SHANG Zeng-qiang,YANG Dong-fu,MA Zhi-pu.Automatic Identification of Soybean Leaf Diseases Based on UAV Image and Deep Convolution Neural Network[J].Soybean Science,2021,40(05):662-668.[doi:10.11861/j.issn.1000-9841.2021.05.0662]
基于深度卷积神经网络的大豆叶片多种病害分类识别
- Title:
- Automatic Identification of Soybean Leaf Diseases Based on UAV Image and Deep Convolution Neural Network
- 文献标志码:
- A
- 摘要:
- \为实现复杂田间背景下快速高效分类识别多种大豆叶片病害图像,以准确性、训练时间和训练误差为深度学习模型性能指标,对比评估不同深度卷积神经网络模型。首先将无人机收集到的大豆叶片病害图像数据集按7∶3的比例分为训练集与测试集。为扩充数据图像,对训练集原图进行数据增强。基于不同权重的微调和迁移学习训练策略,采用Inception-v3、VGG-19、ResNet-50和Xception 4种深度神经网络模型对大豆叶片多种病害进行分类识别,并进行田间验证试验。结果表明,在FT 75%训练策略下的Inception-v3模型准确性最高,为99.04%;与其他模型相比,FT 100%和FT 75%训练策略下的深度学习模型显示出较高的准确性和较低的训练误差,但训练时间也更长。将训练好的Inception-v3 FT 75%模型用于计算机视觉系统的结果表明该系统可有效实现田间复杂背景下大豆叶片病害的智能识别。
- Abstract:
- In order to realize fast and efficient classification and recognition of soybean leaf disease images under complex field background,convolution neural network models with different depth were compared and evaluated with accuracy,training time and learning error as performance indexes of deep learning model.Firstly,the soybean leaf disease image data set collected by UAV was divided into training set and test set according to the ratio of 7∶3.In order to expand the data image,the original image of training set was enhanced.Four kinds of deep neural network models,including perception-v3,vgg-19,resnet-50 and xception,were used to test the models based on different weights of fine-tuning and transfer learning training strategies,and field validation experiments were carried out.The results showed that the accuracy of perception-v3 model was the highest (99.04%) under the FT 75% training strategy;Compared with other models,FT 100% and FT 75% deep learning models showed higher accuracy and lower learning error,but the training time was longer.The trained perception-v3 ft 75% model was applied to the computer vision system,and the results showed that the system could effectively realize the intelligent recognition of soybean leaf diseases under the complex field background.
参考文献/References:
[1]查霆,钟宣伯,周启政,等.我国大豆产业发展现状及振兴策略[J].大豆科学,2018,37(3):458-463.(Zha T,Zhong X B,Zhou Q Z,et al.Development status of China’s soybean industry and strategies of revitalizing[J].Soybean Science,2018,37(3):458-463.)[2]Pires R D L,Goncalves D N,Oruê J P M,et al.Local descriptors for soybean disease recognition[J].Computers and Electronics in Agriculture,2016,125:48-55.[3]Oerke E C,Steiner U,Dehne H W,et al.Thermal imaging of cucu-mber leaves affected by downy mildew and environmental conditions[J].Journal of Experimental Botany,2006,57(9):2121-2132.[4]Mahlein A K,Oerke E C,Steiner U,et al.Recent advances in sen-sing plant diseases for precision crop protection[J].European Journal of Plant Pathology,2012,133(1):197-209.[5]Weiss U,Biber P,Laible S,et al.Plant species classification using a 3D LIDAR sensor and machine learning[C]∥2010 Ninth International Conference on Machine Learning and Applications.IEEE,2010:339-345.[6]Chelladurai V,Karuppiah K,Jayas D S,et al.Detection of Calloso-bruchus maculatus(F.) infestation in soybean using soft X-ray and NIR hyperspectral imaging techniques[J].Journal of Stored Products Research,2014,57:43-48.[7]Oppenheim D,Shani G,Erlich O,et al.Using deep learning for image-based potato tuber disease detection[J].Phytopathology,2019,109(6):1083-1087.[8]郝惠敏,梁永国,武海彬,等.对称点模式-深度卷积神经网络的红外光谱识别方法[J].光谱学与光谱分析,2021,41(3):782-788.(Hao H M,Liang Y G,Wu H B,et al.Infrared spectrum recognition method based on symmetrized dot patterns coupled with deep convolutional neural network[J].Spectroscopy and Spectral Analysis,2021,41(3):782-788.)[9]黄双萍,孙超,齐龙,等.基于深度卷积神经网络的水稻穗瘟病检测方法[J].农业工程学报,2017,33(20):169-176.(Huang S P,Sun C,Qi L,et al.Rice panicle blast identification method based on deep convolution neural network[J].Transactions of the Chinese Society of Agricultural Engineering,2017,33(20):169-176.)[10]曹英丽,江凯伦,于正鑫,等.基于深度卷积神经网络的水稻纹枯病检测识别[J].沈阳农业大学学报,2020,51(5):568-575.(Cao Y L,Jiang K L,Yu Z X,et al.Detection and recognition of rice sheath blight based on deep convolutional neural network[J].Journal of Shenyang Agricultural University,2020,51(5):568-575.)[11]金秀,卢杰,傅运之,等.基于深度卷积神经网络的小麦赤霉病高光谱病症点分类方法[J].浙江农业学报,2019,31(2):315-325.(Jin X,Lu J,Fu Y Z,et al.A classification method for hyperspectral imaging of Fusarium head blight disease symptom based on deep convolutional neura network[J]. Acta Agriculturae Zhejiangensis, 2019, 31(2):315-325.)[12]Achanta R,Shaji A,Smith K,et al.SLIC superpixels compared to state-of-the-art superpixel methods[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2012,34(11):2274-2282.[13]张兴政,孙一闻,吕世翔,等.大豆霜霉病抗病机制和防治研究[J].大豆科学,2020,39(1):152-159.(Zhang X Z,Sun Y W,Lyu S X.Research progress and prospect on disease resistance mechanism and control of soybean downy mildew[J].Soybean Science,2020,39(1):152-159.)[14]柳建,姜文涛,安保宁,等.大豆白粉病病原菌鉴定[J].植物病理学报,2015,45(5):548-551.(Liu J,Jiang W T,An B N,et al.The identification of soybean pow dery mildew fungus[J].Acta Phytopathologica Sinica,2015,45(5):548-551.)[15]Szegedy C,Vanhoucke V,Ioffe S,et al.Rethinking the inception architecture for computer vision[C]∥Proceedings of the IEEE conference on computer vision and pattern recognition.IEEE,2016:2818-2826.[16]Wei Y C,Yuan Q Q,Shen H F,et al.Boosting the accuracy of multispectral image pansharpening by learning a deep residual network[J].IEEE Geoscience and Remote Sensing Letters,2017,14(10):1795-1799.[17]Zhang K,Zuo W M,Chen Y J,et al.Beyond a gaussian denoiser:Residual learning of deep CNN for image denoising[J].IEEE Transactions on Image Processing,2017,26(7):3142-3155.[18]Chollet F.Xception:Deep learning with depthwise separable convolutions[C]∥Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition,2017:1251-1258.[19]陶砾,杨朔,杨威.深度学习的模型搭建及过拟合问题的研究[J].计算机时代,2018(2):14-21.(Tao S,Yang S,Yang W.Research on the model building and over-fitting of deep learning[J].Computer Era,2018(2):14-21.)[20]吉海彦,任占奇,饶震红.基于高光谱成像技术的不同产地小米判别分析[J].光谱学与光谱分析,2019,39(7):2271-2277.(Ji H Y,Ren Z Q,Rao Z H.Discriminant analysis of millet from different origins based on hyperspectral imaging technology[J]. Spectroscopy and Spectral Analysis, 2019,39(7):2271-2277.)[21]樊湘鹏,周建平,许燕.数据集对基于深度学习的作物病害识别有效性影响[J].中国农机化学报,2021,42(1):192-200.(Fan X P,Zhou J P,Xu Y,et al.Influence of data sets on the effectiveness of crop disease recognition based on deep learning[J].Journal of Chinese Agricultural Mechanization,2021,42(1):192-200.)[22]Mohanty S P,Hughes D P,Salathé M.Using deep learning for image-based plant disease detection[J].Frontiers in Plant Science,2016,7:1419.
相似文献/References:
[1]马 晓,董天亮,钟闻宇,等.基于改进ConvNeXt的大豆叶片病害分类研究[J].大豆科学,2023,42(06):733.[doi:10.11861/j.issn.1000-9841.2023.06.0733]
备注/Memo
收稿日期:2021-04-06