模糊目标识别检测是视频监控领域中研究的主要内容。通过将采集到的模糊图像进行去模糊处理，并结合YOLOX目标检测算法来实现更高的目标检测精度和速度。因此，进行模糊目标检测算法研究具有一定的研究参考和应用价值。

针对图像模糊而导致的跟踪丢失问题，本文对Deblur GAN-v2去模糊算法进行改进，引入Pyramid Pooling模块增大感受野，提升改进算法的多尺度目标特征提取能力；在ResBlock中，引入Coordinate Attention模块，提供改进算法的空间注意力和通道注意力；在鉴别器中，引入SPNorm模块，有助于提高图像清晰度。通过公开数据集GoPro进行仿真，仿真结果表明本文改进Deblur GAN-v2去模糊算法相较于经典去模糊算法的PSNR（Peak Signal-to-Noise Ratio）值和SSIM（Structural SIMilarity）值分别提升了5%-8%和1%-8%，且在真实环境中进行实验验证，其结果相较于经典去模糊算法的PSNR值和SSIM值分别提升了3%-8%和1%-4%。

在去模糊效果的基础上，为提高目标检测算法检测精度和速度，本文对YOLOX算法进行改进，引入MobileNet-V3模块替换原始YOLOX目标检测算法中的backbone模块，在原始YOLOX目标检测算法中添加Coordinate Attention模块，实现YOLOX目标检测算法的改进。通过公开数据集VOC进行仿真，仿真结果表明本文改进的YOLOX目标检测算法相较于其他目标检测算法的MAP（Mean Average Precision）值和FPS（Frames Per Second）值分别提升了1%-20%和2%-10%。

本文通过DeepSort算法进行了实验验证。验证结果表明本文改进算法的MAP值和FPS值相较于其他算法分别提升了3%-23%和2%-23%，说明本文基于YOLOX的模糊目标检测算法具有良好的抗模糊能力和更快的识别速度，为模糊目标识别检测算法研究提供了一种借鉴和参考。

Fuzzy object recognition and detection is the main research content in video surveillance field. By deblurring the collected fuzzy images and combining with YOLOX target detection algorithm, higher accuracy and speed of target detection are achieved. Therefore, the research of fuzzy object detection algorithm has a certain reference and application value.

In order to solve the problem of tracking loss caused by image blurring, Deblur GAN-v2 de-blurring algorithm is improved in this thesis. Pyramid Pooling module is introduced to increase the perception field and improve the multi-scale target feature extraction capability of the improved algorithm. In ResBlock, Coordinate Attention module is introduced to provide space attention and channel attention of the improved algorithm. In the discriminator, the SPNorm module is introduced to improve the image clarity. Through GoPro simulations with open data sets, Simulation results show that compared with the classical Deblur algorithm, the modified deblur GAN-v2 deblur algorithm has an improved PSNR (Peak Signal-to-Noise Ratio) value and 1%-8% Structural SIMilarity (SSIM) value, respectively. In addition, the results are verified by experiments in real environment, and the PSNR value and SSIM value are increased by 3%-8% and 1%-4% respectively compared with the classical defuzzification algorithm.

On the basis of deblurring effect, in order to improve the detection accuracy and speed of the target detection algorithm, this thesis improves the YOLOX algorithm. MobileNet-V3 module is introduced to replace the backbone module of the original YOLOX target detection algorithm. Add Coordinate Attention module to the original YOLOX target detection algorithm to improve the YOLOX target detection algorithm. The simulation results showed that the improved YOLOX target detection algorithm improved the MAP (Mean Average Precision) value and Frames Per Second (FPS) value by 1%-20% and 2%-10%, respectively, compared with other target detection algorithms.

In this thesis, the DeepSort algorithm is used for experimental verification. The verification results show that the MAP value and FPS value of the improved algorithm in this thesis are increased by 3%-23% and 2%-23% respectively compared with other algorithms, which indicates that the fuzzy target detection algorithm based on YOLOX in this thesis has good anti-fuzzy ability and faster recognition speed, and provides a reference for the research of fuzzy target recognition and detection algorithms.

[1] 袁珊,瞿安朝,钱伟行,等.一种结合深度学习的运动去模糊视觉SLAM方法[J].飞控与探测,2022,5(03):28-36.

[2] 刘晨辉,尹增山,高爽.基于遥感图像序列的运动去模糊算法研究[J].激光与光电子学进展,2022,59(08):496-504.

[3] 刘巍,严洪悦,李肖,等.机床动态检测中的高速图像运动去模糊还原[J].仪器仪表学报, 2018,39(05):224-232.

[4] Tao X, Gao H, Shen X, et al. Scale-recurrent network for deep image deblurring[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. 2018: 8174-8182.

[5] Wang Y, Lu Q, Ren B. Wind Turbine Crack Inspection Using a Quadrotor With Image Motion Blur Avoided[J]. IEEE Robotics and Automation Letters, 2023, 8(2): 1069-1076.

[6] Cho S J, Ji S W, Hong J P, et al. Rethinking coarse-to-fine approach in single image deblurring[C]//Proceedings of the IEEE/CVF international conference on computer vision. 2021: 4641-4650.

[7] Hu X, Ren W, Yu K, et al. Pyramid architecture search for real-time image deblurring[C]//Proceedings of the IEEE/CVF International Conference on Computer Vision. 2021: 4298-4307.

[8] 张潮,宋亚林,袁明阳.基于MultiResUNet-SMIS的皮肤黑色素瘤图像分割[J/OL].计算机系统应用:1-10[2023-03-23].

[9] Wang C, Zhang L, Wei W, et al. Dynamic Super-Pixel Normalization for Robust Hyperspectral Image Classification[J]. IEEE Transactions on Geoscience and Remote Sensing, 2023, 61: 1-13.

[10] 江沸菠,彭于波,董莉.面向6G的深度图像语义通信模型[J/OL].通信学报:1-11[2023-03-23].

[11] 陈炜玲,邱艳玲,赵铁松,等.面向海洋的水下图像处理与视觉技术进展[J/OL].信号处理:1-17[2023-03-23].

[12] 余海洋,范之国,金海红,等.融合图像灰度比值的离焦图像自动判别分离方法[J/OL].激光与光电子学进展:1-19[2023-03-23].

[13] 黄洁,姜志国,张浩鹏,等.基于卷积神经网络的遥感图像舰船目标检测[J].北京航空航天大学学报,2017,43(09):1841-1848.

[14] 王一娴,叶尊忠,斯城燕,等.适配体生物传感器在病原微生物检测中的应用[J].分析化学,2012,40(04):634-642.

[15] 程换新,蒋泽芹,程力,等.基于改进YOLOX-S的安全帽反光衣检测算法[J].电子测量技术,2022,45(06):130-135.

[16] 曹捷,郭志彬,潘立志,等.高空作业场景下的安全带穿戴检测[J].湖南科技大学学报(自然科学版),2022,37(01):92-99.

[17] 宋玉存,葛泉波,朱军龙,等.基于梯度差自适应学习率优化的改进YOLOX目标检测算法[J/OL].航空学报:1-16[2023-03-23].

[18] Liao S, Huang C, Liang Y, et al. Solder Joint Defect Inspection Method Based on ConvNeXt-YOLOX[J]. IEEE Transactions on Components, Packaging and Manufacturing Technology, 2022, 12(11): 1890-1898.

[19] Guo Q, Liu J, Kaliuzhnyi M. YOLOX-SAR: High-Precision Object Detection System Based on Visible and Infrared Sensors for SAR Remote Sensing[J]. IEEE Sensors Journal, 2022, 22(17): 17243-17253.

[20] Wang X, Zhang Z, Xu Y, et al. Real-Time Terahertz Characterization of Minor Defects by the YOLOX-MSA Network[J]. IEEE Transactions on Instrumentation and Measurement, 2022, 71: 1-10.

[21] Chunxiang Z, Jiacheng Q, Wang B. YOLOX on Embedded Device With CCTV & TensorRT for Intelligent Multicategories Garbage Identification and Classification[J]. IEEE Sensors Journal, 2022, 22(16): 16522-16532.

[22] 曾理,熊西林,陈伟.低剂量CT图像降噪的深度图像先验的目标偏移加速算法[J/OL].电子与信息学报:1-9[2022-11-08].

[23] 闫晓鹏.离散小波变换在水下航行器目标图像降噪中的应用[J].舰船科学技术,2021,43(22):67-69.

[24] 张翔松,高炜欣,穆向阳,王思宇,朱士玲.X射线环焊缝图像降噪及增强方法研究[J].激光杂志,2021,42(03):79-85.

[25] 孙凤山,范孟豹,曹丙花,叶波,刘林.基于几何纹理与Anscombe变换的蜂窝材料太赫兹图像降噪模型[J].机械工程学报,2021,57(22):96-105.

[26] 熊海晨,么娆,孙乐萌.基于SRAD-DWT算法的显微图像降噪模型[J].智能计算机与应用,2021,11(10):26-32+37.

[27] 解涛,郭建胜,张晓丰,顾涛勇,赵博欣.基于生成对抗网络的SAR图像降噪模型[J].电光与控制,2022,29(09):48-52.

[28] 王晓红,卢辉,黄中秋等.基于颜色空间变换和CNN的自适应去模糊方法[J].包装工程,2020,41(07):224-233.

[29] 王栋. 基于自主导航机器人的设施内植物三维重建方法研究[D].浙江大学,2020.

[30] Ji S W, Lee J, Kim S W, et al. XYDeblur: divide and conquer for single image deblurring[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 2022: 17421-17430.

[31] 邹冲. 复杂背景下行人检测研究与实现[D]. 武汉工程大学, 2017.

[32] Zha Z, Zhang X, Wu Y, et al. Non-convex weighted Lp nuclear norm based ADMM framework for image restoration[J]. Neurocomputing, 2018, 311: 209-224.

[33] Souibgui M A, Kessentini Y. De-gan: A conditional generative adversarial network for document enhancement[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2020, 44(3): 1180-1191.

[34] Zhang K, Luo W, Zhong Y, et al. Deblurring by realistic blurring[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 2020: 2737-2746.

[35] .Zhao S, Zhang Z, Hong R, et al. FCL-GAN: A lightweight and real-time baseline for unsupervised blind image deblurring[C]//Proceedings of the 30th ACM International Conference on Multimedia. 2022: 6220-6229.

[36] Zou W, Jiang M, Zhang Y, et al. SDWNet: A straight dilated network with wavelet transformation for image deblurring[C]//Proceedings of the IEEE/CVF international conference on computer vision. 2021: 1895-1904.

[37] Girshick R, Donahue J, Darrell T, et al. Region-Based Convolutional Networks for Accurate Object Detection and Segmentation[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2016, 38(1): 142-158.

[38] Girshick R. Fast r-cnn[C]//Proceedings of the IEEE international conference on computer vision. 2015: 1440-1448.

[39] 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.

[40] .He K, Gkioxari G, Dollár P, et al. Mask r-cnn[C]//Proceedings of the IEEE international conference on computer vision. 2017: 2961-2969.

[41] Redmon J, Divvala S, Girshick R, et al. You only look once: Unified, real-time object detection[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. 2016: 779-788.

[42] Redmon J, Farhadi A. YOLO9000: better, faster, stronger[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. 2017: 7263-7271.

[43] Farhadi A, Redmon J. Yolov3: An incremental improvement[C]//Computer vision and pattern recognition. Berlin/Heidelberg, Germany: Springer, 2018, 1804: 1-6.

[44] Wang C Y, Bochkovskiy A, Liao H Y M. Scaled-yolov4: Scaling cross stage partial network[C]//Proceedings of the IEEE/cvf conference on computer vision and pattern recognition. 2021: 13029-13038.

[45] Ge Z, Liu S, Wang F, et al. Yolox: Exceeding yolo series in 2021[J]. arXiv preprint arXiv:2107.08430, 2021.

[46] Liu M, Zhu C. Residual YOLOX-based ship object detection method[C]//2022 2nd International Conference on Consumer Electronics and Computer Engineering (ICCECE). IEEE, 2022: 427-431.

[47] Zhao Y, Xiao M Y, Lv H, et al. Research on scanning acoustic image defects detection of integrated circuits based on yolox[C]//2022 23rd International Conference on Electronic Packaging Technology (ICEPT). IEEE, 2022: 1-4.

[48] Feng L, Jia Y. Traffic sign recognition based on YOLOX in extreme weather[C]//2022 Global Conference on Robotics, Artificial Intelligence and Information Technology (GCRAIT). IEEE, 2022: 299-303.

[49] Guo Q, Liu J, Kaliuzhnyi M. YOLOX-SAR: High-Precision Object Detection System Based on Visible and Infrared Sensors for SAR Remote Sensing[J]. IEEE Sensors Journal, 2022, 22(17): 17243-17253.

[50] Wei Y, Guo Z, Dai C, et al. Distracted Driver Behavior Detection Based-on An Improved YOLOX Framework[C]//2022 27th International Conference on Automation and Computing (ICAC). IEEE, 2022: 1-6.

[51] Wang X, Zhang Z, Xu Y, et al. Real-Time Terahertz Characterization of Minor Defects by the YOLOX-MSA Network[J]. IEEE Transactions on Instrumentation and Measurement, 2022, 71: 1-10.

[52] Chunxiang Z, Jiacheng Q, Wang B. YOLOX on Embedded Device With CCTV & TensorRT for Intelligent Multicategories Garbage Identification and Classification[J]. IEEE Sensors Journal, 2022, 22(16): 16522-16532.