随着社会现代化建设的推进和人们生活质量的提高，图像处理和目标识别领域得到了飞速的发展和广泛的应用。由于近年来雾霾天气愈发严重，道路交通所采集到的图像不清晰，导致车牌识别准确率降低。因此，针对雾霾天气下的车牌识别任务，着力于多尺度生成对抗网络图像去雾与车牌识别方法研究，论文的主要工作和成果如下：

针对传统的图像去雾算法对特征提取不完全、模型鲁棒性低的问题，本文提出了一种基于MPGAN（Multi-scale convolution perceptual generative adversarial network）的图像去雾算法。首先，该算法以生成对抗网络为主干构建，结合多尺度卷积来获取更多的特征，直接学习有雾图像和清晰图像之间的非线性映射关系，提高模型的去雾效果；其次，算法提出感知均方误差损失，增强模型的鲁棒性。实验结果表明，基于MPGAN的图像去雾算法的去雾结果更清晰，相对于经典去雾算法AOD-Net在峰值信噪比上提高了4.18dB，在结构相似度上提高了6.2%。

针对车牌识别算法对空间信息提取不完全、受外界环境影响较大导致车牌定位的准确率低的问题，本文提出了一种基于SDCNet（Shuffle dilated convolution network）车牌识别算法，包括车牌定位和车牌识别两部分。首先，该算法以ShuffleNetv2为主干构建，结合空洞卷积和全局上下文块，来提高模型的感受野和空间信息；其次，提出了多因素联合损失，并结合角定位的模式提高模型在复杂环境下定位的准确率。实验结果表明，SDCNet车牌识别算法的准确率达到了98.1%，与经典车牌识别算法LPRNet、MTCNN_LPRNet和RPNet相比分别提高了5.2%、3.6%和2.8%。

针对雾霾天气下的车牌识别任务，本文基于MPGAN图像去雾算法和SDCNet车牌识别算法，设计并实现了一套基于B/S架构的雾天车牌识别系统。该系统不仅能满足正常条件下的车牌识别任务，而且实现了在雾霾环境下对车牌的去雾、定位和识别功能，并对结果进行可视化的展示。

With the advancement of social modernization and the improvement of people's quality of life, the fields of image processing and object recognition have undergone rapid development and extensive application. However, in recent years, the increasingly severe haze weather has led to unclear images collected by road traffic, resulting in a decrease in the accuracy of license plate recognition. Therefore, focusing on the license plate recognition task in hazy weather, this paper focuses on the research of multi-scale generative adversarial network image dehazing and license plate recognition methods. The main work and achievements of this paper are as follows:

To address the problem of incomplete feature extraction and low model robustness in traditional image dehazing algorithms, this paper proposes an MPGAN-based(Multi-scale convolution perceptual generative adversarial network) image dehazing algorithm. Firstly, the algorithm employs a generative adversarial network as the backbone and combines multi-scale convolution to obtain more features, directly learning the nonlinear mapping relationship between hazy and clear images to improve the model's dehazing effect. Secondly, the algorithm proposes perceptual mean square error loss to enhance the robustness of the model. Experimental results show that the dehazing results of the MPGAN-based image dehazing algorithm are clearer, with a 4.18dB increase in peak signal-to-noise ratio and a 6.2% increase in structural similarity compared to the classic dehazing algorithm AOD-Net.

To address the problem of incomplete extraction of spatial information and low accuracy in license plate positioning caused by external environmental factors in license plate recognition algorithms, this paper proposes a SDCNet-based(Shuffle dilated convolution network) license plate recognition algorithm, consisting of license plate positioning and license plate recognition. Firstly, the algorithm employs ShuffleNetv2 as the backbone and combines dilated convolution and global contextual blocks to improve the model's receptive field and spatial information. Secondly, a multi-factor joint loss is proposed, combined with the angle positioning pattern to improve the model's accuracy in complex environments. Experimental results show that the accuracy of the SDCNet-based license plate recognition algorithm reaches 98.1%, which is 5.2%, 3.6%, and 2.8% higher than the classic license plate recognition algorithms LPRNet, MTCNN_LPRNet, and RPNet, respectively.

To address the license plate recognition task in hazy weather, this paper designs and implements license plate recognition system under haze weather based on the MPGAN image dehazing algorithm and the SDCNet license plate recognition algorithm, using a B/S architecture. This system not only meets the task of license plate recognition under normal conditions, but also achieves the functions of dehazing, positioning, and recognition of license plates in haze environments, and visualizes the results.

[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, 70(8): 8092-8107.

[2]蔡允岚, 覃智泉, 李凡,等. 一种黑名单车辆识别及报警方法及应用该方法的视频行车记录设备 [P] . 中国专利：CN201811536204.1，2019-05-03.

[3]吕晨, 程德强, 寇旗旗,等. 基于YOLOv3和ASMS的目标跟踪算法 [J] . 光电工程, 2021, 48(2): 67-77.

[4]汪辉, 高尚兵, 周君等. 基于YOLOv3的多车道车流量统计及车辆跟踪方法 [J] . 国外电子测量技术, 2020, 39(02): 42-46.

[5]Sun Z, Zhang Y, Bao F, et al. ICycleGAN: Single image dehazing based on iterative dehazing model and CycleGAN [J] . Computer Vision and Image Understanding, 2021, 203(2): 103133.

[6]Xue W D, Zhao F Q. Partial differential equation model involving depth of scene for hazy image enhancement [J] . Computer Engineering and Applications, 2017, 53(19): 192-197.

[7]He K M, Sun J, Tang X O. Single image haze removal using dark channel prior [J] . IEEE Transactions on Pattem Analysis and Machine Intelligence, 2011, 33(12): 2341-2353.

[8]BERMAN D, TREIBITZ T, AVIDAN S. Single image dehazing using haze-lines [J] . IEEE Transactions on Pattem Analysis and Machine Intelligence, 2020, 42 (3): 720-734.

[9]Zhu Q, Mai J, Shao L. A Fast Single Image Haze Removal Algorithm Using Color Attenuation Prior [J] . IEEE Transactions on Image Processing, 2015, 24(11): 3522-3533.

[10]Zhang X, Jiang R, Wang T, et al. Single image dehazing via dual-path recurrent network [J] . IEEE Transactions on Image Processing, 2021, 30: 5211-5222.

[11]He L, Zhao J, Zheng N, et al. Haze Removal Using the Difference-Structure-Preservation Prior [J] . IEEE Trans Image Process, 2017, 26(3): 1063-1075.

[12]Sebastian, Salazar-Colores, Eduardo, et al. A Fast Image Dehazing Algorithm Using Morphological Reconstruction [J] . IEEE transactions on image processing : a publication of the IEEE Signal Processing Society, 2018, 28(5): 2357-2366.

[13]Qi L, Gao X, He L, et al. Single Image Dehazing With Depth-Aware Non-Local Total Variation Regularization [J] . IEEE Transactions on Image Processing, 2018, 27(10): 5178-5191.

[14]Zhang H, Patel V M. Densely Connected Pyramid Dehazing Network [J] . 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2018, 3194-3203.

[15]Zhao S, Zhang L, Shen Y, et al. RefineDNet: A Weakly Supervised Refinement Framework for Single Image Dehazing [J] . IEEE Transactions on Image Processing, 2021, 30: 3391-3404.

[16]Li L, Dong Y, Ren W , et al. Semi-Supervised Image Dehazing [J] . IEEE Transactions on Image Processing, 2020, 29: 2766-2779.

[17]Li P, Tian J, Tang Y, et al. Deep retinex network for single image dehazing [J] . IEEE Transactions on Image Processing, 2020, 30: 1100-1115.

[18]Li B, Peng X, Wang Z, et al. AOD-net: all-in-one dehazing network [C] . IEEE International Conference on Computer Vision ( ICCV). 2017: 4780-4788.

[19]Zhang W, Dong L, Pan X, et al. Single Image Defogging Based on Multi-Channel Convolutional MSRCR [J] . IEEE Access, 2019, 7(1): 72492-72504.

[20]Lin H, Zhao J, Li S, et al. License plate location method based on edge detection and mathematical morphology [C] . 2020 IEEE 4th Information Technology, Networking, Electronic and Automation Control Conference (ITNEC). IEEE, 2020, 1: 853-857.

[21]He Z. License Plate Location Method Based on Color Processing and Mathematical Morphology [J] . Smart Factory, 2019.

[22]陈宏照, 谢正光, 卢海伦. 颜色与边缘纹理相结合的车牌定位方法 [J] . 现代电子技术,2018,41(21): 67-70.

[23]胡峰松, 朱浩. 基于多重特征和SURF算法的车牌定位研究 [J] . 计算机工程与应用,2015,51(17): 182-187.

[24]王磊, 王瀚漓, 何良华. 基于双边缘检测的车牌识别算法 [J] . 计算机工程与应用, 2013 (8): 169-173.

[25]Yuan Y, Zou W, Yong Z, et al. A Robust and Efficient Approach to License Plate D etection [J] . IEEE Transactions on Image Processing, 2017, 26(6):1102-1114.

[26]杨人豪, 任斌. 基于颜色特征和边缘检测的车牌识别算法 [J] . 工业控制计算机, 2021, 34(4): 100-103.

[27]罗山. 基于改进数学形态学的车牌定位 [J] . 山西电子技术, 2019, 000(005): 9-11.

[28]Ying T, Xin L, Wanxiang L. License Plate Detection and Localization in Complex Scenes Based on Deep Learning [C] . 东北大学, 中国自动化学会信息物理系统控制与决策专业委员会. 第30届中国控制与决策会议论文集(5), 2018: 1035-1040.

[29]Yao L, Zhao Y, Fan J, et al. Research and Application of License Plate Recognition Technology Based on Deep Learning [C] . Journal of Physics: Conference Series. IOP Publishing, 2019, 1237(2): 022155.

[30]Du L, Li L, Wei D, et al. Saliency-Guided Single Shot Multibox Detector for Target Detection in SAR Images [J] . IEEE Transactions on Geoscience and Remote Sensing, 2020, 58(5): 3366-3376.

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

[32]Ren S, He K, Girshick R, et al. Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks [J] . IEEE Transactions on Pattern Analysis & Machine Intelligence, 2017, 39(6): 1137-1149.

[33]姚文凤, 甄彤, 吕宗旺, 等. 车牌字符分割与识别技术研究 [J] . 现代电子技术, 2020, 43(19): 65-69.

[34]Ansari N N. License Number Plate Recognition using Template Matching [J] . International Journal of Emerging Trends & Technology in Computer Science, 2016, 35(4): 175-178.

[35]李强, 张娟. 一种改进的基于模板匹配的污损车牌识别方法 [J] . 智能计算机与应用, 2019, 009(003): 113-117.

[36]Tian J，Wang G , Liu J，et al. Chinese license plate character segmentation using multiscale template matching [J] . Journal of Electronic Imaging, 2016, 25(5): 053005.1-053005.9.

[37]Li H，WangP，Shen C . Towards End-to-End Car License Plates Detection and Recognition with Deep Neural Networks [J] . IEEE Transactions on Intelligent Transportation Systems, 2019, 20(3): 1126-1136.

[38]Wang W, Yang J, Chen M, et al. A Light CNN for End -to-End Car License Plates Detection and Recognition [J] . IEEE Access, 2019, 7: 173875-173883.

[39]Zherzdev S, Gruzdev A. Lprnet: License plate recognition via deep neural networks [J] . arXiv preprint arXiv:1806.10447, 2018.

[40]Xue X. A two stage lightweight and high performance license plate recognition in MTCNN and LPRNet (2019) [J] . 2019.

[41]Xu Z, Yang W, Meng A, et al. Towards end-to-end license plate detection and recognition: A large dataset and baseline [C] . Proceedings of the European conference on computer vision (ECCV). 2018: 255-271.

[42]Bi G, Ren J, Fu T, et al. Image Dehazing Based on Accurate Estimation of Transmission in the Atmospheric Scattering Model [J] . IEEE Photonics Journal, 2017, 9(4): 1-18.

[43]Cai B, Xu X, Jia K, et al. DehazeNet: An End-to-End System for Single Image Haze Removal [J] . IEEE Transactions on Image Processing, 2016, 25(11): 5187-5198.

[44]陈云彪, 张丽琴, 李汉明等. 基于边缘检测的车牌识别算法的实现 [J] . 电子测试, 2017, 000(021): 41-42.

[45]Chaple G N, Daruwala R D, Gofane M S. Comparisions of Robert, Prewitt, Sobel operator based edge detection methods for real time uses on FPGA [C] . 2015 International Conference on Technologies for Sustainable Development (ICTSD). IEEE, 2015: 1-4.

[46]陆宗骐, 梁诚. 用Sobel算子细化边缘 [J] . 中国图象图形学报, 2000, 05A(6): 516-520.

[47]李牧, 闫继红, 李戈等. 自适应Canny算子边缘检测技术 [J] . 哈尔滨工程大学学报, 2007, 28(9): 1002-1007.

[48]甘玲, 林小晶. 基于连通域提取的车牌字符分割算法 [J] . 计算机仿真, 2011, 28(4): 336-339.

[49]Zhang K, Zuo W, Zhang L. FFDNet: Toward a Fast and Flexible Solution for CNN based Image Denoising [J] . IEEE Transactions on Image Processing, 2017, 27(9): 4608-4622.

[50]He K, Zhang X, Ren S, et al. Deep Residual Learning for Image Recognition [J] . IEEE, 2016, 770-778.

[51]Ioffe S, Szegedy C. Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift [J] . JMLR.org, 2015, 37(9): 448-456.

[52]Creswell A, White T, Dumoulin V, et al. Generative adversarial networks: An overview [J] . IEEE signal processing magazine, 2018, 35(1): 53-65.

[53]Zhou Y, Chen Z, Sheng B, et al. AFF-Dehazing: Attention-based feature fusion network for low-light image Dehazing [J] . Computer Animation and Virtual Worlds, 2021, 32(3/4): e2011.1-e.2011.12.

[54]Ren W, Liu S, Zhang H, et al. Single Image Dehazing via Multi-scale Convolutional Neural Networks [C] . European Conference on Computer Vision Springer. Cham, 2016, 154-169.

[55]Mathieu M, Couprie C, Lecun Y. Deep multi-scale video prediction beyond mean square error [C] . ICLR. 2016.

[56]Simonyan K, Zisserman A. Very Deep Convolutional Networks for Large-Scale Image Recognition [J] . Computer Science, 2014.

[57]Zhao T, Wang J, Wang Z, et al. SSIM-Based Coarse-Grain Scalable Video Coding [J] . IEEE Transactions on Broadcasting, 2015, 61(2): 210-221.

[58]Malarvel M, Sethumadhavan G, Bhagi P , et al. Anisotropic diffusion based denoising on X-radiography images to detect weld defects [J] . Digital Signal Processing, 2017, 68: 112-126.

[59]Ma N, Zhang X, Zheng H T, et al. Shufflenet v2: Practical guidelines for efficient cnn architecture design [C] . Proceedings of the European conference on computer vision (ECCV). 2018: 116-131.

[60]Gomes R, Rozario P, Adhikari N. Deep Learning optimization in remote sensing image segmentation using dilated convolutions and ShuffleNet [C] . 2021 IEEE International Conference on Electro Information Technology (EIT). IEEE, 2021, 244-249.

[61]Yang Y, Deng H. GC-YOLOv3: You Only Look Once with Global Context Block [J] . Electronics, 2020, 9(8): 1235.

[62]An Q, Pan Z, You H, et al. Transitive Transfer Learning Based Anchor Free Rotatable Detector For SAR Target Detection With Few Samples [J] . IEEE Access, 2021, 9: 24011-24025.

[63]Kang X, Huang H, Hu Y, et al. Connectionist temporal classification loss for vector quantized variational autoencoder in zero-shot voice conversion [J] . Digital Signal Processing, 2021, 116(6): 103110.

[64]Shen Z, Zhuang L, Li J, et al. DSOD: Learning Deeply Supervised Object Detectors from Scratch [J] . IEEE, 2017, 1937-1945.

[65]Szegedy C, Vanhoucke V, Ioffe S, et al. Rethinking the Inception Architecture for Computer Vision [J] . IEEE, 2016: 2818-2826.

[66]Szegedy C, Ioffe S, Vanhoucke V, et al. Inception-v4, Inception-ResNet and the Impact of Residual Connections on Learning [P] .10.48550/arXiv.1602.07261. 2016.

[67]Zheng Z , Wang P , Liu W , et al. Distance-IoU Loss: Faster and Better Learning for Bounding Box Regression [J] . arXiv, 2019.

[68]Pan S, Liu J, Chen D. Research on License Plate Detection and Recognition System based on YOLOv7 and LPRNet [J] . Academic Journal of Science and Technology, 2022, 4(2): 62-68.

[69]Awalgaonkar N, Bartakke P, Chaugule R. Automatic license plate recognition system using ssd [C] . 2021 International Symposium of Asian Control Association on Intelligent Robotics and Industrial Automation (IRIA). IEEE, 2021: 394-399.