中红外光谱结合机器学习对不同产地平菇鉴别

doi:10.3964/j.issn.1000-0593(2023)02-0577-06

摘要
参考文献
相关文章 (15)

全文: PDF (3533 KB)
输出: BibTeX | EndNote (RIS)

摘要：平菇味道鲜美、营养丰富，深受消费者喜爱。平菇在我国的栽培范围较广，产地分散，每个产地的气候条件、栽培基质、栽培方式的差异，使不同产地生产的平菇在口感、营养价值方面会有不同。为规范平菇产品的市场管理，更为打造区域内特色平菇品牌，借助中红外光谱技术无污染、高效、低成本等特点，突破目前化学分析、生物学鉴别方法的限制，提出一种中红外光谱结合机器学习鉴别不同产地平菇的方法。对10个不同产地的平菇子实体进行红外光谱数据采集，每个地区各60份共600份样本。光谱数据经分析表明，在波段530～1 660 cm^-1范围内红外光谱的相关性表现出较明显的差异。同时，基于K-S法按照训练集和测试集比例为7∶3对样品划分，得训练集为420份，测试集为180份。采用多元散射校正（MSC），标准正态变量变换（SNV），平滑（SG），一阶导数（FD），二阶导数（SD）等预处理方法进行光谱优化，去除噪声，并结合支持向量机（SVM）进行初步建模对比，得出MSC预处理后光谱数据差异性最大，预测集识别效果最好为84.44%。将MSC光谱数据进行0～1区间的归一化处理，并采用主成分分析（PCA）对其进行降维，选择满足训练集中主成分个数累积贡献率≥85%，且主成分方差百分比≥1%的前7个主成分作为输入变量与支持向量机（SVM）、随机森林（RF）、极限学习机（ELM）进行建模识别比较。实验结果表明，在识别不同产地平菇模型中，SVM模型识别效果最佳，训练集和测试集识别率均为100%；RF模型训练集识别率为100%，测试集识别率略低，为98.89%；ELM模型对比其他模型识别率较差，训练集识别率为99.28%，测试集识别率为98.33%。3种模型的识别率均高于98%，说明采用红外光谱结合机器学习的方法可以简单、快速、低成本的实现对不同产地平菇的鉴别，不仅为平菇产品产地识别提供方法依据，也为其他种类食用菌产品的产地鉴别提供参考。

关键词：平菇；中红外光谱；机器学习；支持向量机；随机森林；极限学习机

Abstract：Pleurotus ostreatus is popular with consumers because of its delicious taste and rich nutrition. Pleurotus ostreatus is widely cultivated in China, and its producing areas are scattered. The differences in climate conditions, cultivation matrix and cultivation mode of each producing area make the Pleurotus ostreatus produced in different producing areas different in taste and nutritional value. In order to standardize the market management of Pleurotus ostreatus products and create regional characteristics of Pleurotus ostreatus brands, with the help of the characteristics of non-pollution, high efficiency and low cost of mid-infrared spectroscopy, this paper broke through the limitations of chemical analysis and biological identification methods at present, and put forward a method of identifying Pleurotus ostreatus from different producing areas by mid-infrared spectroscopy combined with machine learning. The infrared spectrum data of fruiting bodies of Pleurotus ostreatus from 10 different producing areas were collected, and 60 samples were collected from each area. The analysis of the spectral data showed that the correlation of the infrared spectra showed significant differences in the band 530～1 660 cm^-1. At the same time, based on the K-S method, the samples were divided according to the ratio of the training set to test set of 7∶3, 420 training sets and 180 test sets were obtained. Multiplicative scatters correction (MSC), standard normal variable transformation (SNV), Smoothing(SG), first derivative (FD), second derivative (SD) and other preprocessing methods were used to optimize the spectrum and remove the noise. In addition, it combined with a support vector machine (SVM) for preliminary modeling comparison. It was concluded that the difference in spectral data after MSC pretreatment was the largest, and the recognition performance of the prediction set was the best at 84.44%. The MSC spectral data is normalized in 0-1, and principal component analysis (PCA) was used to reduce the dimension. The first seven principal components, which satisfy the cumulative contribution rate of principal components in the training set ≥85% and the variance percentage of principal components ≥1%, were selected as input variables for modeling identification comparison with support vector machine (SVM), random forest (RF) and extreme learning machine (ELM). The experimental results showed that the SVM model had the best recognition effect in identifying Pleurotus ostreatus models from different producing areas, and the recognition rate of the training set and test set was 100%. The recognition rate of the RF model training set was 100%, and the recognition rate of the test set was slightly lower, 98.89%. Compared with other models, the recognition rate of the ELM model was poor, the recognition rate of the training set was 99.28%, and that of the test set was 98.33%. The recognition rates of the three models were all higher than 98%, indicating that the identification of Pleurotus ostreatus from different producing areas can be realized, quickly and at low cost using infrared spectroscopy combined with machine learning. This provided a method basis for the producing areas identification of Pleurotus ostreatus products and a reference for the identification of other kinds of edible fungi products’ producing areas.

Key words：Pleurotus ostreatus; Mid-infrared spectrum; Machine learning; Support vector machine; Random forest; Extreme learning machine

收稿日期: 2021-09-23 修订日期: 2022-05-05

中图分类号:

TP391.41

基金资助: 现代农业产业技术体系项目（CARS-20），中国博士后科学基金面上项目一等资助（2016M600237），吉林省教育厅“十三五”科学技术项目（JJKH20180686KJ）资助

通讯作者: 武海巍，王琦 E-mail: haiwei@jlau.edu.cn; q-wang2006@126.com

作者简介: 杨承恩，1996年生，吉林农业大学工程技术学院硕士研究生 e-mail: 928618077@qq.com

引用本文:

杨承恩，苏玲，冯伟志，周建宇，武海巍，袁月明，王琦. 中红外光谱结合机器学习对不同产地平菇鉴别[J]. 光谱学与光谱分析, 2023, 43(02): 577-582.
YANG Cheng-en, SU Ling, FENG Wei-zhi, ZHOU Jian-yu, WU Hai-wei, YUAN Yue-ming, WANG Qi. Identification of Pleurotus Ostreatus From Different Producing Areas Based on Mid-Infrared Spectroscopy and Machine Learning. SPECTROSCOPY AND SPECTRAL ANALYSIS, 2023, 43(02): 577-582.

链接本文:

https://www.gpxygpfx.com/CN/10.3964/j.issn.1000-0593(2023)02-0577-06 或 https://www.gpxygpfx.com/CN/Y2023/V43/I02/577

[1] HE Wang-xing, LI Yan-sheng, SHI Xu-ping, et al(贺望兴, 李延升, 石旭平, 等). Edible Fungi of China(中国食用菌), 2021, 40(1): 153.
[2] LIN Li-ming, ZHANG Zhen-wen, CAI Kun, et al（林立铭, 张振文, 蔡坤, 等）. Chinese Journal of Tropical Crops（热带作物学报）, 2017, 38(11): 2008.
[3] HU Su-juan, DUAN Ya-kui, KANG Yuan-chun, et al（胡素娟, 段亚魁, 康源春, 等）. Journal of Henan Agricultural Sciences（河南农业科学）, 2018, 47(3): 96.
[4] LONG Rui, SU Ling, WANG Qi（龙瑞, 苏玲, 王琦）. Edible Fungi of China（中国食用菌）, 2020, 39(5): 49.
[5] LIU Xiao-huan, LIU Cui-ling, SUN Xiao-rong, et al（刘晓欢, 刘翠玲, 孙晓荣, 等）. Food Science and Technology（食品科技）, 2021, 46(4): 244.
[6] SHI Xiao-ni, TIAN Jing, JIA Zheng, et al（石晓妮, 田静, 贾铮, 等）. Journal of Food Safety & Quality（食品安全质量检测学报）, 2020, 11(9): 2733.
[7] CHEN Lin-jie, CHENG Jun-wen, WEI Hai-long, et al（陈林杰, 程俊文, 魏海龙, 等）. China Food Additives（中国食品添加剂）, 2020, 31(9): 1.
[8] LI Chao, HUANG Xian-zhang, ZHANG Chao-yun, et al（李超, 黄显章, 张超云, 等）. Journal of Chinese Medicinal Materials（中药材）, 2019, 42(1): 51.
[9] AN Shu-jing, WANG Ting, NIU Dou, et al（安淑静, 王婷, 牛豆, 等）. Acta Chinese Medicine and Pharmacology（中医药学报）, 2021, 49(8): 49.
[10] LIU Yan, CHENG Lu, SUN Lin（刘艳, 程璐, 孙林）. Journal of Henan Normal University·Natural Science Edition（河南师范大学学报·自然科学版）, 2019, 47(2): 22.
[11] LU Lu-lu, FAN Yi-ling, DENG Ke, et al(卢路路, 樊怡灵, 邓珂, 等). Journal of Nuclear Agricultural Sciences(核农学报), 2021, 35(7): 1605.
[12] Hu Minwei, Zou Ling, Lu Jiong, et al. Bioengineered, 2021, 12(1): 6821.
[13] LI Heng-kai, WANG Li-juan, XIAO Song-song(李恒凯, 王利娟, 肖松松). Transactions of the Chinese Society of Agricultural Engineering(农业工程学报), 2021, 37(7): 247.
[14] XIONG Zhi-wen（熊治文）. Ship Science and Technology（舰船科学技术），2021, 43(2): 215.
[15] KANG Li, YUAN Jian-qing, GAO Rui, et al（康丽, 袁建清, 高睿, 等）. Spectroscopy and Spectral Analysis（光谱学与光谱分析）, 2021, 41(3): 900.