Academic Journal of Computing & Information Science, 2026, 9(2); doi: 10.25236/AJCIS.2026.090202.
Junjun Zhang1, Zhen Liu2
1School of Public Safety and Emergency Management, Anhui University of Science and Technology, Hefei, China
2School of Computer Science and Engineering, Anhui University of Science and Technology, Huainan, China
To improve the accuracy of vehicle detection for automotive autonomous driving systems under low-illumination, low-contrast, and dynamic lighting interference conditions at night, this paper introduces a lightweight enhanced model named GCD-YOLO based on the YOLOv12 architecture, focusing on strengthening multi-scale feature extraction and fusion mechanisms. First, to enhance the model’s sensitivity to contours and edges and enable learning of richer image feature representations, a Global Edge Information Transfer (GEIT) network is proposed to refine the backbone network, thereby improving detection accuracy. Second, the bottleneck module (A2C2F) is redesigned by incorporating a Content Aware Mixer (CAMixer) module, which achieves efficient and high-quality image super-resolution reconstruction while facilitating better feature fusion. Finally, the original detection head is replaced with a Dynamic Detection Head (DyHead) to mitigate the negative impact of redundant information generated during feature fusion and to increase inference speed. Experimental results demonstrate that GCD-YOLO achieves an accuracy of 94.7% on the BBD100k dataset, representing a 4.6% improvement over YOLOv12, with a detection frame rate of 59.8 frames per second, thus balancing detection accuracy and speed.
Vehicle Detection, Low-light Environments, Feature Fusion, Yolov12
Junjun Zhang, Zhen Liu. GCD-YOLO: Enhanced Vehicle Detection in Low-Light Conditions via Edge Information Transfer and Dynamic Head. Academic Journal of Computing & Information Science (2026), Vol. 9, Issue 2: 11-17. https://doi.org/10.25236/AJCIS.2026.090202.
[1] FAN J Q, LI X, HUO T J. Research on cross domain detection in intelligent vehicles based on one-stage algorithm[J]. China Journal of Highway and Transport,2022,35(3):249-262(in Chinese).
[2] ZHANG X , STORY B, RAJAN D. Night time vehicledetection and tracking by fusing vehicle parts from multiplecameras[J]. IEEE Transactions on Intelligent TransportationSystems,2021,23(7):8136-8156.
[3] LEI Z. Vehicle detection method research based on multi-feature fusion under complicated environment[D]. Jiangxi: East China JiaotongUniversity,2013:44(in Chinese).
[4] SONG P, et al. Detection and tracking of ground moving target by uav in complex environment[D]. Shanxi: Xi’an TechnologicalUniversity,2023:5-7(in Chinese).
[5] GAO Y, et al. Research on nighttime vehicle detection system using feature fusion CNN [D]. Nanjing: Nanjing University,2018.
[6] Hua B S, Tran M K, Yeung S K, et al. Pointwise convolutional neural networks[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. 2018: 984-993
[7] Chollet F, et al. Xception: Deep learning with depthwise separable convolutions[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. 2017: 1251-1258.
[8] Shaw P, Uszkoreit J, Vaswani A, et al. Self-attention with relative position representations[J]. arxiv preprint arxiv:1803.02155, 2018.
[9] Dai X, Chen Y, et al. Dynamic head: Unifying object detection heads with attentions[C]//Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. 2021: 7373-7382.
[10] YU F, CHEN H F, WANG X, et al. Bdd100k: a diverse driving dataset for heterogeneous multitask learning[C]. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2020:2636-2645.
[11] Liu W, Quijano K, Crawford M M. YOLOv5-Tassel: Detecting tassels in RGB UAV imagery with improved YOLOv5 based on transfer learning[J]. IEEE Journal of Selected Topics in Applied Earth
Observations and Remote Sensing, 2022, 15: 8085-8094.
[12] Wang C, Sun W, Wu H, et al. A low-altitude remote sensing inspection method on rural living environments based on a modified YOLOv5s-ViT[J]. Remote Sensing, 2022, 14(19): 4784.
[13] Zhang Y, Guo Z, Wu J, et al.Real-time vehicle detection based on improved yolo v5[J]. Sustainability,2022,14(19): 12274.
[14] Wang C, I-Hau Y, Hong Y, et al. Yolov9: Learning what you want to learn using programmable gradient information[J]. European conference on computer vision, 2024.
[15] Zhang Y, Liu H, Jia S, et al. Lightweight detection method for safety helmet and reflective vest in construction environments[J/OL]. Electronic Measurement Technology, 2025:1-13.
[16] Sun G, Wang X, Li Y, et al. Improved helmet detection algorithm for two-wheeled vehicles of RT-DETR[J/OL]. Journal of Electronic Measurement and Instrument, 2025: 1-13