应用高光谱图像技术对林下作物质量等级鉴别方法——以黄芪为例  

Identification Method of Understory Crops Quality Grade Using Hyperspectral Image Technology:A Case of Astragalus membranaceus

作　　者：张佳薇[1] 支佶豪 管雪梅[1] 张颂 苏田 林舒杨 余佩龙 李明宝[1] Zhang Jiawei;Zhi Jihao;Guan Xuemei;Zhang Song;Su Tian;Lin Shuyang;Yu Peilong;Li Mingbao(Northeast Forestry University,Harbin 150040,P.R.China)

机构地区：[1]东北林业大学,哈尔滨150040

出　　处：《东北林业大学学报》2024年第6期79-84,共6页Journal of Northeast Forestry University

基　　金：国家自然科学基金面上项目(32171691)。

摘　　要：黄芪作为一种林下多年生草本,具有极高的经济和药用价值。黄芪粉是黄芪的重要消费形式,不同质量等级的黄芪粉由于内部成分差异,在近红外光谱下具有不同的特性,而肉眼却难以区分。针对不同质量等级间黄芪粉难以鉴别的问题,利用高光谱成像技术对312组黄芪粉样本进行数据采集,再对光谱信息采用标准正态变化(SNV)、多元散射校正(MSC)和卷积平滑(SG)3种预处理,再利用竞争性自适应重加权采样(CARS)、变量组合集群分析(VCPA)和区间变量迭代空间收缩法(IVISSA)对全波段光谱进行特征提取,以优选的特征波长作为输入,建立K-近邻判别(KNN)和支持向量机(SVM)分类模型。结果表明:经过竞争性自适应重加权采样的支持向量机模型分类效果最好,训练集和测试集准确率分别达到100.00%和98.94%,能够实现黄芪粉的准确分类,为林下作物的等级鉴别提供了新的研究思路。Astragalus membranaceus is an understory perennial herbaceous plant with extremely high economic and medicinal value.A.membranaceus powder is an important consumption form of A.membranaceus,and different quality grades of A.membranaceus powder have different characteristics in near-infrared spectroscopy due to differences in internal components,which are difficult to distinguish by the naked eye.To address the problem of difficulty in distinguishing between different quality grades of A.membranaceus powder,hyperspectral imaging technology was used to collect data on 312 groups of A.membranaceus powder samples.The spectral information was preprocessed using Standard Normal Variate(SNV),Multiplicative Scatter Correction(MSC),and Savitzky-Golay(SG)smoothing.Competitive Adaptive Reweighted Sampling(CARS),Variable Combination Population Analysis(VCPA),and Interval Variable Iterative Space Shrinkage Algorithm(IVISSA)were used to extract features from the full spectrum,and the selected feature wavelengths were used as input to establish K-Nearest Neighbor(KNN)and Support Vector Machine(SVM)classification models.The results showed that the SVM model with CARS had the best classification performance,with training and testing set accuracies reaching 100%and 98.94%,respectively.It can accurately classify A.membranaceus powder,providing a new research approach for the grading identification of understory crops.

关 键 词：高光谱图像 黄芪鉴别 特征波长提取 机器学习 

分 类 号：S789.9[农业科学—林学] R282.5[医药卫生—中药学]