Hyperspectral Imaging Technology-based Early Diagnosis of a Serious Agaricus Bisporus Disease
-
摘要:
目的 有害疣孢霉菌（Mycogone perniciosa）引起的双孢蘑菇疣孢霉病，是破坏性极强的真菌类病害，本研究拟基于高光谱成像技术，建立双孢蘑菇疣孢霉病早期快速检测方法。 方法 对200个健康双孢蘑菇样本与200个染病双孢蘑菇样本采集全波段（401～1046 nm）可见/近红外高光谱图像信息，利用Savitzky-Golay卷积一阶求导、Savitzky-Golay卷积平滑（SG平滑）、多元散射校正（MSC）分别对360个波段（全波段）的高光谱图像信息进行预处理后，对比随机森林（Random forest，RF）、支持向量机（Support vector machine，SVM）和极限学习机（Extreme learning machine，ELM）3种模型对健康和染病双孢蘑菇鉴别准确度进行分析。 结果 3种鉴别模型的结果接近，其中，MSC-SVM模型检测效果最优，将原始测试集和预测集总体样本鉴别准确度分别由85.02%和87.38%提升至92.21%和91.04%。 结论 本研究建立的MSC-SVM模型可以有效提高基于全波段的双孢蘑菇疣孢霉病早期的鉴别准确度，同时，为进一步开发双孢蘑菇病害早期的快速无损鉴别设备提供了理论依据和方法。 Abstract:Objective A nondestructive, effective method was developed based on the hyperspectral imaging technology for early diagnosis of the highly destructive wet bubble disease on Agaricus Bisporus caused by Mycogone perniciosa. Method Information on the full band (401-1046nm) visible/near-infrared hyperspectral images on 200 healthy and 200 infected A. bisporus specimens was collected. After a preprocess using Savitzky-Golay 1st order derivative, Savitzky-Golay smoothing, or multiple scattering correction (MSC) on the obtained information of the 360 full bands, accuracy of the methodology in separating the healthy from the infected samples was scrutinized by using the Random Forest (RF), Support Vector Machine (SVM), and Extreme Learning Machine (ELM) models. Result The 3 models yielded similar results and the MSC-SVM combination had the best detection effect with the identification accuracy on the test set increased from 85.02% to 92.21%, and on the prediction set, from 87.38% to 91.04%. Conclusion The MSC-SVM model appeared to significantly improve the identification accuracy using the full band. It provided a basis for the development of rapid, nondestructive diagnostic device on the devastating disease of A. bisporus at early stage which has been conventionally conducted by expert visual examination, PCR analysis on the internal transcribed spacer gene, or traditional Koch's postulation. -
图 1 可见/近红外高光谱成像系统
注：1：CCD相机和扫描结构；2：便携式地物光谱成像仪；3：成像镜头；4：光源；5：样品；6：样品升降台；7：多功能暗箱；8：平板电脑。
Figure 1. Visible/near-infrared hyperspectral imaging system
Note: 1: CCD camera and scanning structures; 2: portable ground object spectral imager; 3: imaging lens; 4: light source; 5: sample; 6: sample lifting table; 7: multifunctional dark box; 8: iPad.
图 2 健康和染病双孢蘑菇高光谱平均光谱曲线
注：（a）：健康双孢蘑菇高光谱平均光谱曲线；（b）：染病双孢蘑菇高光谱平均光谱曲线。
Figure 2. Average hyperspectral curves on healthy and infected mushrooms
Note: (a): Average hyperspectral curve on healthy mushrooms; (b): Average hyperspectral curve on infected mushrooms.
图 3 双孢蘑菇高光谱曲线
注：（a）：原始高光谱曲线；（b）：SG卷积一阶求导处理后高光谱曲线；（c）：SG卷积平滑处理后高光谱曲线；（d）：MSC处理后高光谱曲线。
Figure 3. Hyperspectral curve of mushroom
Note: (a): Original hyperspectral spectra curve; (b): Hyperspectral spectra curve after SG 1st order derivative preprocessing; (c): Hyperspectral spectra curve after SG smoothing preprocessing; (d): Hyperspectral spectra curve after MSC preprocessing.
图 4 随机森林分类结果
注：（a）：NONE-RF分类结果；（b）：MSC-RF分类结果。
Figure 4. Random Forest classification
Note: (a): NONE-RF classification on full bands; (b): MSC-RF classification on full bands.
图 5 SVM分类结果
注：（a）：NONE-SVM分类结果；（b）：MSC-SVM分类结果。
Figure 5. SVM classification
Note: (a): NONE-SVM classification on 360 bands; (b): MSC-SVM classification on 360 bands.
图 6 ELM分类结果
注：（a）：NONE-ELM分类结果；（b）：MSC- ELM分类结果。
Figure 6. ELM classification
Note: (a): NONE-ELM classification on 360 bands; (b): MSC-ELM classification on 360 bands.
表 1 样本信息
Table 1. List of samples
接种时间
Processed time/d生长期
Stage健康/染病
Healthy/Infected数量
Number7 原基期
Primordial stage健康
Healthy50 7 原基期
Primordial stage染病
Infected50 9 菇蕾期
Mushroom bud stage健康
Healthy50 9 菇蕾期
Mushroom bud stage染病
Infected50 10 幼菇期
Young mushroom stage健康
Healthy50 10 幼菇期
Young mushroom stage染病
Infected50 11 小菇期
Little mushroom stage健康
Healthy50 11 小菇期
Little mushroom stage染病
Infected50 总计
Total400 
下载: 导出CSV
表 2 不同预处理与不同建模方法结果
Table 2. Results of different preprocessing and modeling methods
方法
Methods识别准确度 Identification accuracy/% 测试集 Test set 预测集 Prediction set 健康样本
Healthy samples染病样本
Infected samples总体样本
All samples健康样本
Healthy samples染病样本
Infected samples总体样本
All samplesNONE-RF 93.99 96.27 95.13 89.71 87.22 88.06 NONE-SVM 82.84 87.22 85.02 89.69 85.07 87.38 NONE-ELM 89.39 84.63 87.01 92.54 86.69 89.62 MSC-RF 92.24 90.86 91.55 91.25 87.42 89.34 MSC-SVM 92.71 91.71 92.21 90.56 91.52 91.04 MSC-ELM 91.77 89.18 90.48 89.48 92.26 90.87 SG smoothing--RF 92.88 93.34 93.11 87.68 88.81 88.25 SG smoothing--SVM 93.00 91.13 92.07 88.71 92.05 90.38 SG smoothing--ELM 88.47 89.81 89.14 87.90 91.09 89.50 SG 1st order derivative-RF 91.03 93.23 92.13 89.35 87.48 88.42 SG 1st order derivative-SVM 91.18 89.30 90.24 91.59 89.81 90.70 SG 1st order derivative-ELM 87.65 91.70 89.68 90.51 91.34 90.93
下载: 导出CSV
-
[1] 秦文韬, 王守现, 荣成博, 等. 我国食用菌病害发生与防控概况 [J]. 中国食用菌, 2020, 39(12):1−7.QIN W T, WANG S X, RONG C B, et al. Occurrence and management of edible fungus diseases in China [J]. Edible Fungi of China, 2020, 39(12): 1−7.(in Chinese) [2] 谭琦, 王镭, 王永红, 等. 不同地区有害疣孢霉菌株生物学特性研究 [J]. 食用菌学报, 1996, 3(1):46−50.TAN Q, WANG L, WANG Y H, et al. A study on biological characteristics in isolates of Mycogone perniciosa magn [J]. Acta Edulis Fungi, 1996, 3(1): 46−50.(in Chinese) [3] YANG Y, SOSSAH F L, LI Z, et al. Genome-wide identification and analysis of chitinase GH18 gene family in Mycogone perniciosa [J]. Frontiers in Microbiology, 2021, 11: 596719. doi: 10.3389/fmicb.2020.596719 [4] GEA F J, TELLO J C, NAVARRO M J. Efficacy and effects on yield of different fungicides for control of wet bubble disease of mushroom caused by the mycoparasite Mycogone perniciosa [J]. Crop Protection, 2010, 29(9): 1021−1025. doi: 10.1016/j.cropro.2010.06.006 [5] PANG L, WANG J H, MEN S, et al. Hyperspectral imaging coupled with multivariate methods for seed vitality estimation and forecast for Quercus variabilis [J]. Spectrochimica Acta Part A:Molecular and Biomolecular Spectroscopy, 2021, 245: 118888. doi: 10.1016/j.saa.2020.118888 [6] SKONECZNY H, KUBIAK K, SPIRALSKI M, et al. Fire blight disease detection for apple trees: Hyperspectral analysis of healthy, infected and dry leaves. [J]. Remote Sensing, 2020, 12(13): 2101. doi: 10.3390/rs12132101 [7] SOBEJANO-PAZ V, MIKKELSEN T N, BAUM A, et al. Hyperspectral and thermal sensing of stomatal conductance, transpiration, and photosynthesis for soybean and maize under drought [J]. Remote Sensing, 2020, 12(19): 3182. doi: 10.3390/rs12193182 [8] PARRAG V, FELFÖLDI J, BARANYAI L, et al. Early detection of cobweb disease infection on Agaricus bisporus sporocarps using hyperspectral imaging[J]. Acta Alimentaria, 2014, 43(Supplement 1): 107−113. [9] GASTON E, FRÍAS J M, CULLEN P J, et al. Visible-near infrared hyperspectral imaging for the identification and discrimination of brown blotch disease on mushroom (Agaricus bisporus) caps [J]. Journal of Near Infrared Spectroscopy, 2010, 18(5): 341−353. doi: 10.1255/jnirs.894 [10] LIN X H, XU J L, SUN D W. Investigation of moisture content uniformity of microwave-vacuum dried mushroom (Agaricus bisporus) by NIR hyperspectra limaging [J]. LWT-Food Science and Technology, 2019, 109: 108−117. doi: 10.1016/j.lwt.2019.03.034 [11] TAGHIZADEH M, GOWEN A, WARD P, et al. Use of hyperspectral imaging for evaluation of the shelf-life of fresh white button mushrooms (Agaricus bisporus) stored in different packaging films [J]. Innovative Food Science & Emerging Technologies, 2010, 11(3): 423−431. [12] GOWEN A A, TAGHIZADEH M, O’DONNELL C P. Identification of mushrooms subjected to freeze damage using hyperspectral imaging [J]. Journal of Food Engineering, 2009, 93(1): 7−12. doi: 10.1016/j.jfoodeng.2008.12.021 [13] 张春兰, 徐济责, 柿岛真, 等. 双孢蘑菇疣孢霉病的发病过程及病原菌的核相研究 [J]. 微生物学报, 2017, 57(3):422−433.ZHANG C L, XU J Z, SHI D Z, et al. The development of Agaricus bisporus wet bubble disease and the nuclear phase of pathogen [J]. Acta Microbiologica Sinica, 2017, 57(3): 422−433.(in Chinese) [14] 郭倩, 凌霞芬, 王志强, 等. 双孢蘑菇工厂化栽培过程中环境因子的调控 [J]. 食用菌学报, 2020, 9(3):38−41.GUO Q, LING X F, WANG Z Q, et al. The control of environmental factors in modern industrial cultivation ofAgaricus bisporus [J]. Acta Edulis Fungi, 2020, 9(3): 38−41.(in Chinese) [15] BELGIU M, DRĂGUŢ L. Random forest in remote sensing: A review of applications and future directions [J]. ISPRS Journal of Photogrammetry and Remote Sensing, 2016, 114: 24−31. doi: 10.1016/j.isprsjprs.2016.01.011 [16] DALPONTE M, ØRKA H O, GOBAKKEN T, et al. Tree species classification in boreal forests with hyperspectral data [J]. IEEE Transactions on Geoscience and Remote Sensing, 2013, 51(5): 2632−2645. doi: 10.1109/TGRS.2012.2216272 [17] HUANG G B, ZHU Q Y, SIEW C K. Extreme learning machine: A new learning scheme of feedforward neural networks [C]// 2004 IEEE International Joint Conference on Neural Networks (IEEE Cat. No. 04CH37541). Budapest, Hungary: IEEE, 2004: 985 − 990. [18] DING S F, XU X Z, NIE R. Extreme learning machine and its applications [J]. Neural Computing and Applications, 2014, 25(3/4): 549−556. [19] 李鸿强. 基于高光谱分析的蔬菜品质检测方法研究[D]. 北京: 中国农业大学, 2019.LI H Q. Study on detection method of vegetable quality based on hyperspectral analysis[D]. Beijing: China Agricultural University, 2019. (in Chinese) -
点击查看大图
计量
- 文章访问数: 659
- HTML全文浏览量: 234
- PDF下载量: 29
- 被引次数: 0
