双孢蘑菇疣孢霉病早期高光谱图像鉴别  

Identification of Agaricus Bisporus Disease in Early Stage by Hyperspectral Image

作　　者：陈子涵 廖小玲 CHEN Zi-han;LIAO Xiao-ling(College of Mechanical and Electronic Engineering,Fujian Agriculture andForestry University,Fuzhou,Fujian 350002,China)

机构地区：[1]福建农林大学机电工程学院,福建福州350002

出　　处：《福建农业科技》2021年第4期16-22,共7页Fujian Agricultural Science and Technology

基　　金：国家自然科学基金项目(61705037)。

摘　　要：双孢蘑菇疣孢霉病是由有害疣孢霉菌Mycogone perniciosa引起的、破坏性极强的真菌类病害,且该病害检测困难耗时,往往导致菇房绝收,菇农收益损失严重。早发现、早处理能够有效解决病害带来的经济损失和农药残留超标等质检问题。因此,本研究将能够快速无损检测的高光谱成像技术应用到双孢蘑菇病害早期鉴别。以双孢蘑菇菌Agaricus bisporus子实体为试材,对健康染病双孢蘑菇生长早期子实体样本采集菌盖的全波段(401~1046 nm)可见/近红外高光谱图像信息,利用多元散射校正(MSC)进行预处理,采用决策树(DT)提取特征波段,对比随机森林(RF)和极限学习机(ELM)两种模型对健康和染病双孢蘑菇鉴别准确度。利用DT选取401.00、951.59、978.09、1006.59和1044.90 nm为鉴别病害的特征波段。对比RF和ELM所建模型效果,得到MSC DT ELM模型检测效果最优,测试集和预测集总体样本鉴别准确度分别为92.39%和91.32%。结果表明,该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,得到基于高光谱成像技术的便捷准确鉴别双孢蘑菇病害早期的模型,同时,为进一步开发双孢蘑菇病害早期的多光谱设备提供了理论依据和方法。The wet bubble disease of Agaricus bisporus is a highly destructive fungal disease caused by Mycogone perniciosa,which is difficult to detect and time-consuming,often leading to the failure of mushroom houses and the serious loss of income of the mushroom farmers.Early detection and treatment for this disease can effectively solve the economic loss caused by the diseases and the quality control problems such as excessive pesticide residues.Therefore,in this study,the hyperspectral imaging technology,which could be used for the rapid nondestructive detection,was applied to the early identification of Agaricus bisporus diseases.By using the fruiting body of Agaricus bisporus as the test material,the full-band(401-1046 nm)visible/near-infrared hyperspectral image information of the pileus was collected from the fruiting body samples at the early growth stage of healthy infected Agaricus bisporus.The multiplicative scatter correction(MSC)was used for the preprocessing.The decision tree(DT)was used to extract the characteristic bands.The accuracy of the two models,Random Forest(RF)and Extreme Learning Machine(ELM),for the identification of healthy and infected Agaricus bisporus was compared.By DT,401.00,951.59,978.09,1006.59 and 1044.90 nm were selected as the characteristic bands for the disease identification.By comparing the effect of RF model and ELM model,the MSC-DT-ELM model had the best detection effect,and the identification accuracy of the test set and the prediction set were 92.39%and 91.32%,respectively.The results showed that the MSC-DT-ELM model could effectively improve the early identification accuracy of the wet bubble disease of Agaricus bisporus based on full band,and obtain a convenient and accurate model for the early identification of Agaricus bisporus disease based on the hyperspectral imaging technology.At the same time,the theoretical basis and method were provided for the further development of multispectral equipment for the rapid and nondestructive identification of Agaricus bisporus disease

关 键 词：双孢蘑菇 高光谱成像技术 病害早期检测 鉴别 

分 类 号：S646[农业科学—蔬菜学]