高光谱成像
主成分分析
降维
化学计量学
模式识别(心理学)
支持向量机
狗尾草
狐尾
计算机科学
人工智能
判别式
特征选择
线性判别分析
数学
作者
Fuxiang Wang,Chunguang Wang,Shiyong Song
标识
DOI:10.1016/j.infrared.2022.104179
摘要
• NIR-HSI coupled with chemometrics was used for origin identification of millet. • The accuracy of the discriminant model for origin identification reached 95.00%. • Data dimension reduction can effectively improve model accuracy. • It provides reference for the origin tracing of agricultural products. Foxtail millet is an important edible crop in China and its demand continues to increase with the growth of the population. There are differences in the quality of millet from different origins, leading to differences in price. Therefore, an effective and rapid technique is needed to monitor the authenticity of millet origin. This study evaluates the feasibility of hyperspectral imaging (HSI) in identifying the origin of millet. A total of 120 millet samples from four regions in Inner Mongolia of China were analyzed and the hyperspectral data in the wavelength of 900-1700 nm were collected. The support vector machine (SVM) model built on full-band spectral data achieved acceptable results with a discriminative accuracy of 87.50% for the prediction set. Compared with the full-band spectra, the performance of the SVM model built on the basis of data dimensionality reduction was improved. Among them, the principal component analysis (PCA)-SVM model achieved the best discriminative performance with an accuracy of 95.00% for the prediction set. This study demonstrates the feasibility of NIR-HSI for origin traceability of millet, which provides a high-throughput, rapid and non-destructive quality control means to ensure the commodity flow of millet.
科研通智能强力驱动
Strongly Powered by AbleSci AI