大豆水分含量的高光谱无损检测及可视化研究

doi:10.3964/j.issn.1000-0593(2022)10-3052-06

摘要
参考文献
相关文章 (15)

全文: PDF (2740 KB)
输出: BibTeX | EndNote (RIS)

摘要：采用近红外高光谱成像技术对大豆水分含量进行快速无损检测，实现大豆水分含量可视化。采集了96个不同品种大豆样本在900～2 500 nm的高光谱图像，采用直接干燥法测量每个大豆样品的水分含量。利用系统自带的HSI Analyzer软件提取图像感兴趣区域（ROI）的平均光谱信息，代表样品的光谱信息。利用SPXY算法划分样品校正集和预测集，并保留938～2 215 nm波段范围内的光谱数据。采用移动平滑（moving average）、S-G平滑、基线校正（baseline）、归一化（normalize）、标准正态变量变换（standard normal variate，SNV）、多元散射校正（multiple scattering correction，MSC）、去趋势（detrending）共7种光谱预处理方法，发现Normalize方法处理后的PLSR模型效果较好。为了去除光谱冗余信息，简化预测模型，采用连续投影算法（SPA）、竞争性自适应加权算法（CARS）、无信息消除变量法（UVE）提取特征波长，其中SPA，CARS和UVE三种算法优选出14，16和29个波长，分别占总波长的6.5%，7.4%和13.4%。分别对938～2 215 nm波段光谱和特征波长建立预测模型，并将效果较优的模型与Normalize方法结合。建立的14种预测模型效果相比较，发现SPA算法筛选的特征波长建模预测效果较好，并优选出Normalize-SPA-PCR模型，模型的R^２_C和Ｒ^２_P值较高，分别为0.974 6和0.977 8，RMSEP和RMSECV值较低，分别为0.238和0.313，模型的稳定性和预测性较好，可以对大豆水分含量进行准确预测。将Normalize-SPA-PCR模型作为大豆水分含量可视化预测模型，计算高光谱图像每个像素点的水分含量，得到灰度图像，对灰度图像进行伪彩色变换，得到大豆水分含量可视化彩色图像。对预测集的24个大豆品种进行可视化处理，发现不同水分含量大豆的可视化图像颜色不同，水分含量变化对应的颜色变化较为明显。结果表明，高光谱成像技术结合化学计量学可以准确快速无损预测大豆水分含量，实现大豆水分含量可视化，为大豆收获、贮藏加工过程中水分含量检测提供了技术支持。

关键词：高光谱成像技术；水分含量；大豆；无损检测；可视化

Abstract：NIR Hyperspectral imaging technology was used to detect soybean moisture content rapidly and non-destructively and realized the visualization of soybean moisture content. A total of 96 soybean samples of hyperspectral images in the region of 900～2 500 nm were acquired, and the moisture content of each soybean sample was measured by the direct drying method. The average spectral information of the region of interest（ROI）of the image was extracted by HSI Analyzer software, representing the sample’s spectral information. The SPXY algorithm was used to divide the sample calibration set and prediction set，and the spectral data in the band range of 938 to 2 215 nm were retained. The spectral’s pretreatment was analyzed, such as Moving Average, Smoothing S-G, Baseline, Normalize, Standard Normal Variate（SNV）, Multiple Scattering Correction（MSC）and Detrending, and the PLSR model established after Normalize pretreatment had the best effect. The characteristic wavelengths were selected by successive projections algorithm（SPA）, competitive adaptive reweighted sampling（CARS）and uninformative variable elimination（UVE）. 14,16 and 29 characteristic wavelengths were selected by SPA, CARS and UVE, accounting for 6.5%，7.4% and 13.4% of the total wavelengths. The prediction models were established for the spectra and characteristic wavelengths of 938～2 215 nm, and the model with better effect was combined with the Normalize method. Compared with the 14 prediction models established, it was found that the modeling and prediction effect of characteristic wavelengths selected by the SPA algorithm was good, and the Normalize-SPA-PCR model was optimized. The values of R²_C and R²_P in the model were higher, which were 0.974 6 and 0.977 8, respectively, while the values of RMSEP and RMSECV in the model were lower, which were 0.238 and 0.313, respectively. The stability and predictability of the model were good, which could be used to predict the soybean moisture content accurately. The Normalize-SPA-PCR model was used as a visual prediction model for soybean moisture content, and the moisture content of each pixel in the hyperspectral image was calculated to obtain a gray image. The gray image was transformed by pseudo-color transformation to obtain a visual color image of soybean moisture content. The 24 soybean varieties in the prediction set were visualized. The color of the visualized image was different with different moisture content, and the color of the visualized image was more evident with different moisture content. The results showed that hyperspectral imaging combined with stoichiometry could accurately, rapidly, and non-destructive predict soybean moisture content. They realized the visualization of soybean moisture content, which provided technical support for soybean moisture content detection in the process of soybean harvest, storage and processing.

Key words：Hyperspectral imaging; Moisture content; Soybean; Non-destructive detection; Visualization

收稿日期: 2021-08-11 修订日期: 2021-11-11

中图分类号:

O433.5

基金资助: 国家自然科学基金项目（32171910），国家重点研发计划项目（2017YFD0700305），现代农业产业技术体系建设专项资金项目（CARS-04-PS26）资助

作者简介: 金诚谦，1973年生，山东理工大学农业工程与食品科学学院教授，农业农村部南京农业机械化研究所研究员
e-mail: 412114402@qq.com

引用本文:

金诚谦，郭榛，张静，马成业，唐小涵，赵男，印祥. 大豆水分含量的高光谱无损检测及可视化研究[J]. 光谱学与光谱分析, 2022, 42(10): 3052-3057.
JIN Cheng-qian, GUO Zhen, ZHANG Jing, MA Cheng-ye, TANG Xiao-han, ZHAO Nan, YIN Xiang. Non-Destructive Detection and Visualization of Soybean Moisture Content Using Hyperspectral Technique. SPECTROSCOPY AND SPECTRAL ANALYSIS, 2022, 42(10): 3052-3057.

链接本文:

https://www.gpxygpfx.com/CN/10.3964/j.issn.1000-0593(2022)10-3052-06 或 https://www.gpxygpfx.com/CN/Y2022/V42/I10/3052

[1] ZHANG Feng，LIU Mei，YANG Cui-cui，et al（张凤，刘美，杨翠翠，等）. Shandong Agricultural Sciences（山东农业科学），2014，46（8）：37.
[2] Yuan Lin，Yan Peng，Han Wenyan，et al. Computers and Electronics in Agriculture，2019，167: 105039.
[3] Deng Shuiguang，Xu Yifei，Li Xiaoli，et al. Computers and Electronics in Agriculture，2015，118: 38.
[4] Wei Yuzhen，Wu Feiyue，Xu Jie，et al. Journal of Food Engineering，2019，248: 89.
[5] Mohammed Kamruzzaman，Gamal ElMasry，Da-Wen Sun，et al. Analytica Chimica Acta，2011，714: 57.
[6] Yao Xinli，Cai Fuhong，Zhu Peiyi，et al. Meat Science，2019，152: 73.
[7] Nicola Caporaso，Martin B Whitworth，Ian D Fisk. Food Chemistry，2018，240.
[8] Nicola Caporaso，Martin B Whitworth，Stephen Grebby，et al. Journal of Food Engineering，2018，227.
[9] Xu Yunfei，Zhang Haijun，Zhang Chi，et al. Infrared Physics and Technology，2019，102.
[10] Jennyfer D Rabanera，Jose D Guzman，Kevin F Yaptenco. Journal of Food Measurement and Characterization，2021（prepublish）.
[11] Wang Zheli，Fan Shuxiang，Wu Jingzhu，et al. Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy，2021，254: 119666.
[12] ZHU Jie，HUANG Hua，ZHU Shi-ping，et al（朱洁，黄华，祝诗平，等）. Journal of the Chinese Cereals and Oils Association（中国粮油学报），2019，34（8）：133.
[13] XU Ying-chao，WANG Xiang-you，YIN Xiang，et al（许英超，王相友，印祥，等）. Transactions of the Chinese Society for Agricultural Machinery（农业工程学报），2018，49（2）：339.
[14] Wang Qiaoyun，Wu Guangfei，Pian Feifei，et al. Spectrochimica Acta Part A：Molecular and Biomolecular Spectroscopy，2021，260: 119906.
[15] Ma Chengye，Ren Zhishang，Zhang Zhehao，et al. Vibrational Spectroscopy，2021，114: 113230.
[16] WANG Song-lei，WU Long-guo，WANG Cai-xia，et al（王松磊，吴龙国，王彩霞，等）. Journal of Optoelectronics·Laser（光电子·激光），2019，30（9）：941.
[17] REN Zhi-shang，PENG Hui-hui，HE Zhuang-zhuang（任志尚，彭慧慧，贺壮壮，等）. Transactions of the Chinese Society for Agricultural Machinery（农业机械学报），2020，51（S2）：466.
[18] Mário César Ugulino Araújo，Teresa Cristina Bezerra Saldanha，Roberto Kawakami Harrop Galvão，et al. Chemometrics and Intelligent Laboratory Systems，2001，57（2）: 65.
[19] Li Hongdong，Liang Yizeng，Xu Qingsong，et al. Analytica Chimica Acta，2009，648（1）: 77.