A review on the improvement of remote sensing image object detection methods based on the YOLO series algorithms
TL;DR
The future of the field lies in the transition from single-modality visual perception to multi-dimensional collaborative detection systems, and this review serves as a comprehensive reference for optimizing object detection algorithms tailored for complex remote sensing and low-altitude security environments.
Abstract
Object detection in remote sensing images is critical for disaster monitoring, urban planning, and low-altitude defense. However, challenges such as extremely small target scales, complex background interference, and limited edge computing power hinder the performance of standard algorithms. This paper provides a systematic review of improvement strategies for YOLO-based object detection in remote sensing. We analyze key evolutionary directions, including perception-enhanced backbone networks, multi-scale feature fusion mechanisms, background-suppressing attention modules, and high-precision bounding box regression loss functions. Furthermore, this review categorizes cutting-edge applications in scenarios such as UAV monitoring, smart transportation, and multi-source sensor collaborative defense (e.g., electro-optical and radar integration). By evaluating the advantages and limitations of current methodologies, we conclude that the future of the field lies in the transition from single-modality visual perception to multi-dimensional collaborative detection systems. This review serves as a comprehensive reference for optimizing object detection algorithms tailored for complex remote sensing and low-altitude security environments.