刘三雄

• 1:山西大学物理与电子工程学院

摘要(Abstract)：

结核病已经被列为全球十大主要死亡原因之一，针对痰涂片图像背景复杂，结核杆菌目标尺寸小等难题，提出基于痰涂片图像的单阶段YOLOv7结核杆菌检测网络。骨干网络中，分别在不同尺度的高效聚合网络后嵌入CBAM注意力模块，以提取不同尺度的特征，在颈部网络中引入分离合并操作以改进MPConv结构，减少深层网络因卷积下采样引起的图像特征损失，并在头部网络中引进高斯分布距离来避免因边界框重叠而引起的小目标漏检，得到一种精度高、检测速度快的结核杆菌方法。实验表明，改进模型的均值平均精度达到了87.8%,相比基线网络模型提升5.1%,且优于同类算法，对推进结核杆菌的智能化检测具有重要意义。

关键词(KeyWords)： 结核杆菌检测;YOLOv7;注意力机制;分离合并操作;高斯分布

Abstract:

Keywords:

基金项目(Foundation): 国家重点研发计划（编号：2022ZD0118300）

作者(Author): 刘三雄

DOI: 10.20064/j.cnki.2095-347X.2024.03.004

参考文献(References)：

• [1] WHO G.Global tuberculosis report 2020[EB/OL].(2020-10-16)/[2023-05-25].http://appswho.int/iris/handle/10665/336069.
• [2] An L,Peng K,Yang X,et al.E-TBNet:Light Deep Neural Network for automatic detection of tube-rculosis with X-ray DR Imaging[J].Sensors,2022,22(3):821.
• [3] Negi S S,Singh P,Sharma K.Molecular Diagnosis of Tuberculosis[M]//Diagnosis of Mycobacterium.Singapore:Springer Nature Singapore,2023:65-85.
• [4] Panicker R O,Kalmady K S,Rajan J,et al.Automatic detection of tuberculosis bacilli from micros-copic sputum smear images using deep learning methods[J].Biocybernetics and Biomedical Engineering,2018,38(3):691-699.
• [5] Swetha K,Sankaragomathi B,Thangamalar J B.Convolutional neural network based automated detec-tion of mycobacterium bacillus from sputum images[C]//2020 International Conference on Inventive Comp-utation Technologies(ICICT).IEEE,2020:293-300.
• [6]Yusoff A K M,Raof R A A,Mahrom N,et al.Enhancement and Segmentation of Ziehl Neelson Sputum Slide Images using Contrast Enhancement and Otsu Threshold Technique[J].Journal of Advanced Research in Applied Sciences and Engineering Technology,2023,30(1):282-289.
• [7] Chen L C,Papandreou G,Schroff F,et al.Rethinking atrous convolution for semantic image segmentation[EB/OL].arXiv preprint arXiv:1706.05587.(2017-06-17)/[2023-05-25].https://arxiv.org/abs/1706.05587.
• [8] 蒋婷，沈旭东，陆伟，等.基于多特征融合的人脸颜值预测[J].网络新媒体技术，2017,6(2):7-13.
• [9] Yang S,Liu J,Lu S,et al.Collaborative learning of gesture recognition and 3D hand pose estimati-on with multi-order feature analysis[C]//Computer Vision-ECCV 2020:16th European Conference.Glasgo-w,UK:Springer International Publishing,2020:769-786.
• [10] 龙凌，陈浩，梁昊，等.基于轻量化YOLO网络的实时X射线焊缝缺陷检测[J].网络新媒体技术，2023,12(2):30-38.
• [11] Lin T Y,Goyal P,Girshick R,et al.Focal loss for dense object detection[C]//Proceedings of the IEEE international conference on computer vision.2017:2980-2988.
• [12] Ultralytics.YOLOv5[EB/OL].(2020-05-18)/[2023-05 -25].https://github.com/ultralytics/yolov5.
• [13] 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[EB/OL].arXiv preprint arXiv:2207.02696.(2022-07-06)/[2023-05-25].https://arxiv.org/abs/2207.02696.
• [14] Liu W,Anguelov D,Erhan D,et al.SSD:single shot multibox detector[C]//Proceedings of the European Conference on Computer Vision.Netherlands,Berlin,Heidelberg:Springer-Verlag,2016:21-37.
• [15] Ren S Q,He K M,Girshick R,et al.Faster RCNN:towards real-time object detection with reg-ion proposal networks[J].IEEE transactions on pattern analysis and machine intelligence,2017,39(6):1137-1149.
• [16] Cai Z,Vasconcelos N.Cascade r-cnn:Delving into high quality object detection[C]//Proceedings of the IEEE conference on computer vision and pattern recognition.2018:6154-6162.
• [17] He K,Gkioxari G,Dollár P,et al.Mask r-cnn[C]//Proceedings of the IEEE international conference on computer vision.2017:2961-2969.
• [18] He K M,Zhang X Y,Ren S Q,et al.Spatial pyramid pooling in deep convolutional networks for visual recognition[J].IEEE transactions on pattern analysis and machine intelligence,2015,37(9):1904-1916.
• [19] Kuok C P,Horng M H,Liao Y M,et al.An effective and accurate identification system of Mycoba-cterium tuberculosis using convolution neural networks[J].Microscopy research and technique,2019,82(6):709-719.
• [20] Woo S,Park J,Lee J Y,et al.CBAM:Convolutional block attention module[C]//Proceedings of the European conference on computer vision(ECCV).2018:3-19.
• [21] Niu Z Y.A review on the attention mechanism of deep learning[J].Neurocomputing,2021,452:48-62.
• [22] Wang J,Xu C,Yang W,et al.A normalized Gaussian Wasserstein distance for tiny object detection[EB/OL].arXiv preprint arXiv:2110.13389.(2021-10-26)/[2023-05-25].https://arxiv.org/abs/2110.13389.
• [23] Zheng Z,Wang P,Ren D,et al.Enhancing geometric factors in model learning and inference for ob-ject detection and instance segmentation[J].IEEE transactions on cybernetics,2021,52(8):8574-8586.
• [24] Zheng Z,Wang P,Liu W,et al.Distance-IoU loss:Faster and better learning for bounding box regr-ession[C]//Proceedings of the AAAI conference on artificial intelligence.2020,34(7):12993-13000.
• [25] Kaggle.Tuberculosis image dataset[DB/OL].(2020-01-16)/[2023-05-25].https://www.kaggle.com/saife245/tuberculosis-image-dataset.
• [26] Carion N,Massa F,Synnaeve G,et al.End-to-end object detection with transformers[C]//European conference on computer vision.Cham:Springer International Publishing,2020:213-229.
• [27] Liu S,Li F,Zhang H,et al.Dab-detr:Dynamic anchor boxes are better queries for detr[EB/OL].arXiv preprint arXiv:2201.12329.(2022-01-28)/[2023-05-25].https://arxiv.org/abs/2201.12329.
• [28] Bochkovskiy A,Wang C Y,Liao H Y M.YOLOv4:Optimal speed and accuracy of object detection[EB/OL].arXiv preprint arXiv:2004.10934.(2020-04-23)/[2023-05-25].https://arxiv.org/abs/2004.10934.