the obtained results with state-of-the-art methods is presented. The experimental results show the robustness and high precision of our approach for drone detection.
- Subjects
UNIVERSITY of Southern California; DRONE aircraft; THERMOGRAPHY; DEEP learning; DATA augmentation
- Publication
Multimedia Tools & Applications, 2023, Vol 82, Issue 13, p20039
- ISSN
1380-7501
- Publication type
Article
- DOI
10.1007/s11042-022-14271-3