The light condition of driving at night is poor, and the target is easily blocked, which makes it difficult for the detection algorithm to accurately determine the edges and shapes of targets. In addition, in the process of vehicle moving, fuzzy sense is easily generated on the captured objects, resulting in the difficulty of feature extraction of targets. To address these issues, this paper proposes an improved object detection algorithm for night-time driving based on YOLOv8n. Firstly, the DCN model is introduced into the C2f module and improved into the DCN_CSP2 module, which is used to replace the C2f module in the Backbone. This enhances the algorithm to capture the shape and edge information of target objects more accurately, improves the feature extraction capability, and reduces the computational burden. Secondly, the DWConv module is specifically introduced into the Neck to reduce the number of model parameters while maintaining computational performance, and improve the computational efficiency and achieve the lightweight of the model. Additionally, to address the issue of the original NMS algorithm potentially causing the loss of important targets when they are partially obscured, a decay function based on the overlap size between the candidate detection boxes and the reference box is introduced. This makes important targets more likely to be retained and participate in the subsequent NMS process, thereby improving the detection performance of the model. Analysis results show that compared with YOLOv8n, the improved algorithms achieve a 6.2% increase in mAP@50 and a 5.7% increase in mAP@50:95 on the rmsw_5k_night dataset, and reduce the computational burden and the number of model parameters, achieving a balance between model lightweight and high performance. The improved algorithm effectively enhances the detection capability for night-time targets, and it lays a solid foundation for the algorithm to be applied to the terminal devices.
WOODJ M .Nighttime driving:visual, lighting and visibility challenges[J].Ophthalmic and Physiological Optics, 2020, 40(2):187-201.
[2]
MAG S, CAIM T, CHENG L, et al .Vehicle detection at night based on the feature of taillight and license plate[C]// Green Connected Automated Transportation and Safety.Singapore:Springer Singapore, 2022:723-744.
[3]
胡忆平 .基于深度学习的夜间车辆检测算法研究[D].长沙:湖南大学, 2020.
[4]
HUY P. Research on night vehicle detection algorithm based on deep learning[D]. Changsha:Hunan University, 2020.(in Chinese)
[5]
BELLA, MANTECÓNT, DÍAZC, et al .A novel system for nighttime vehicle detection based on foveal classifiers with real-time performance[J]. IEEE Transactions on Intelligent Transportation Systems, 2022, 23(6): 5421-5433.
[6]
XUET, ZHANGZ Q, MAW N, et al .Nighttime pedestrian and vehicle detection based on a fast saliency and multifeature fusion algorithm for infrared images[J]. IEEE Transactions on Intelligent Transportation Systems, 2022, 23(9): 16741-16751.
GUOQ M, LIUN B, WANGZ X, et al. Review of deep learning based object detection algorithms[J]. Journal of Detection and Control, 2023, 45(6): 10-20. (in Chinese)
ZHANGB L, QINH R, JIANGS, et al. A method of vehicle detection at night based on RetinaNet and optimized loss functions[J]. Automotive Engineering, 2021, 43(8):1195-1202.(in Chinese)
KONGX, PENGJ Q, ZHANGJ, et al .Vehicle object detection method for low-light environment[J]. Journal of Hunan University (Natural Sciences), 2025, 52(1): 187-195.(in Chinese)
JIANGS, YINS R, LUOT H, et al .Multi-object vehicle detection algorithm based on improved YOLOv4[J].Computer Engineering and Design, 2024, 45(4):1181-1188.(in Chinese)
[15]
CHENZ, CHENX, YANX, et al. A lightweight pedestrian vehicle detection algorithm based on YOLOv5[J].Journal of Advances in Artificial Life Robotics, 2022, 2(4):171-176.
[16]
XIAW J, LIP Q, HUANGH Y, et al .TTD-YOLO:a real-time traffic target detection algorithm based on YOLOV5[J].IEEE Access, 2024, 12:66419-66431.
ZHANGJ Y, DINGB, DINGJ .Research on vehicle detection based on low light environments[J]. Modern Electronics Technique, 2024, 47(7): 17-24.(in Chinese)
WANGZ Y, XUH Y, ZHUX Z, et al .An improved dense pedestrian detection algorithm based on YOLOv8:MER-YOLO[J].Computer Engineering & Science, 2024, 46(6):1050-1062.(in Chinese)
[21]
WANGR X, SHIVANNAR, CHENGD, et al .DCN V2:improved deep & cross network and practical lessons for web-scale learning to rank systems[C]//Proceedings of the Web Conference 2021.April 19-23, 2021, Ljubljana, Slovenia:ACM, 2021:1785-1797.
[22]
ZHOUG, CHENW, LIX, et al. Split depth-wise separable graph-convolution network for road extraction in complex environments from high-resolution remote-sensing images[J]. IEEE Transactions on Geoscience and Remote Sensing, 2022, 60: 1-15.
[23]
ZHANGJ M, SCLAROFFS, LINZ, et al .Unconstrained salient object detection via proposal subset optimization[C]//2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).June 27-30, 2016, Las Vegas, NV, USA.IEEE, 2016:5733-5742.
[24]
RENS Q, HEK M, GIRSHICKR, et al .Faster R-CNN:towards real-time object detection with region proposal networks[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2017, 39(6): 1137-1149.
[25]
LIUW, ANGUELOVD, ERHAND, et al .SSD:single shot MultiBox detector[M]//Computer Vision – ECCV 2016.Cham:Springer International Publishing, 2016:21-37.
[26]
LINT Y, GOYALP, GIRSHICKR, et al .Focal loss for dense object detection[C]//2017 IEEE International Conference on Computer Vision (ICCV).October 22-29, 2017, Venice, Italy.IEEE, 2017:2999-3007.
基金资助
国家自然科学基金资助项目(52375093)
National Natural ScienceFoundation of China(52375093)
广东省自然科学基金资助项目(2023A15 15011585)
Natural Science Foundation of Guangdong Province(2023A1515011585)