DustRobust-YOLO: Enhanced UAV Detection in Dusty Conditions

Unmanned Aerial Vehicles pose significant concerns in safety, security, and privacy, which drives the need for robust detection mechanisms to prevent unauthorized usage. Traditional detection methods face significant challenges when deployed in dusty environments. In this research, we propose a novel approach for detecting UAVs in dusty backgrounds. We contribute three key components to address this challenge. Firstly, we introduce synthetic datasets UAV Dust Test Set (UAV-DTS) tailored for UAV detection in dusty environments, facilitating comprehensive model evaluation. Secondly, we evaluate the performance of four state-of-the-art object detection models including YOLOv5, YOLOv8, Faster-RCNN, and Retina-Net by using our proposed UAV-DTS. Thirdly, we introduced a comprehensive UAV detection model DustRobust-YOLO (DR-YOLO) comprising the Dust-Image Enhancement Network (DIE-Net) and an enhanced YOLOv8 object detection model. DIE-Net integrates Adaptive Color Correction and Image Enhancement with Laplacian Pyramids modules to enhance the visibility of UAVs in dust-laden scenes. Our proposed model surpasses existing state-of-the-art models. This work advances UAV detection capabilities while emphasizing the need for tailored datasets and sophisticated deep-learning architectures to address challenging environmental conditions.