随着智能驾驶和交通系统的迅速发展，图像恢复和目标检测领域得到了飞速的发展和广泛的应用。其中，交通标志的精准识别是智能交通系统的关键任务。然而，雨天天气产生的雨痕会对图像造成遮挡，从而影响交通标志识别的精度。因此，本文针对该问题，对相关算法进行了研究与应用。本课题完成的主要工作与创新如下：

（1）针对雨天产生的雨痕会掩盖部分图像特征导致后续识别任务性能下降的问题，本文提出了一种渐进式多尺度残差去雨网络PMRNet（Progressive multiscale residual network）。首先，该方法在残差结构的基础上加入了不同尺度的卷积，并且该网络采用跳跃连接的方式，以此来增强网络的特征提取能力。其次，网络采用多阶段的方式，以此来弥补单一阶段网络的雨痕残留问题，从而提高网络的去雨效果。实验结果表明，相对于经典去雨算法RESCAN在峰值信噪比上提高了3.6dB，在结构相似度上提高了4.3%。

（2）针对卷积神经网络在交通标志识别任务中特征提取不足的问题，本文提出了CoT-RDYOLOv8（Contextual Transformer-Residual Dense YOLOv8）交通标志识别算法。首先，该算法针对YOLOv8的C2f模块进行了改进，采用密集残差连接方式对该模块的结构进行优化。该方法能增强特征的传递和融合，以此来提高网络的识别精度。其次，本文将网络中部分的3×3卷积替换为CoT模块以增强全局特征表达，并且该模块保留了卷积神经网络的局部特征提取能力。实验结果表明，CoT-RDYOLOv8交通标志识别算法的准确率达到了91.1%。本文的算法与经典交通标志识别算法Faster R-CNN + ACFPN、SSD、Yolov5和Yolov8相比分别提高了1.0%、7.4%、4.0%和3.0%。

（3）针对雨天天气下的交通标志识别任务，本文开发了一套基于浏览器/服务器（B/S）架构的雨天交通标志识别系统。该系统基于PMRNet图像去雨算法和CoT-RDYOLOv8交通标志识别算法实现。系统在满足正常天气情况下的交通标志识别任务的同时，实现了在雨天环境下对交通标志图像的去雨、和识别功能。

         With the rapid development of intelligent driving and traffic systems, the fields of image restoration and object detection have seen swift progress and widespread application. Accurate recognition of traffic signs is a key task in intelligent traffic systems. However, rain streaks caused by rainy weather can obstruct images, thereby impacting the accuracy of traffic sign recognition. Therefore, this paper focuses on the issue of low accuracy in traffic sign recognition algorithms under rainy conditions, and studies and applies relevant algorithms. The main work and innovations completed in this project are as follows：

In order to solve the problem that rain streaks produced on rainy days will cover up some image features, leading to a decrease in the performance of subsequent recognition tasks, this paper proposes a progressive multiscale residual network to remove rain, PMRNet (Progressive multiscale residual network). First, this method adds convolutions of different scales to the residual structure, and the network uses skip connections to improve the feature extraction capability of a single-size residual structure. Secondly, the network adopts a multi-stage approach to make up for the problem of residual rain marks in a single-stage network, thereby improving the rain removal effect of the network. Experimental results show that compared with the classic rain removal algorithm RESCAN, the peak signal-to-noise ratio is improved by 3.6dB and the structural similarity is improved by 4.3%.

In order to address the problem of insufficient feature extraction by convolutional neural network in traffic sign recognition task, this paper proposes the CoT-RDYOLOv8 (Contextual Transformer-Residual Dense YOLOv8) algorithm. Firstly, the algorithm is improved for the C2f module of YOLOv8, and the structure of the module is optimized by using dense residual connection. This method enhances the transfer and fusion of features as a way to improve the detection and recognition accuracy of the network. Secondly, this paper replaces part of the 3×3 convolution in the network with the CoT module to enhance the global feature representation, and the module retains the local feature extraction capability of convolutional neural networks. The experimental results show that the CoT-RDYOLOv8 traffic sign recognition recognition algorithm achieves an accuracy of 91.1%, which is 1.0%, 7.4%, 4.0%, and 3.0% higher than the classical traffic sign recognition algorithms Faster R-CNN + ACFPN, SSD, Yolov5, and Yolov8, respectively.

In order to address the task of traffic sign recognition in rainy weather, this paper develops a rainy weather traffic sign recognition system based on browser/server (B/S) architecture. The system is realized based on PMRNet image de-raining algorithm and CoT-RDYOLOv8 traffic sign detection algorithm. The system realizes the rain removal, and recognition functions of traffic sign images in rainy day environment while satisfying the traffic sign recognition task in normal weather conditions.

[1]Chattaraj D , Bera B , Das A K , et al. Block-CLAP: Blockchain-assisted Certificateless Key Agreement Protocol for Internet of Vehicles in Smart Transportation[J]. IEEE Transactions on Vehicular Technology, 2021, PP(99):1-1.

[2]Chang C C , Lina H Y , Huang S C . Traffic Sign Detection and Recognition for Driving Assistance System[J]. Advances In Image and Video Processing, 2018, 6(3).

[3]Aswathy Madhu;Veena S Nair.Traffic Sign Detection and Recognition for Automated Driverless Cars Based on SSD[J].International Journal of Trend in Scientific Research and Development,2020,Vol.4(5).

[4]Shih-Chia Huang;Quoc-Viet Hoang;Trung-Hieu Le.SFA-Net: A Selective Features Absorption Network for Object Detection in Rainy Weather Conditions[J].IEEE transactions on neural networks and learning systems,2022.

[5]Wu Y, Li Z, Chen Y, et al. Real-time traffic sign detection and classification towards real traffic scene[J]. Multimedia Tools and Applications, 2020, 79(25): 18201-18219.

[6]陈飞, 刘云鹏, 李思远. 复杂环境下的交通标志检测与识别方法综述[J]. 计算机工程与应用, 2021, 57(16):9.

[7]Li R, Cheong L F, Tan R T. Heavy rain image restoration: Integrating physics model and conditional adversarial learning[C].Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. 2019: 1633-1642.

[8]黄淑英,许亚婷,杨勇.联合形态学滤波和卷积稀疏编码的图像去雨[J].中国图象图形学报, 2022(005):027.

[9]刘亚楠,曹风云,王璐.基于双阈值引导滤波的单幅图像去雨算法[J].甘肃联合大学学报（自然科学版）, 2022(003):036.

[10]杨蕾蕾,吐尔洪江·阿布都克力木.基于滚动导向滤波的单幅图像去雨算法[J].河南科技, 2021, 40(20):3.

[11]Deng L J , Huang T Z , Zhao X L , et al. A directional global sparse model for single image rain removal[J]. Applied Mathematical Modelling, 2018, 59.

[12]肖进胜,王文,邹文涛. 基于景深和稀疏编码的图像去雨算法[J]. 计算机学报,2019,42(9):2024-2034.

[13]Son C H，Zhang Xiaoping． Rain detection and removal via shrinkage-based sparse coding and learned rain dictionary[J]．Journal of Imaging Science and Technology，2020，64( 3) : 30501.

[14]汤红忠，刘婷，曾淑英． 群稀疏残差约束的引导字典学习算法及其单幅图像去雨[J]．计算机辅助设计与图形学学报，2020，32( 8) : 1267-1277.

[15]Luo Yu，Ling Jie． Single-image de-raining using low-rank matrix approximation[J]．Neural Computing and Applications，2020，32( 11) : 7503-7514．

[16]Hong Wang;Qi Xie;Qian Zhao;Deyu Meng.A Model-driven Deep Neural Network for Single Image Rain Removal[J].2020.

[17]Fu X, Huang J, Zeng D, et.al. Removing rain from single images via a deep detail network[C].Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. 2017: 3855-3863.

[18]Xia H, Zhuge R, Li H, et al. Single image rain removal via a simplified residual dense network[J]. IEEE Access, 2018, 6: 66522-66535.

[19]Li X, Wu J, Lin Z, et al. Recurrent squeeze-and-excitation context aggregation net for single image deraining[C].Proceedings of the European conference on computer vision (ECCV). 2018: 254-269.

[20]Ren DW, Zuo WM, Hu QH, Zhu PF, Meng DY. Progressive image deraining networks: A better and simpler baseline. In: Proc. of the IEEE Conf. on Computer Vision and Pattern Recognition. 2019. 3937-3946.

[21]Zamir S.W , Arora A , Khan S , et al. Multi-Stage Progressive Image Restoration[J].2021.

[22]Zhang H, Sindagi V, Patel V M. Image de-raining using a conditional generative adversarial network[J]. IEEE transactions on circuits and systems for video technology, 2019, 30(11): 3943-3956.

[23]Zhang H, Patel V M. Density-aware single image de-raining using a multi-stream dense network[C].Proceedings of the IEEE conference on computer vision and pattern recognition. 2018: 695-704.

[24]Wei W, Meng D, Zhao Q, et al. Semi-supervised transfer learning for image rain removal[C].Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. 2019: 3877-3886.

[25]Yasarla R, Sindagi V A, Patel V M. Syn2real transfer learning for image deraining using gaussian processes[C].Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. 2020: 2726-2736.

[26]Jin X, Chen Z, Lin J, et al. Unsupervised single image deraining with self-supervised constraints[C].2019 IEEE International Conference on Image Processing (ICIP). IEEE, 2019: 2761-2765.

[27]蔡俊,赵超,沈晓波.SIFT算法在指示类交通标志识别中的应用[J].太原学院学报：自然科学版, 2020, 38(4):6

[28]戴雪瑞,袁雪,乐国庆.复杂场景下基于颜色对和MSER的交通标志检测方法[J].北京交通大学学报, 2018, 42(5):9.

[29]苗丹,卢伟,高娇娇.基于聚类与Hough变换的交通标志检测方法[J].计算机系统应用, 2019, 28(11):5.

[30]吴迪,屈宗顺,赵葵银.基于HOGv-CLBP特征融合和ELM的交通标志识别[J].光电子．激光, 2020, 31(6):7.

[31]刘亚辰,陈跃鹏,张赛硕.融合式空间塔式算子和HIK-SVM的交通标志识别研究[J].交通运输系统工程与信息, 2017, 17(1):7.

[32]Li X, Xie Z, Deng X, et al. Traffic sign detection based on improved faster R-CNN for autonomous driving[J]. The Journal of Supercomputing, 2022: 1-21.

[33]徐国整,周越,董斌.基于改进Cascade R-CNN的交通标志牌识别[J].传感器与微系统, 2021, 40(5):5.

[34]Wu J, Liao S. Traffic sign detection based on SSD combined with receptive field module and path aggregation network[J]. Computational intelligence and neuroscience, 2022.

[35]Redmon J, Farhadi A. Yolov3: An incremental improvement[J]. arXiv preprint arXiv:1804.02767, 2018.

[36]Zhang B, Wang G, Wang H, et al. Detecting small chinese traffic signs via improved yolov3 method[J]. Mathematical Problems in Engineering, 2021, 2021: 1-10.

[37]Bochkovskiy A,Wang C Y,Liao H Y M.YOLOv4:optimal speed and accuracy of object detection[J].arXiv preprint arXiv:2004.10934,2020.

[38]Wu X, Cao H. Traffic Sign Detection Algorithm Based On Improved YOLOv4[C]//Journal of Physics: Conference Series. IOP Publishing, 2022, 2258(1): 012009.

[39]Tung P X, Thien N L, Ngoc P V B, et al. Research and Development of a Traffic Sign Recognition Module in Vietnam[J]. Engineering, Technology & Applied Science Research, 2024, 14(1): 12740-12744.

[40]Chen X, Pan J, Dong J, et al. Towards unified deep image deraining: A survey and a new benchmark[J]. arXiv preprint arXiv:2310.03535, 2023.

[41]Song T, Fan S, Jin J, et al. Exploring an efficient frequency-guidance transformer for single image deraining[J]. Signal, Image and Video Processing, 2023: 1-10.

[42]Song T, Li P, Jin G, et al. Image Deraining transformer with sparsity and frequency guidance[C]//2023 IEEE International Conference on Multimedia and Expo (ICME). IEEE, 2023: 1889-1894.

[43]Sarkar M, Mondal U, Pal U, et al. Image deraining via multi-level decomposition and empirical wavelet transform[J]. Multimedia Tools and Applications, 2024: 1-23.

[44]Kang L W，Lin C W，Fu Y H．Automatic single-image-based rain streaks removal via image decomposition［J］．IEEE Trans on Image Processing，2012，21( 4) : 1742-1755．

[45]Hu M, Yang J, Yu J, et al. Prior-Dualgan: Rain Rendering from Coarse to Fine[J]. Available at SSRN 4605631.

[46]Zheng C, Jiang J, Ying W, et al. Single Image Deraining via Feature-based Deep Convolutional Neural Network[C]//2023 IEEE 18th Conference on Industrial Electronics and Applications (ICIEA). IEEE, 2023: 84-89.

[47]Badrieva S, Bernabel S, Agaian S. Comprehensive analysis and benchmarking of de-raining techniques through a rain model approach[C].Multimodal Image Exploitation and Learning 2022. SPIE, 2022, 12100: 135-146.

[48]Hu J, Shen L, Sun G. Squeeze-and-excitation networks[C].Proceedings of the IEEE conference on computer vision and pattern recognition. 2018: 7132-7141.

[49]Wang Z, Ji S. Smoothed dilated convolutions for improved dense prediction[C].Proceedings of the 24th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining. 2018: 2486-2495.

[50]Yang Y, Luo H, Xu H, et al. Towards real-time traffic sign detection and classification[J]. IEEE Transactions on Intelligent transportation systems, 2015, 17(7): 2022-2031.

[51]Lu E H C, Gozdzikiewicz M, Chang K H, et al. A hierarchical approach for traffic sign recognition based on shape detection and image classification[J]. Sensors, 2022, 22(13): 4768.

[52]Akatsuka H, Imai S. Road signposts recognition system[R]. SAE Technical Paper, 1987.

[53]Benallal M, Meunier J.Real-time color segmentation of road signs[C].Canadian Conference on Electrical and Computer Engineering, 2003: 1823-1826.

[54]Syed M H, Kumar S. Road lane line detection based on roi using hough transform algorithm[C].Proceedings of Third International Conference on Computing, Communications, and Cyber-Security: IC4S 2021. Singapore: Springer Nature Singapore, 2022: 567-580.

[55]Barnes N, Zelinsky A, Fletcher L S. Real-Time Speed Sign Detection Using the Radial Symmetry Detector[J]. IEEE Transactions on Intelligent Transportation Systems, 2008,9(2):322-332.

[56]李厚杰，邱天爽，宋海玉,等. 基于径向对称变换的自适应交通禁止标志检测[J]. 光电子·激光，2014,25(3):532-539.

[57]Bharati P, Pramanik A. Deep learning techniques—R-CNN to mask R-CNN: a survey[J]. Computational Intelligence in Pattern Recognition: Proceedings of CIPR 2019, 2020: 657-668.

[58]Lu X, Ji J, Xing Z, et al. Attention and feature fusion SSD for remote sensing object detection[J]. IEEE Transactions on Instrumentation and Measurement, 2021, 70: 1-9.

[59]Kim J U, Kim S T, Kim E S, et al. Towards high-performance object detection: Task-specific design considering classification and localization separation[C].ICASSP 2020-2020 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). IEEE, 2020: 4317-4321.

[60]Pham T D. Time–frequency time–space LSTM for robust classification of physiological signals[J]. Scientific reports, 2021, 11(1): 6936.

[61]Zhou J , Liu J , Li J , et al. Mixed Attention Densely Residual Network for Single Image Super-Resolution[J]. Computer Systems Science and Engineering, 2021, 39(1):133-146.

[62]Bai Y. RELU-function and derived function review[C]//SHS Web of Conferences. EDP Sciences, 2022, 144: 02006.

[63]Bande S , Bhatia V , Prakash S . MSE-based analysis of circular grating self-images fortesting beam collimation[J]. Applied Optics, 2020, 59(24).

[64]Yang W, Tan R T, Feng J, et al. Deep joint rain detection and removal from a single image[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. 2017: 1357-1366.

[65]Zhang H, Sindagi V, Patel V M. Image de-raining using a conditional generative adversarial network[J]. IEEE transactions on circuits and systems for video technology, 2019, 30(11): 3943-3956.

[66]Zhu Z, Liang D, Zhang S, et al. Traffic-sign detection and classification in the wild[C].Proceedings of the IEEE conference on computer vision and pattern recognition. 2016: 2110-2118.

[67]Kirillov A, Girshick R, He K, et al. Panoptic feature pyramid networks[C].Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. 2019: 6399-6408.