猕猴桃软腐病无损高光谱诊断技术研究  

Nondestructive Hyperspectral Imaging for the Detection of Kiwifruit Soft Rot Disease

作　　者：翟依晨 张林[1] 祝佳佳 龙艳 杨鑫睿 赵志博 Zhai Yichen;Zhang Lin;Zhu Jiajia;Long Yan;Yang Xinrui;Zhao Zhibo(College of Agriculture,Guizhou University,Guiyang 550025,Guizhou,China)

机构地区：[1]贵州大学农学院,贵州贵阳550025

出　　处：《山地农业生物学报》2025年第2期78-86,共9页Journal of Mountain Agriculture and Biology

基　　金：大学生创新创业训练计划项目(贵大(省)创字[2023]108)。

摘　　要：基于高光谱成像技术实现对猕猴桃软腐病的室内无损快速检测,以期为病果诊断、果实分选提供技术方案。在可见光-近红外波长范围内(400~1000 nm)采集病果与健康果的高光谱图像,并提取相应的感兴趣区域(ROI),获得样本128个波段的高光谱数据,基于最小冗余最大相关(mRMR)特征排序算法,从中选择8个最优特征波段。使用标准正态变量变换(SNV)方法对高光谱数据进行预处理,将样本随机分配为测试和训练数据集,分别基于全波段和特征波段建立分类模型。同时,从自然发病果实分离软腐病菌,并鉴定其种类。结果表明:猕猴桃软腐病果实与健康果实的高光谱曲线存在明显区别。针对猕猴桃软腐病识别,可优化神经网络模型识别效果最好,从发病果实分离到葡萄座腔菌(Botryosphaeria dothidea)和甜樱间座壳属(Diaporthe eres)均为已知的猕猴桃软腐病菌,表明自然发病组分类结果可靠。本研究利用高光谱成像技术能准确区分贵长猕猴桃软腐病果实与健康果实,可实现猕猴桃软腐病早期无症状时期的无损快速检测,为采后果实分选提供了可靠技术。This study aims to achieve rapid,nondestructive indoor detection of kiwifruit soft rot disease using hyperspectral imaging technology,providing a technical basis for disease diagnosis and fruit sorting.Hyperspectral images of diseased and healthy kiwifruits were acquired within the visible-near infrared wavelength range(400~1000 nm),and regions of interest(ROIs)were extracted to obtain spectral data from 128 bands.Using the minimum Redundancy and Maximum Relevance(mRMR)feature selection algorithm,eight optimal spectral bands were identified.The hyperspectral data were normalized using the Standard Normal Variate(SNV)method and randomly divided into test and validation datasets to build classification models based on both full-band and feature-band data.Soft rot pathogens were isolated from naturally infected kiwifruits,and their species were identified.The results demonstrated clear differences in the hyperspectral reflectance curves of diseased and healthy fruits.Among the classification models,the optimized neural network model achieved the best recognition performance for identifying soft rot disease.The isolated pathogens were identified as Botryosphaeria dothidea and Diaporthe eres,known causative agents of kiwifruit soft rot,confirming the reliability of the classification results for naturally infected fruits.This research highlights the potential of hyperspectral imaging technology for accurately distinguishing soft rot-diseased Guichang kiwifruits from healthy ones during the early symptom-free stage.The findings provide a robust foundation for postharvest fruit sorting based on rapid,nondestructive testing methods.

关 键 词：猕猴桃软腐病 高光谱成像技术 无损快速检测 机器学习 

分 类 号：S4[农业科学—植物保护]