Rice pest recognition based on multi-scale feature extraction depth residual network
-
摘要:目的
在水稻生产过程中，针对不同虫害需要采用不同的防治方案，水稻害虫的准确识别分类是制定针对性防治方案的前提。
方法采用深度学习结合机器视觉的方法，基于Res2Net结构提出了一种多尺度特征提取的深度残差网络，通过准确地提取害虫特征实现复杂自然背景下的水稻害虫识别；采用改进的残差结构，使用等级制的类残差连接取代了原本的3×3卷积核，增加了每个网络层的感受野，可以更细粒度地提取多尺度特征。
结果本网络训练的模型能够有效地识别自然背景下的水稻害虫，在自建的包含22类常见水稻害虫的图像数据集上，平均识别准确率达到了92.023%，优于传统的ResNet、VGG等网络。
结论本文提出的模型可应用于水稻虫情自动监测系统，为实现水稻害虫虫情的机器视觉监测提供参考。
Abstract:ObjectiveIn the process of rice production, different control schemes need to be adopted for different pests. The accurate identification and classification of rice pests are the premise of formulating targeted control program.
MethodA deep residual network of multi-scale feature extraction was proposed based on the Res2Net structure, which could extract pest characteristics more accurately and realize rice pest identification in complex natural background. This network adopted an improved residual structure, replaced the original convolutional kernel with hierarchical class residual connections, increased the sensing field of each network layer, and could extract multi-scale features at a more fine-grained degree.
ResultThe results showed that the model trained by this network could effectively identify rice pests in natural background. The average recognition accuracy of proposed model reached 92.023% on the self-built image dataset containing 22 kinds of the common rice pests, which was superior to the traditional ResNet, VGG and other networks.
ConclusionThis network can be applied to the automatic monitoring system of rice insect status, which provides a reference for the realization of machine vision monitoring of rice pests.
-
Keywords:
- Rice pest /
- Res2Net /
- Residual network /
- Deep learning /
- Image recognition /
- Image classification /
- Multi-scale feature
-
-
![]()
图 1 传统卷积结构(a)和残差卷积结构(b)
Figure 1. Conventional convolutional structure(a) and residual convolutional structure (b)
![]()
图 3 多尺度特征提取的深度残差网络结构图
Figure 3. Network structure diagram of depth residual network based on multi-scale feature extraction
![]()
图 7 ResNet50与本文网络_s=6的整体准确率和各种害虫准确率对比图
Figure 7. Comparisons of the overall accuracy rates and the accuracy rates of various pests between ResNet50 and this network _s=6
表 1 各类网络模型的性能对比
Table 1 Performance comparison of various network models
模型1)Model 准确率/% Accuracy 模型大小/MB Model size t识别2)/ms Recognition time t训练3)/min Training time AlexNet 80.566 233.0 61.81 135.3 ResNet18 88.397 42.7 60.88 133.1 ResNet34 89.123 81.3 89.73 138.1 ResNet50 90.345 90.1 136.60 145.8 VGG16 87.382 528.0 205.41 154.1 s=4 90.935 90.7 169.40 148.8 s=6 92.023 134.0 215.42 166.5 s=8 91.588 177.0 271.23 172.5 1) s=4、6或8分别代表特征维为4、6或8的本文网络模型；2)单张图片识别时间；3)迭代100次的训练时间 1) s=4, 6 or 8 represents the network model of this paper with characteristic dimension of 4, 6 or 8 respectively; 2) Recognition time of single image; 3) Training time of 100 epoches 
下载: 导出CSV
-
[1] 梁勇, 邱荣洲, 李志鹏, 等. 基于深度学习的水稻主要害虫识别方法[J/OL]. 农业机械学报, (2022-05-19)[2022-06-08]. http://kns.cnki.net/kcms/detail/11.1964.S.20220519.0919.002.html. [2] 杨红云, 肖小梅, 黄琼, 等. 基于卷积神经网络和迁移学习的水稻害虫识别[J]. 激光与光电子学进展, 2022, 59(16): 333-340. [3] 姚青, 谷嘉乐, 吕军, 等. 改进RetinaNet的水稻冠层害虫为害状自动检测模型[J]. 农业工程学报, 2020, 36(15): 182-188. doi: 10.11975/j.issn.1002-6819.2020.15.023 [4] 刘德营, 王家亮, 林相泽, 等. 基于卷积神经网络的白背飞虱识别方法[J]. 农业机械学报, 2018, 49(5): 51-56. doi: 10.6041/j.issn.1000-1298.2018.05.006 [5] 谢成军, 李瑞, 董伟, 等. 基于稀疏编码金字塔模型的农田害虫图像识别[J]. 农业工程学报, 2016, 32(17): 144-151. doi: 10.11975/j.issn.1002-6819.2016.17.020 [6] XIAO D Q, FENG J Z, LIN T Y, et al. Classification and recognition scheme for vegetable pests based on the BOF-SVM model[J]. International Journal of Agricultural and Biological Engineering, 2018, 11(3): 190-196. doi: 10.25165/j.ijabe.20181103.3477
[7] 张博, 张苗辉, 陈运忠. 基于空间金字塔池化和深度卷积神经网络的作物害虫识别[J]. 农业工程学报, 2019, 35(19): 209-215. doi: 10.11975/j.issn.1002-6819.2019.19.025 [8] 程科, 孙玮, 高尚. 一种水稻田稻飞虱图像识别的混合算法[J]. 农机化研究, 2015, 37(11): 17-21. doi: 10.3969/j.issn.1003-188X.2015.11.004 [9] 鲍文霞, 吴德钊, 胡根生, 等. 基于轻量型残差网络的自然场景水稻害虫识别[J]. 农业工程学报, 2021, 37(16): 145-152. doi: 10.11975/j.issn.1002-6819.2021.16.018 [10] LU X Y, YANG R, ZHOU J, et al. A hybrid model of ghost-convolution enlightened transformer for effective diagnosis of grape leaf disease and pest[J]. Journal of King Saud University (Computer and Information Sciences), 2022, 34(5): 1755-1767. doi: 10.1016/j.jksuci.2022.03.006
[11] LI H, LI S F, YU J G, et al. Plant disease and insect pest identification based on vision transformer[C]// SPIE. International Conference on Internet of Things and Machine Learning (IoTML 2021). Shanghai: SPIE , 2021: 194-201.
[12] YUAN L, CHEN Y, WANG T, et al. Tokens-to-token vit: Training vision transformers from scratch on imagenet[C]//IEEE/CVF. IEEE/CVF International Conference on Computer Vision. Montreal : IEEE/CVF, 2021: 558-567.
[13] HE K M, ZHANG X Y, REN S Q, et al. Deep Residual Learning for Image Recognition[C]//IEEE. IEEE Conference on Computer Vision and Pattern Recognition. CA: IEEE, 2016: 770-778.
[14] GAO S H, CHENG M M, ZHAO K, et al. Res2Net: A new multi-scale backbone architecture[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2021, 43(2): 652-662. doi: 10.1109/TPAMI.2019.2938758
[15] HE T, ZHANG Z, ZHANG H, et al. Bag of tricks for image classification with convolutional neural networks [C]//IEEE/CVF. IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR 2019). Los Angeles: IEEE/CVF, 2019: 558-567.
[16] IOFFE S, SZEGEDY C. Batch normalization: Accelerating deep network training by reducing internal covariate shift[C]//JMLR. Proceedings of the 32nd International Conference on Machine Learning. Lille: W&CP, 2015: 448-456.
[17] LIN M, CHEN Q, YAN S. Network in network[J/OL]. Neural and Evolutionary Computing, (2013-10-16)[2022-04-26]. https://doi.org/10.48550/arXiv.1312.4400.
[18] WU X P, ZHAN C, LAI Y K, et al. IP102: A large-scale benchmark dataset for insect pest recognition[C]// IEEE/CVF. IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR 2019) . Los Angeles: IEEE/CVF, 2019: 8787-8796.
[19] 中国科技出版传媒股份有限公司. 中国生物志库[DB/OL]. (2008-08-28)[2022-04-26]. https://species.sciencereading.cn/biology/v/biologicalIndex/122.html. [20] 彭旭, 饶元, 乔焰. 基于宽度卷积神经网络的异常农情数据检测方法[J]. 华南农业大学学报, 2022, 43(2): 113-121. doi: 10.7671/j.issn.1001-411X.202103050 [21] PENG Z L, HUANG W , GU S Z , et al. Conformer: Local features coupling global representations for visual recognition[C]//IEEE/CVF. Proceedings of the IEEE/CVF International Conference on Computer Vision. Los Angeles: IEEE/CVF, 2021: 367-376.
计量
- 文章访问数: 529
- HTML全文浏览量: 8
- PDF下载量: 38
