外文摘要：This work explored the possibility of using hyperspectral microscope imaging (HMI) technique coupled with advanced chemometric methods to evaluate the cell wall microstructure and physiochemical properties of 'Korla' fragrant pear disease caused by Alternaria alternata. The physicochemical characteristics such as SSC, firmness and L* value of pears undergo successive decreases and the microstructure of the cell wall breaks down during the process of pathogen infection. Principal component analysis was applied on the HMI of pear tissues at different infected stages, which could clearly visualize the distribution of pigment, carbohydrate compounds and structural changes in parenchyma cells. Further, partial least squares discriminant analysis (PLS-DA), support vector machine (SVM), and convolutional neural network (CNN) model coupled with selected spectral variables, and HMI features were used to identify the diseased 'Korla' fragrant pears. The CNN model based on the fused data showed the best discrimination between healthy and diseased pears (96.72%) and provided a satisfactory discrimination accuracy of 94.74% in successfully identifying the diseased diameter of 1.56 mm after 1 d of storage. This study indicated the HMI combined with CNN has great potential in detecting the early stages of pear infection and provides a possible method for monitoring fruit quality and safety.
外文关键词：Convolutional Neural Network;early disease detection;'Korla' fragrant pear;Alternaria alternata;Hyperspectral microscope imaging
作者：You, Sicong;Li, Yiting;Song, Jin;Yu, Xiaobo;Tu, Kang;Lan, Weijie;Pan, Leiqing
作者单位：Nanjing Agr Univ
期刊名称：POSTHARVEST BIOLOGY AND TECHNOLOGY
期刊影响因子：0.0
出版年份：2024
出版刊次：212
原文传递申请：江苏省科技资源（工程技术文献）统筹服务平台