A Survey on Low-Light Object Detection: Architectures, Strategies, and Future Directions

Low-light object detection poses a significant challenge in the field of computer vision, as traditional methods struggle to maintain image quality under insufficient illumination, limiting their applicability in scenarios such as autonomous driving and security surveillance. This paper reviews the paradigm shift in this domain from traditional image enhancement techniques to task-driven intelligent perception. It highlights recent advancements in low-light image enhancement methods and object detection algorithms, including YOLO-based optimizations, end-to-end joint design frameworks, and illumination-invariant feature learning. The review identifies major challenges faced by existing approaches, including the mismatch between enhancement techniques and machine vision requirements, the scarcity of high-quality datasets, and computational efficiency bottlenecks. Future research directions should focus on developing illumination-invariant features, integrating multimodal data, deploying lightweight models, and pursuing task-oriented end-to-end optimization.