随着科技的不断进步，无人机对地观测目标检测在无人机巡检、农业监测、城市规划和生态保护等领域得到了广泛应用，无人机航拍图像与深度学习检测技术的结合也逐渐成为当今的热门研究方向。由于复杂的目标检测模型难以在边缘设备中部署，而轻量级目标检测模型难以提高准确性，本课题从生成无人机对地观测数据集、提高目标检测模型的精度和降低模型的复杂度三个方面出发，研究了基于深度学习的无人机对地观测图像的生成与检测问题。课题研究内容具体包括如下：

针对因智能任务非平稳、长周期和小样本等特性导致的基础数据样本集不平衡、数据量不足等问题，在传统的图像生成算法基础上，提出一种基于改进深度卷积生成对抗网络的无人机对地观测图像的生成方法。首先对深度卷积生成对抗网络模型结构进行改进，提高模型训练的稳定性；其次改进模型的损失函数，来平衡生成器和判别器的能力，提高模型的收敛速度；最后通过与不同的图像生成算法进行实验对比。实验结果表明，提出的方法在保持模型训练稳定的同时提高了生成图像的质量，生成的图像FID值比其他生成算法都低，且比原深度卷积生成对抗网络低2.409。

针对无人机对地观测图像包含众多小目标且背景复杂，当前主流算法对这些目标漏检率与误检率较高的问题，提出一种基于改进YOLOv8的对地观测目标检测方法。首先在YOLOv8的头部增加检测头，提高对目标特征的多尺度检测；然后在主干网络中引入注意力机制使模型更加聚焦于重要的区域以减少背景信息对模型的影响；最后改进损失函数，使模型更好地适应各种形状和尺寸的物体来提高整个模型的检测精度。实验结果表明，提出的方法UAV-YOLOv8的检测准确率 、召回率和平均检测精度分别是93.4%、94.0%和88.0%，较基准算法分别提升了3.7％、5.5％和2.7％。

针对复杂的目标检测模型难以在边缘设备中部署问题，本课题在上述研究的基础上提出一种基于UAV-YOLOv8的轻量化改进方法。通过引入深度可分离卷积网络对模型结构进行替换，来降低模型的复杂度，同时为了避免模型精度下降，在网络中引入双路径注意门和特征增强模块。实验结果表明，轻量化后的模型的检测准确率 、召回率和平均检测精度分别是94.5%、92.1%和92.1%，模型精度的变化可以忽略不计，同时模型的检测率为138.8FPS，大小为16.1 MB，比基准模型降低了6.4MB，表明在保证精度的同时，轻量化后的模型QUAV-YOLOv8依旧能满足实时监测要求。

为了进一步验证本课题的模型在边缘设备上的应用性能，首先基于灵汐HS110搭建对地观测机载样机系统，将面向对地观测场景的QUAV-YOLOv8目标检测网络进行模型编译，然后将编译后的模型部署在KA200类脑芯片上，最后测试样机的检测精度。实验结果表明，轻量化后的模型在部署到硬件设备上后，目标检测精度只下降了0.1~0.25，在合理的范围内，表明该模型能够有效地部署到边缘设备中。

本课题提出的基于深度学习的无人机对地观测图像的生成与检测方法，能够有效地对无人机对地观测图像进行检测，最终设计的基于QUAV-YOLOv8的目标检测网络也可以成功部署到硬件设备上，为下一步研究深度学习模型的边缘化部署提高了一定的参考价值。

With the continuous progress of science and technology, UAV ground observation target detection has been widely used in the fields of UAV inspection, agricultural monitoring, urban planning and ecological protection, etc. The combination of UAV aerial images and deep learning detection technology has gradually become a popular research direction today. Since complex target detection models are difficult to deploy in edge devices and lightweight target detection models are difficult to improve accuracy, this topic investigates the generation and detection of UAV ground observation images based on deep learning from the three aspects of generating UAV ground observation datasets, improving the accuracy of target detection models, and reducing the complexity of the models. The research content of the topic specifically includes the following:

(1) Aiming at the problems of unbalanced basic data sample set and insufficient data volume due to the characteristics of non-smooth, long period and small samples of intelligent tasks, based on the traditional image generation algorithm, a method of generating UAV ground observation images based on improved deep convolutional generative adversarial network is proposed. Firstly, the model structure of the deep convolutional generative adversarial network is improved to improve the stability of the model; secondly, the loss function of the model is improved to balance the ability of the generator and the discriminator and to improve the convergence speed of the model; and finally, experimental comparisons are made with different image generation algorithms. The experimental results show that the proposed method improves the quality of the generated image while maintaining the stability of the model training, and the generated image FID value is lower than other generation algorithms, and is 2.409 lower than the original deep convolutional generation adversarial network.

(2) Aiming at the problem that UAV ground observation images contain many small targets with complex backgrounds, and the current mainstream algorithms have high omission and false detection rates for these targets, a ground observation target detection method based on the improved YOLOv8 is proposed. Firstly, the detection head is added to the head of YOLOv8 to improve the multi-scale detection of target features; then the attention mechanism is introduced into the backbone network to make the model more focused on important regions to reduce the influence of background information on the model; finally, the loss function is improved to make the model better adapt to various shapes and sizes of objects to improve the detection accuracy of the whole model. The experimental results show that the detection accuracy, recall, and average detection precision of the proposed method UAV-YOLOv8 are 93.4%, 94.0%, and 88.0%, respectively, which are 3.7%, 5.5%, and 2.7% higher than those of the benchmark algorithm.

(3) Aiming at the problem that complex target detection models are difficult to be deployed in edge devices, this topic proposes a lightweight improvement method based on UAV-YOLOv8 on the basis of the above research. The complexity of the model is reduced by introducing a depth-separable convolutional network to replace the model structure, and at the same time, in order to avoid model accuracy degradation, dual-path attention gates and feature enhancement modules are introduced into the network. The experimental results show that the detection accuracy , recall and average detection precision of the lightweighted model are 94.5%, 92.1% and 92.1%, respectively, and the change of the model precision is negligible, while the detection rate of the model is 138.8 FPS, and the size of the model is 16.1 MB, which is 6.4 MB lower than that of the benchmark model, which indicates that, while guaranteeing the precision, the lightweighted model QUAV -YOLOv8 can still meet the real-time monitoring requirements.

(4) In order to further verify the application performance of the model on edge devices, we first build an airborne prototype system for ground observation based on the Spirit HS110, compile the QUAV-YOLOv8 target detection network for ground observation scenarios, then deploy the compiled model on the KA200 brain-like chip, and finally test the detection accuracy of the prototype. The experimental results show that the target detection accuracy of the lightweighted model only decreases by 0.1~0.25 after deploying it to hardware devices, which is within a reasonable range, indicating that the model can be effectively deployed to edge devices.

The proposed deep learning-based method for generating and detecting UAV ground observation images can effectively detect UAV ground observation images, and the finally designed target detection network based on QUAV-YOLOv8 can also be successfully deployed to hardware devices, which improves a certain reference value for the next step of researching the edge deployment of deep learning models.

[1]安培浚, 高峰, 曲建升．对地观测系统未来发展趋势及其技术需求[J]. 遥感技术与应用, 2007, 22(6): 62-767.

[2]林宗坚, 李德仁, 胥燕婴．对地观测技术最新进展评述[J]. 测绘科学, 2011, 36(4): 5-8.

[3]王毅. 国际新一代对地观测系统的发展[J]. 地球科学进展, 2005, 20(9): 980-989.

[4]庄子尤. 基于生成对抗网络的无人机图像去雾技术研究及系统研制[D]. 中国科学院大学, 2021.

[5]韩东哲, 鲁王泽. 基于深度学习的无人机遥感影像灾损评估[J]. 信息技术与信息化, 2024, 01: 22-26.

[6]Bouguettaya A, Zarzour H, Kechida A, et al. Vehicle Detection from UAV Imagery with Deep Learning: A Review. IEEE Trans. Neural Netw. Learn. Syst., 2021, 33: 6047-6067.

[7]Byun S, Shin I, Moon J, et al. Road Traffic Monitoring from UAV Images Using Deep Learning Networks. Remote Sens, 2021, 13: 4027.

[8]吴海斌, 张亚, 胡鹏. 面向无人机航拍图像小目标检测方法[J]. 安徽工业大学学报(自然科学版), 2024, 41(01): 65-73.

[9]Chen Y, Lee W S, Gan H, et al. Strawberry Yield Prediction Based on a Deep Neural Network Using High-Resolution Aerial Orthoimages. Remote Sens, 2019, 11: 1584.

[10]郝立, 张皓迪. 面向无人机作战的复杂光照环境下小目标检测方法研究[J]. 系统仿真技术, 2022, 18(02): 85-89+95.

[11]Cai W, Wei Z. Remote Sensing Image Classification Based on a Cross-Attention Mechanism and Graph Convolution. IEEE Geosci. Remote Sens. Lett., 2020, 19: 1-5.

[12]Deng S, Li S, Xie K, et al. A Global-Local Self-Adaptive Network for Drone-View Object Detection. IEEE Trans. Image Process, 2020, 30: 1556-1569.

[13]刘雯雯, 孙裕晶, 姜树辉, 等. 遮挡情况下的行人检测方法研究[J]. 中国农机化学报, 2023, 44(07): 179-186.

[14]韩晶, 王伟宇, 吕学强, 等. 基于协同注意力与特征融合的无人机小目标检测[J]. 北京信息科技大学学报(自然科学版), 2023, 38(03): 1-8.

[15]Dong J, Ota K, Dong M. UAV-Based Real-Time Survivor Detection System in Post-Disaster Search and Rescue Operations. IEEE J. Miniat. Air Space Syst., 2021, 2: 209-219.

[16]Hsieh M, Lin Y, Hsu W. Drone-Based Object Counting by Spatially Regularized Regional Proposal Network. In Proceedings of the IEEE International Conference on Computer Vision, Venice, Italy, 2017: 4145-4153.

[17]Liao J, Piao Y, Su J, Cai G, et al. Unsupervised Cluster Guided Object Detection in Aerial Images. IEEE J. Sel. Top. Appl. Earth Obs. Remote Sens., 2021, 14: 11204-11216.

[18]Pan Z, Yu W, Yi X, et al. Recent progress on generative adversarial networks ( GANs): A survey [J]. IEEE Access, 2019, 7: 36322-36333.

[19]赵雁斌, 苏锦钿. 基于变分自编码器的多隐变量双向推理模型[J]. 计算机科学, 2023, 50(10): 176-183.

[20]Goodfellow l, Pouget-Abadie J, Mirza M, et al. Generative adversarial nets [C]// Advances in Neural Information Processing Systems, 2014: 2672-2680.

[21]Shu L, Yongchun Z. MIT technology review released“top 10 global breakthrough technologies”in 2018[J]. China Scitechnology Business, 2018, 3: 34-37.

[22]Nowozin S, Cseke B, Tomioka R. f-GAN: Training generative neural samplers using variational divergence minimization[J]. Advances in Neural Information Processing Systems, 2016, 6: 271-279.

[23]Mao X, Li Q, Xie H, et al. Least squares generative adversarial networks[C]. IEEE International Conference on Computer Vision, 2017: 2813-2821.

[24]Arjovsky M, Chintala S, Bottou L. Wasserstein GAN[J]. arXiv preprint arXiv: 1701.07875, 2017.

[25]Mirza M, Osindero S. Conditional generative adversarial nets[J]. arXiv preprint arXiv: 1411.1784, 2014.

[26]柯研, 王希龙, 郑钰辉. 深度卷积生成对抗网络结构[J]. 电子技术与软件工程, 2018, 24: 5-6.

[27]Ullah H, Khan S D, Ullah M, et al. Generative adversarial networks: A short review [C]. IS and T International Symposium on Electronic Imaging Science and Technology, 2020: 3121-3127.

[28]汪美琴, 袁伟伟, 张继业. 生成对抗网络GAN的研究综述[J]. 计算机工程与设计, 2021, 42(12): 3389-3395.

[29]赵树阳, 李建武. 基于生成对抗网络的低秩图像生成方法[J]. 自动化学报, 2018, 44(05): 829-839.

[30]Hamza Pehlivan, Yusuf Dalva, Aysegul Dundar. StyleRes: Transforming the Residuals for Real Image Editing With StyleGAN. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2023: 1828-1837.

[31]Huang Y, Chen J, Huang D. UFPMP-Det: Toward Accurate and Efficient Object Detection on Drone Imagery. In Proceedings of the AAAI Conference on Artificial Intelligence, Vancouver, BC, Canada, 2022, 36: 1026-1033.

[32]Liu Z, Gao G, Sun L, Fang Z. HRDNet: High-Resolution Detection Network for Small Objects. In Proceedings of the 2021 IEEE International Conference on Multimedia and Expo (ICME), Shenzhen, China, 2021: 1-6.

[33]Qingyun F, Dapeng H, Zhaokui W. Cross-Modality Fusion Transformer for Multispectral Object Detection. arXiv preprint arXiv: 2111.00273, 2021.

[34]Yang F, Fan H, Chu P, et al. Clustered Object Detection in Aerial Images. In Proceedings of the IEEE/CVF International Conference on Computer Vision, Seoul, Republic of Korea, 2019: 8311-8320.

[35]Liang X, Zhang J, Zhuo L, et al. Small Object Detection in Unmanned Aerial Vehicle Images Using Feature Fusion and Scaling-Based Single Shot Detector with Spatial Context Analysis. IEEE Trans. Circuits Syst. Video Technol., 2019, 30: 1758-1770.

[36]林健巍. YOLO图像检测技术综述[J]. 福建电脑, 2019, 35(09): 80-83.

[37]李昊, 于虹, 饶桐, 等. 基于DS证据理论的高光谱图像融合分类[J]. 哈尔滨理工大学学报, 2023, 28(04): 33-41.

[38]鲁晓涵, 李洋, 贾耀东, 等. 基于GAN改进的红外与可见光图像融合算法研究[J/OL]. 电光与控制, 2024: 1-7.

[39]袁慧敏, 张绪红. 目标检测算法综述[J]. 科技经济导刊, 2021, 6: 52-55.

[40]Chen Y, Li R, Li R. HRCP: High-Ratio Channel Pruning for Real-Time Object Detection on Resource-Limited Platform. Neurocomputing, 2021, 463: 155-167.

[41]Ringwald T, Sommer L, Schumann A, et al. UAV-Net: A Fast Aerial Vehicle Detector for Mobile Platforms. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, Long Beach, CA, USA, 2019:544-552.

[42]Deng F, Xie Z, Mao W, et al. Research on Edge Intelligent Recognition Method Oriented to Transmission Line Insulator Fault Detection. Int. J. Electr. Power Energy Syst., 2022, 139: 108054.

[43]Li Y, Yuan H, Wang Y, et al. GGT-YOLO: A Novel Object Detection Algorithm for Drone-Based Maritime Cruising. Drones 2022, 6: 335.

[44]Jocher G, Chaurasia A, Qiu J. YOLO by Ultralytics. Accessed: February 30, 2023.

[45]赵继达, 甄国涌, 储成群. 基于YOLOv8的无人机图像目标检测算法[J/OL]. 计算机工程, 2024: 1-10.

[46]LeCun Y, Bengio Y, Hinton G, Deep learning, Nature, vol.521, pp.436-444, 2015.

[47]Maayan F A, Idit D, Eyal K, et al. GAN base-d synthetic medical image augmentation for increased CNN performance in liver lesion classification. Neuro-computing, 2018, 321: 321-331.

[48]Shao S, Wang P, Yan R. Generative adversarial networks for data augmentation in machine fault diagnosis. Computers in Industry, 2019, 106: 85-93.

[49]丁煦, 王栋, 翟华. 半监督阶梯网络和GAN在滚动轴承故障诊断的应用[J]. 机械设计与制造, 2022, 05: 152-156.

[50]Juan R Terven, Diana M Cordova-Esparza. A Comprehensive Review of YOLO: From YOLOv1 and Beyond[J]. arXiv preprint arXiv: 2304.0050, 2023.

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

[52]牟晋娟. 深度神经网络中激活函数的研究[J]. 电脑编程技巧与维护, 2019, 0(12): 59-61.

[53]Szegedy C, Liu W, Jia Y, et al. Going deeper with convolutions. in Proceedings of the IEEE conference on computer vision and pattern recognition, 2015: 1-9.

[54]Lin M, Chen Q, Yan S. Network in network. arXiv preprint arXiv: 1312.4400, 2013.

[55]Everingham M, Van Goo l L, Williams C K, et al. The pascal visual object classes (voc) challenge. International journal of computer vision, 2010, 88(2): 303-338.

[56]Ren S, He K, Girshick R, et al. Faster R-CNN: Towards real-time object detection with region proposal networks. Advances in neural information processing systems, 2015: 28.

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

[58]Krasin I, Duerig T, Alldrin N, et al. Open images: A public dataset for large-scale multi-label and multi-class image classification. github. com/open images, 2017, 2(3): 18.

[59]夏千涵, 何胜煌, 吴元清, 等. 基于可学习记忆特征金字塔网络的小样本目标检测[J]. 计算机与现代化, 2023, 12: 7-13.

[60]Jocher G. YOLOv5 by Ultralytics. Accessed, 2023.

[61]Zheng Z, Wang P, Liu W, et al. Distance-iou loss: Faster and better learning for bounding box regression. in Proceedings of the AAAI conference on artificial intelligence, 2020, 34: 12993-13000.

[62]张恺翊, 张宁燕, 江志浩, 等. 深度学习模型中不同激活函数的性能分析[J]. 网络安全与数据治理, 2023, 42(S1):149-156.

[63]杨龙欢, 董效杰, 黄施懿, 等. 基于YOLOV5的轻量型绝缘子缺陷检测算法研究[J]. 通信与信息技术, 2023, 5: 13-18.

[64]Zhang H, Cisse M, Dauphin N Y, et al. mixup: Beyond empirical risk minimization. arXiv preprint arXiv: 1710.09412, 2017.

[65]Buslaev A, Iglovikov V I, Khvedchenya E, et al. Albumentations: Fast and flexible image augmentations. Information, 2020, 11: 2.

[66]张光华, 王福豹, 段渭军. 基于 DCGAN 的高分辨率天文图像生成研究[J]. 计算机仿真, 2019, 36(12): 200-204.

[67]SHAO S, WANG P, YAN R. Generative adversarial networks for data augmentation in machine fault diagnosis. Computers in Industry, 2019, 106: 85-93.

[68]尹玉婷, 肖秦琨. 基于深度卷积生成对抗网络的图像生成[J]. 计算机技术与发展, 2021, 31(4): 86-92.

[69]李云红, 朱绵云, 任劼, 等. 改进深度卷积生成式对抗网络的文本生成图像[J]. 北京航空航天大学学报, 2023, 49(08): 1875-1883.

[70]王士斌, 高梓雕, 刘栋. 一种基于有限数据的改进DCGAN图像生成方法[J]. 河南师范大学学报（自然科学版）, 2023, 51(6): 39-46.

[71]Dewi C, Chen R C, Liu Y T, et al. Synthetic Data generation using DCGAN for improved traffic sign recognition. Neural Comput & Applic 34, 2022: 21465-21480.

[72]戚银城, 郎静宜, 赵振兵, 等. 结合注意力机制的相对GAN螺栓图像生成[J]. 电测与仪表, 2019,56(19): 64-69.

[73]矫红岩, 杨彦利. 基于DCGAN 的绝缘子图像生成方法[J]. 自动化与仪表, 2021, 36(01): 5-9.

[74]白由路, 杨俐苹, 王磊, 等. 农业低空遥感技术及其应用前景[J]. 农业网络信息, 2010, 1: 5-7.

[75]Tzutalin D. LabelImg[J]. GitHub repository, 2015:6.

[76]He K, Zhang X, Ren S, et al. Spatial Pyramid Pooling in Deep Convolutional Networks for Visual Recognition. IEEE Trans.Pattern Anal. Mach., 2015, 37: 1904-1916.

[77]Liu S, Qi L, Qin H, et al. Path Aggregation Network for Instance Segmentation. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, Salt Lake City, UT, USA, 2018: 8759-8768.

[78]王传昱, 李为相, 陈震环. 基于语音和视频图像的多模态情感识别研究[J]. 计算机工程与应用, 2021, 57(23): 163-170.

[79]Feng C, Zhong Y, Gao Y, et al. TOOD: Task-Aligned One-Stage Object Detection. In Proceedings of the 2021 IEEE International Conference on Computer Vision (ICCV), Montreal, QC, Canada, 2021: 3490-3499.

[80]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.

[81]Woo S, Park J, Lee J Y, et al. CBAM: Convolutional block attention module[C]// Computer Vision-ECCV 2018. Berlin, Heidelberg, Germany: Springer International Publishing, 2018: 3-19.

[82]Hou Q, Zhou D, Feng J. Coordinate attention for efficient mobile network design, In Proceedings of the 2021 IEEE International Conference on Computer Vision (ICCV), 2021: 13713-13722.

[83]Li Y, Shao Z, Hoffmann N. Global Attention Mechanism: Retain Information to Enhance Channel-Spatial Interactions[J]. arXiv preprint arXiv: 2112.05561, 2021.

[84]胡瑶, 李进, 王远军. 基于多层感知机和多尺度特征提取的肝脏分割网络[J]. 光学技术, 2023, 49(01): 105-112.

[85]Rezatofighi H, Tsoi N, Gwak J, et al. Generalized intersection over union: a metric and a loss for bounding box regression[C]// 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), June 15-20, 2019, Long Beach, CA, USA. New York: IEEE Press, 2019: 658-666.

[86]Zhang Y F, Ren W, Zhang Z, et al. Focal and efficient IOU loss for accurate bounding box regression. Neurocomputing, 2022, 506: 146-157.

[87]Gevorgyan Z. SIoU Loss: More Powerful Learning for Bounding Box Regression. arXiv preprint arXiv: 2205.12740, 2022.

[88]Tong Z J, Chen Y H, Xu Z W, et al. Wise-IoU: bounding box regression loss with dynamic focusing mechanism[J]. arXiv preprint arXiv: 2301.10051, 2023.

[89]Bochkovskiy A, Wang C Y, Liao H Y M. YOLOv4: Optimal Speed and Accuracy of Object Detection. arXiv preprint arXiv: 2004.10934, 2020.

[90]Wang C Y, Bochkovskiy A, Liao H Y M. YOLOv7: Trainable bag-of-freebies sets new state-of-the-art for real-time object detectors. arXiv preprint arXiv:2207.02696, 2022.

[91]Du D, Zhu P, Wen L, et al. VisDrone-DET2019: The vision meets drone object detection in image challenge results[C]. Proceedings of the IEEE/CVF international conference on computer vision workshops, 2019.

[92]朱强军, 胡斌, 汪慧兰, 等. 基于轻量化YOLOv8s交通标志的检测[J/OL]. 图学学报, 2024: 1-10.

[93]Yang G, Wang J, Nie Z, et al. A Lightweight YOLOv8 Tomato Detection Algorithm Combining Feature Enhancement and Attention. Agronomy, 2023, 13(7): 1824.

[94]翟仲毅, 赵胤铎. DSconv-LSTM： 面向边缘环境的轻量化视频行为识别模型[J]. 中国传媒大学学报(自然科学版), 2021, 28(06): 17-22.

[95]张润梅, 贾振楠, 李佳祥, 等. 基于多感受野特征增强的改进EfficientDet遥感目标检测算法[J/OL]. 电光与控制, 2024: 1-15.

[96]Wang X, Girshick R, Gupta A, et al. Non-local neural networks. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, Salt Lake City, UT, USA, 2018: 7794-7803.

[97]魏龙, 胡建鹏, 张庚. 基于门控注意力的双通道情感分析及应用[J]. 计算机工程与应用, 2023, 59(10): 134-141.

[98]Tong X, Wei J, Sun B, et al. ASCU-Net: Attention Gate, Spatial and Channel Attention U-Net for Skin Lesion Segmentation. Diagnostics, 2021, 11: 501.

[99]陈俊. 基于PCIe3.0的高速数据采集与传输系统设计与实现[D]. 西安电子科技大学, 2020.