一种基于YOLOv4-TIA的林业害虫实时检测方法

doi:10.19678/j.issn.1000-3428.0060563

摘要/Abstract

摘要： 针对现有基于深度学习的林业昆虫图像检测方法存在检测精度低和检测速度慢的问题，提出一种结合改进PANet结构与三分支注意力机制的目标检测方法YOLOv4-TIA。通过对样本数量较少的昆虫类别进行数据增强，实现样本均衡分布。利用三分支注意力机制改进YOLOv4中的CSPDarkNet53骨干网络，同时通过旋转操作和残差变换建立维度间的依存关系，以提高有效的特征通道权重，在PANet结构上增加将跳跃连接与跨尺度连接相结合的特征融合方式，从而获取更丰富的语义信息和位置信息。在此基础上，采用Focal loss函数优化分类损失，解决正负样本不均衡的问题。实验结果表明，该方法的精确率和召回率分别达到85.9%和91.2%，相比SSD、Faster R-CNN、YOLOv4方法，其在保证检测速度的同时，能够有效提高检测精度，且实现对林业害虫的实时精确监测。

关键词: 林业害虫检测, YOLOv4模型, 深度学习, 三分支注意力, Focal loss函数, 加权双向特征金字塔网络

Abstract: The existing forestry insect image detection methods based on Deep Learning(DL) have some problems such as low detection accuracy and slow detection speed.To address these problems, a YOLOv4-TIA target detection method combining improved PANet structure with three branch attention mechanisms is proposed.By enhancing the data of insect categories with a small number of samples, a balanced distribution of samples is realized.The three branch attention mechanism is used to improve the CSPDarkNet53 backbone network in YOLOv4.In addition, the rotation operation and residual transformation are used to establish the dependency relationship between dimensions, to improve the effective feature channel weight.The feature fusion method, combining jump connection and efficient multi-directional cross-scale connection, is added to the PANet structure, to obtain richer semantic and location information.The Focal loss function is used to optimize the classification loss to solve the problem of imbalance between positive and negative samples.The experimental results show that the accuracy and recall of the YOLOv4-TIA method reach 85.9% and 91.2%, respectively.Compared with SSD, Faster R-CNN, and YOLOv4 methods, the proposed method improves detection accuracy on the premise that detection speed can be ensured, and real-time monitoring of forestry pests can be realized.

Key words: forestry pests detection, YOLOv4 model, Deep Learning(DL), three branch attention, Focal loss function, weighted Bi-directional Feature Pyramid Network(BiFPN)

中图分类号:

TP399

候瑞环, 杨喜旺, 王智超, 高佳鑫. 一种基于YOLOv4-TIA的林业害虫实时检测方法[J]. 计算机工程, 2022, 48(4): 255-261.

HOU Ruihuan, YANG Xiwang, WANG Zhichao, GAO Jiaxin. A Real-Time Detection Method for Forestry Pests Based on YOLOv4-TIA[J]. Computer Engineering, 2022, 48(4): 255-261.

http://www.ecice06.com/CN/Y2022/V48/I4/255

图/表 15

20230131202310

20230131202314

20230131202317

20230131202320

20230131202323

20230131202326

20230131202329

20230131202332

20230131202336

20230131202339

20230131202342

20230131202345

20230131202348

20230131202351

20230131202354

参考文献

[1] 王正宏.基于分形理论的林业害虫识别与分类方法的研究[D].哈尔滨:东北林业大学, 2010. WANG Z H.Research on recognition and classification method of forestry pests based on fractal theory[D].Ha'er bin:Northeast Forestry University, 2010.(in Chinese)
[2] 周爱明.基于深度学习的农业灯诱害虫自动识别与计数技术的研究[D].杭州:浙江理工大学, 2019. ZHOU A M.Automated recognition and counting technique for agricultural light- trap pests based on deep learning[D].Hangzhou:Zhejiang Sci-Tech University, 2019.(in Chinese)
[3] 潘春华, 肖德琴, 林探宇, 等.基于SVM和区域生长结合算法的南方主要蔬菜害虫分类识别[J].农业工程学报, 2018, 34(8):192-199. PAN C H, XIAO D Q, LIN T Y, et al.Classification and recognition for major vegetable pests in Southern China using SVM and region growing algorithm[J].Transactions of the Chinese Society of Agricultural Engineering, 2018, 34(8):192-199.(in Chinese)
[4] 李衡霞, 龙陈锋, 曾蒙, 等.一种基于深度卷积神经网络的油菜虫害检测方法[J].湖南农业大学学报(自然科学版), 2019, 45(5):560-564. LI H X, LONG C F, ZENG M, et al.A detectiong method for the rape pests based on deep convolutional neural network[J].Hunan Agricultural University(Natural Sciences), 2019, 45(5):560-564.(in Chinese)
[5] SIMONYAN K, ZISSERMAN A.Very deep convolutional networks for large-scale image recognition[EB/OL].[2020-12-10].https://arxiv.org/pdf/1409.1556.pdf.
[6] GIRSHICK R.Fast R-CNN[C]//Proceedings of IEEE International Conference on Computer Vision.New York, USA:IEEE Press, 2015:1440-1448.
[7] 李想.基于物联网的虫情监测系统[D].北京:北京林业大学, 2019. LI X.The pest monitoring system based on the Internet of Things[D].Beijing:Beijing Forestry University, 2019.(in Chinese)
[8] LIU W, ANGUELOV D, ERHAN D, et al.SSD:single shot multibox detector[C]//Proceedings of European Conference on Computer Vision.Berlin, Germany:Springer, 2016:21-37.
[9] BOCHKOVSKIY A, WANG C Y, LIAO H Y M.YOLOv4:optimal speed and accuracy of object detection[EB/OL].[2020-12-10].https://arxiv.org/abs/2004.10934.
[10] SONG H, WILLI M, THIAGARAJAN J J, et al.Triplet network with attention for speaker diarization[EB/OL].[2020-12-10].https://arxiv.org/pdf/1808.01535.pdf.
[11] LIU S, QI L, QIN H, et al.Path aggregation network for instance segmentation[C]//Proceedings of IEEE Conference on Computer Vision and Pattern Recognition.New York, USA:IEEE Press, 2018:8759-8768.
[12] REDMON J, DIVVALA S, GIRSHICK R, et al.You only look once:unified, real-timeobject detection[C]//Proceedings of IEEE Conference on Computer Vision and Pattern Recognition.New York, USA:IEEE Press, 2016:779-788.
[13] 高宗, 李少波, 陈济楠.基于YOLO网络的行人检测方法[J].计算机工程, 2018, 44(5):215-219, 226. GAO Z, LI S B, CHEN J N.Pedestrian detection method based on YOLO network[J].Computer Engineering, 2018, 44(5):215-219, 226.(in Chinese)
[14] REDMON J, FARHADI A.YOLOV3:an incremental improvement[EB/OL].[2020-12-10].https://arxiv.org/pdf/1804.02767.pdf.
[15] WU D, LV S, JIANG M, et al.Using channel pruning-based YOLOv4 deep learning algorithm for the real-time and accurate detection of apple flowers in natural environments[J].Computers and Electronics in Agriculture, 2020, 178:105742.
[16] 鞠默然, 罗海波, 王仲博, 等.改进的YOLOV3算法及其在小目标检测中的应用[J].光学学报, 2019, 39(7):253-260. JU M R, LUO H B, WANG Z B, et al.Improved YOLOV3 algorithm and its application in small target detection[J].Acta Optica Sinica, 2019, 39(7):253-260.(in Chinese)
[17] MISRA D, NALAMADA T, ARASANIPALAI A U, et al.Rotate to attend:convolutional triplet attention module[C]//Proceedings of IEEE/CVF Winter Conference on Applications of Computer Vision.New York, USA:IEEE Press, 2021:3139-3148.
[18] TAN M, PANG R, LE Q V.Efficientdet:scalable and efficient object detection[C]//Proceedings of IEEE/CVF Conference on Computer Vision and Pattern Recognition.New York, USA:IEEE Press, 2020:10781-10790.
[19] 卞永明, 高飞, 李梦如, 等.结合Kmeans++聚类和颜色几何特征的火焰检测方法[J].中国工程机械学报, 2020, 18(1):1-6. BIAN Y M, GAO F, LI M R, et al.Fire detection method using Kmeans++ clustering and features of mixed color and geometry[J].Chinese Journal of Coustruction Machinery, 2020, 18(1):1-6.(in Chinese)
[20] YU Z, SHEN Y, SHEN C.A real-time detection approach for bridge cracks based on YOLOv4-FPM[J].Automation in Construction, 2021, 122:1-10.
[21] 杨明欣, 张耀光, 刘涛.基于卷积神经网络的玉米病害小样本识别研究[J].中国生态农业学报, 2020, 28(12):1924-1931. YANG M X, ZHANG Y G, LI T.Research on small sample recognition of corn disease based on convolutional neural network[J].Chinese Journal of Eco-Agriculture, 2020, 28(12):1924-1931.(in Chinese)

选择文件类型/文献管理软件名称

选择包含的内容