Current methods for traffic sign detection primarily rely on single-stage deep learning algorithms to construct target detection models, which suffer from low detection accuracy and weak model generalizability. To address these issues, a traffic-sign detection method based on an improved the YOLOv8 model was proposed. This method introduces an attention-scale sequence fusion mechanism that enhances the ability of the neural network to extract multi-scale information. The addition of a small-object detection layer makes this method more suitable for small-object detection. Additionally, it adopts the RT-DETR detector head to process multi-scale features efficiently by decoupling the intra-scale interactions and cross-scale fusions. Moreover, to overcome the limitations of existing methods in terms of weak generalization and to improve the accuracy and efficiency of bounding box regression, a novel loss function, inner-mpdiou, is employed, which effectively improves the training efficiency and accuracy of the model. Experimental results on the Tsinghua-Tencent 100K (TT100K) dataset showed that under the premise of real-time performance, this method achieved an average accuracy of up to 84.0%. Compared with the current international mainstream YOLOv8 model, its average accuracy was improved by 7.1%, and the overall model size was reduced by 12.9%, thereby enhancing the effectiveness of low-resolution small object detection.
上述方法虽然解决了模型对于小目标检测精度较低的问题,但交通标志检测同时存在着实时性的要求。Wang等[21]在改进的Faster R-CNN网络中引入了Score-IOU平衡抽样方法,能获得质量更好的候选框,高效地解决现实中小交通标志的检测与识别问题。Hoang等[22]基于自适应ROI(Region of Interest)和RetinaNet检测,分类道路上的箭头和自行车标记,丢弃了其他不相关的噪声,增强了模型的性能。为了追求实时检测,Zhang等[23]提出了一种由预处理、道路交通标志检测和分割组成的方法,在道路交通标志检测阶段采用轻量级网络和注意力机制相结合的方法提高准确性和对交通标志的敏感性。这些多阶段检测方法虽然解决了交通标志检测的实时性问题,但在实际应用场景需要将模型部署在移动设备上,以便在真实的环境中快速、准确地检测出不同大小和形状的交通标志。因此,模型需要在参数个数较少、检测速度较快的情况下,提高多尺度目标的检测性能。基于以上需求,本文选取最新的单阶段检测器YOLOv8模型。
ZOUZ X, CHENK Y, SHIZ W, et al. Object detection in 20 years: A survey[J]. Proceedings of the IEEE, 2023, 111(3): 257-276. DOI: 10.1109/JPROC.2023.3238524 .
PANX Y, JIAN X, MUY Z, et al. Survey of small object detection[J]. Journal of Image and Graphics, 2023, 28(9): 2587-2615. DOI: 10.11834/jig.220455(Ch ).
[4]
CHUJ Q, ZHANGC, YANM M, et al. TRD-YOLO: A real-time, high-performance small traffic sign detection algorithm[J]. Sensors, 2023, 23(8): 3871. DOI: 10.3390/s23083871 .
JIANGY N, ZHANGS Y. A review of research on detection and recognition methods of traffic signs [J]. Automotive Engineer, 2021(8): 15-18. DOI: 10.3969/j.issn.1674-6546.2021.08.004(Ch ).
YUP P, QIL, MAM L, et al. Traffic signs detection based on visual attention mechanism and shape feature[J]. Mathematics in Practice and Theory, 2019, 49(21): 123-131 (Ch).
[11]
DALALN, TRIGGSB. Histograms of oriented gradients for human detection[C]//2005 IEEE Computer Society Conference on Computer Vision and Pattern Recognition. New York: IEEE Press, 2005: 886-893. DOI: 10.1109/CVPR.2005.177 .
[12]
TAKAKIM, FUJIYOSHIH. Traffic sign recognition using SIFT features[J]. IEEJ Transactions on Electronics, Information and Systems, 2009, 129(5): 824-831. DOI: 10.1541/ieejeiss.129.824 .
[13]
XUEB, YIW J, JINGF, et al. Complex ISAR target recognition using deep adaptive learning[J]. Engineering Applications of Artificial Intelligence, 2021, 97: 104025. DOI: 10.1016/j.engappai.2020.104025 .
[14]
XUEB, TONGN N. DIOD: Fast and efficient weakly semi-supervised deep complex ISAR object detection[J]. IEEE Transactions on Cybernetics, 2019, 49(11): 3991-4003. DOI: 10.1109/TCYB.2018.2856821 .
[15]
XUEB, HEY, JINGF, et al. Robot target recognition using deep federated learning[J]. International Journal of Intelligent Systems, 2021, 36(12): 7754-7769. DOI: 10.1002/int.22606 .
[16]
XUEB, TONGN N. Real-world ISAR object recognition using deep multimodal relation learning[J]. IEEE Transactions on Cybernetics, 2020, 50(10): 4256-4267. DOI: 10.1109/TCYB.2019.2933224 .
[17]
XUEB, HEY, JINGF, et al. Dynamic coarse-to-fine ISAR image blind denoising using active joint prior learning[J]. International Journal of Intelligent Systems, 2021, 36(8): 4143-4166. DOI: 10.1002/int.22454 .
[18]
WUX W, SAHOOD, HOIS C H. Recent advances in deep learning for object detection[J]. Neurocomputing, 2020, 396: 39-64. DOI: 10.1016/j.neucom.2020.01.085 .
[19]
LIZ M, PENGC, YUG, et al. Light-head R-CNN: In defense of two-stage object detector[EB/OL]. 2017: 1711.07264.
[20]
ZHOUW, LIC Y, YEZ W, et al. An efficient tiny defect detection method for PCB with improved YOLO through a compression training strategy[J]. IEEE Transactions on Instrumentation and Measurement, 2024, 73: 2003514. DOI: 10.1109/TIM.2024.3390198 .
[21]
REDMONJ, DIVVALAS, GIRSHICKR, et al. You only look once: Unified, real-time object detection[C]//2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). New York: IEEE Press, 2016: 779-788. DOI: 10.1109/CVPR.2016.91 .
[22]
LUX C, JIJ, XINGZ Q, et al. Attention and feature fusion SSD for remote sensing object detection[J]. IEEE Transactions on Instrumentation and Measurement, 2021, 70: 5501309. DOI: 10.1109/TIM.2021.3052575 .
[23]
HUANGJ, CHENZ X, JONATHAN WUQ M, et al. CATFPN: Adaptive feature pyramid with scale-wise concatenation and self-attention[J]. IEEE Transactions on Circuits and Systems for Video Technology, 2022, 32(12): 8142-8152. DOI: 10.1109/TCSVT.2021.3087002 .
[24]
HONGZ, YAOJ, LIK H, et al. Conjunction of active and semi-supervised learning for wireline logs-based automatic lithology identification[J]. IEEE Geoscience and Remote Sensing Letters, 2022, 19: 3008905. DOI: 10.1109/LGRS.2022.3214929 .
[25]
WANGF, LIY D, WEIY C, et al. Improved faster RCNN for traffic sign detection[C]//2020 IEEE 23rd International Conference on Intelligent Transportation Systems (ITSC). New York: IEEE Press, 2020: 1-6. DOI: 10.1109/ITSC45102.2020.9294270 .
[26]
HOANGT M, NAMS H, PARKK R. Enhanced detection and recognition of road markings based on adaptive region of interest and deep learning[J]. IEEE Access, 2019, 7: 109817-109832. DOI: 10.1109/ACCESS.2019.2933598 .
[27]
ZHANGW W, MIZ Y, ZHENGY C, et al. Road marking segmentation based on siamese attention module and maximum stable external region[J]. IEEE Access, 2019, 7: 143710-143720. DOI: 10.1109/ACCESS.2019.2944993 .
[28]
TANC, YUR W, XIONGS W, et al. Making contract users safer: Towards building a safe browsing platform on Ethereum[J]. ISA Transactions, 2023, 141: 121-131. DOI: 10.1016/j.isatra.2023.04.021 .
[29]
ELFWINGS, UCHIBEE, DOYAK. Sigmoid-weighted linear units for neural network function approximation in reinforcement learning[J]. Neural Networks, 2018, 107: 3-11. DOI: 10.1016/j.neunet.2017.12.012 .
[30]
ZHENGZ H, WANGP, LIUW, et al. Distance-IoU loss: Faster and better learning for bounding box regression[EB/OL]. [2019-11-19]. DOI: 10.1609/aaai.v34i07.6999 .
[31]
ZHANGH, XUC, ZHANGS J. Inner-IoU: More effective intersection over union loss with auxiliary bounding box[EB/OL]. [2023-11-14].
[32]
MAS L, XUY. MPDIoU: A loss for efficient and accurate bounding box regression[EB/OL]. [2023-07-14]. DOI: 10.1016/j.imavis.2024.105381 .
[33]
ZHUZ, LIANGD, ZHANGS H, et al. Traffic-sign detection and classification in the wild[C]//2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). New York: IEEE Press, 2016: 2110-2118. DOI: 10.1109/CVPR.2016.232 .
[34]
GEZ, LIUS T, WANGF, et al. YOLOX: Exceeding YOLO series in 2021[EB/OL]. 2021: arXiv: 2107.08430.
[35]
LIUW, ANGUELOVD, ERHAND, et al. SSD: Single shot MultiBox detector[EB/OL]. 2015: arXiv: 1512.02325. DOI: 10.1007/978-3-319-46448-0_2 .
[36]
SUNC, AIY B, WANGS, et al. Mask-guided SSD for small-object detection[J]. Applied Intelligence, 2021, 51(6): 3311-3322. DOI: 10.1007/s10489-020-01949-0 .