高光谱成像
油茶
线性判别分析
人工智能
模式识别(心理学)
霉病
主成分分析
采后
VNIR公司
计算机科学
数学
偏最小二乘回归
统计
园艺
生物
作者
Hongzhe Jiang,Xuesong Jiang,Ruihang Yu,Qing Chen,Xingpeng Li,Xu Li,Hu Zhou,Minghong Shi
标识
DOI:10.1016/j.infrared.2022.104169
摘要
Camellia oleifera fruit is a kind of popular woody oil crops in China, which is susceptible to natural mildew damage during storage. This problem significantly depreciates its value and poses a potential threat to human health. A hyperspectral imaging (HSI) system covered visible and near-infrared (Vis-NIR, 400–1000 nm) range was employed to identify Camellia oleifera fruit with three different natural mildew degrees (slight, moderate, and severe). Spectra were primarily extracted from representative regions of interest (ROIs), and principal component analysis (PCA) of extracted spectra revealed that PC1 and PC2 were effective for the identification. Hyperspectral images were then processed using PC transformation for exploratory visual detection by displaying PC1 and PC2 score images. Three modeling methods including partial least squares-discriminant analysis (PLS-DA), k-nearest neighbor (KNN), and classification and regression tree (CART) were subsequently evaluated in terms of their capacity to establish the natural mildew degree estimation model developed by full spectra. Raw spectra without any preprocessing constructed the optimal PLS-DA model with the highest 90.8% correct classification rate (CCR) of success in external prediction set. After that, key wavelengths selected by competitive adaptive reweighted sampling (CARS) algorithm built an optimal simplified model, achieving the best performance with the highest CCR of 83.3%, AUC value of 0.89, and Kappa coefficient of 0.75 in prediction set. Therefore, CARS-PLS-DA model was finally employed to visually classify the three different natural mildew degrees. As a result, the general natural mildew degrees of Camellia oleifera samples were readily discernible in pixel-wise manner by generating classification maps. The overall results illustrated that HSI offered an alternative way in detecting and visualizing natural mildew degrees of Camellia oleifera fruit.
科研通智能强力驱动
Strongly Powered by AbleSci AI