Images captured in low-illumination environments typically suffer from issues such as low contrast, significant noise interference, and uneven light distribution, severely compromising the performance of object detection models. To address these challenges, an end-to-end low-illumination object detection algorithm, LLAN-YOLO, is proposed based on the YOLOv11s detection framework. Firstly, an adaptive illumination module is designed to correct both the overall and local light distribution within images. Subsequently, a multi-scale feature enhancement module is constructed to bolster the model's perception of objects across various scales, effectively improving feature extraction for targets of differing dimensions. Finally, a noise correction module is introduced at the end of the YOLOv11s backbone network. This module dynamically assesses the noise level at each pixel location within feature maps, thereby adaptively integrating denoising and enhancement processing strategies. Experiments demonstrate that on the ExDark and DarkFace datasets, LLAN-YOLO achieves mAP50 scores of 73.7% and 52.5% respectively, representing improvements of 1.9 percentage points and 2.2 percentage points over the baseline YOLOv11s model. This confirms LLAN-YOLO’s robust performance in low-illumination object detection tasks.
JiangY, GongX, LiuD, et al. EnlightenGAN: deep light enhancement without paired supervision[J]. IEEE Transactions on Image Processing, 2021, 30: 2340-2349.
[2]
GuoC, LiC, GuoJ, et al. Zero-reference deep curve estimation for low-light image enhancement[C]//2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2020: 1777-1786.
[3]
HnewaM, RadhaH. Multiscale domain adaptive Yolo for cross-domain object detection[C]//2021 IEEE International Conference on Image Processing (ICIP), 2021: 3323-3327.
LaiXin, WangChu, ChenHang. MSRCR optical frequency segmented filter enhancement algorithm in low-light face detection[J]. Journal of Electronic Measurement and Instrumentation, 2022, 36(2): 96-106. (in Chinese)
[6]
LiuW, RenG, YuR, et al. Image-adaptive YOLO for object detection in adverse weather conditions[J]. Proceedings of the AAAI Conference on Artificial Intelligence, 2022, 36(2): 1792-1800.
[7]
KalwarS, PatelD, AanegolaA, et al. GDIP: gated differentiable image processing for object detection in adverse conditions[C]//2023 IEEE International Conference on Robotics and Automation (ICRA), 2023: 7083-7089.
[8]
QinQ, ChangK, HuangM, et al. DENet: detection-driven enhancement network for object detection under adverse weather conditions[C]//Computer Vision-ACCV, 2023: 491-507.
[9]
XiaoY, LiaoH. LIDA-YOLO: an unsupervised low-illumination object detection based on domain adaptation[J]. IET Image Processing, 2024, 18(5): 1178-1188.
[10]
KhanamR, HussainM. YOLOv11: an overview of the key architectural enhancements[PP/OL]. (2024-10-23)[2026-02-06].
MaXin, YUChunyu, CHENGang, et al. Adaptive gamma correction of subregion for non-uniform illumination image enhancement[J]. Laser & Optoelectronics Progress, 2024, 61(14): 383-390. (in Chinese)
[13]
HowardA G, ZhuM, ChenB, et al. MobileNets: efficient convolutional neural networks for mobile vision applications[PP/OL]. (2017-04-17)[2026-02-06].
[14]
LohY P, ChanC. Getting to know low-light images with the exclusively dark dataset[J]. Computer Vision and Image Understanding, 2019, 178: 30-42.
[15]
VargheseR, SambathM. YOLOv8: a novel object detection algorithm with enhanced performance and robustness[C]//2024 International Conference on Advances in Data Engineering and Intelligent Computing Systems (ADICS), 2024: 1-6.
[16]
WangC, YehI H, MarkL H Y. YOLOv9: Learning what you want to learn using programmable gradient information[C]//Computer Vision-ECCV, 2025: 1-21.
[17]
WangA, HuiC, LiuL, et al. YOLOv10: real-time end-to-end object detection[EB/OL].(2024-05-28)[2026-02-06].