该文采用傅里叶变换中红外光谱和矿质元素信息结合多种特征变量选取方法,建立偏最小二乘判别分析(partial least squares discriminant analysis,PLS-DA)模型实现冰糖橙的产地鉴别,并通过低级和中级两种数据融合策略进一步提高鉴别能力...该文采用傅里叶变换中红外光谱和矿质元素信息结合多种特征变量选取方法,建立偏最小二乘判别分析(partial least squares discriminant analysis,PLS-DA)模型实现冰糖橙的产地鉴别,并通过低级和中级两种数据融合策略进一步提高鉴别能力。结果显示,冰糖橙的中红外吸收峰主要与糖类和蛋白质有关,不同产地冰糖橙的中红外光谱在糖类相关的吸收峰上存在差异。不同产地冰糖橙的矿质元素含量有所不同,其中Fe、K、Mg、P、Se 5种矿质元素在4个产地的冰糖橙之间呈现显著性差异。采用变量投影重要性指标、反向区间偏最小二乘法和组合区间偏最小二乘法等特征变量选取方法可以提高产地鉴别模型的正确率。与使用单一技术获得数据建立的PLS-DA模型相比,使用中级数据融合策略建立的模型在训练集和测试集上分别取得98%和100%的正确率。研究表明中红外光谱和矿质元素信息都能反映不同产地冰糖橙间的差异,基于中级数据融合策略融合两种信息建立的PLS-DA模型能够实现冰糖橙产地的准确鉴别,可以作为冰糖橙溯源研究的一种可靠方法。展开更多
[Objective] The aim was to indentify diseased leaves of broad bean by vibra- tional spectroscopy. [Method] In this paper, broad bean rust, fusarium rhizome rot, broad bean zonate spot, yellow leaf curl virus and norma...[Objective] The aim was to indentify diseased leaves of broad bean by vibra- tional spectroscopy. [Method] In this paper, broad bean rust, fusarium rhizome rot, broad bean zonate spot, yellow leaf curl virus and normal leaves were studied using Fourier transform infrared spectroscopy combined with chemometrics. [Result] The spectra of the samples were similar, only with minor differences in absorption inten- sity of several peaks. Second derivative analyses show that the significant difference of all samples was in the range of 1 200-700 cm2. The data in the range of 1 200- 700 cm＇ were selected to evaluate correlation coefficients, hierarchical cluster analy- sis （HCA） and principal component analysis （PCA）. Results showed that the correla- tion coefficients are larger than 0.928 not only between the healthy leaves, but also between the same diseased leaves. The values between healthy and diseased leaves, and among diseased leaves, are all declined. HCA and PCA yielded about 73.3% and 82.2% accuracy, respectively. [Conclusion] This study demonstrated that FTIR techniques might be used to detect crop diseases.展开更多
Fourier transform infrared （FTIR） spectroscopy was used to study diseased leaves in broad bean. Results showed that the infrared spectra of different broad bean diseased leaves were similar, which were mainly made u...Fourier transform infrared （FTIR） spectroscopy was used to study diseased leaves in broad bean. Results showed that the infrared spectra of different broad bean diseased leaves were similar, which were mainly made up of the vibrational absorption bands of protein,lipid and polysaccharide.There were minor differences in-cluding the spectral peak position, peak shape and the absorption intensity in the range of 1 800-1 300 cm-1. There were obvious differences among their second derivative spectra in the range of 1 800-1 300 cm-1. After the procedure of the Fourier self-deconvolution and curve fitting of health bean leaves and broad bean diseased leaves in the range of 1 700-1 500 cm-1, three sub-peaks were obtained at 1 550 cm-1 （protein amide Ⅱ band）, 1 605 cm-1 （lignin） and 1 650 cm-1 （protein amide I band）.The ratios of relative areas of the bands of amide Ⅱ, lignin, and amide I were 38.86%, 28.68% and 32.47% in the spectra of healthy leaves, respec-tively. It was distinguished from the diseased leaves （chocolate spot leaf： 15.42%, 42.98% and 41.61%, ring spot leaf：32.39%, 35.63% and 31.98%, rust leaf： 13.97%, 46.40% and 39.65%, yel owing leaf curl disease leaf： 24.01%,36.55% and 39.44%）. For sub-peak area ratios （A1 563/A1 605, A1 650/A1 605 and A1 563/A1 654）, those of four kinds of diseased leaves were smal er than that of healthy leaves, and there were also differences among four kinds of diseased leaves. The results proved that FTIR combining with curve fitting might be a potential y useful tool for detecting different kinds of broad bean diseases.展开更多
In order to distinguish 8 kinds of rhizome crops, the 40 samples were studied by Fourier transform infrared spectroscopy （FTIR） combined with wavelet transform （WT）, principal component analysis （PCA） and hieram...In order to distinguish 8 kinds of rhizome crops, the 40 samples were studied by Fourier transform infrared spectroscopy （FTIR） combined with wavelet transform （WT）, principal component analysis （PCA） and hieramhical cluster analysis （HCA）. The results showed that the infrared spectra were similar on the whole, but there were differences in peak position, peak shape and peak absorption intensity in the range of 1 800-700 cm-1. The infrared spectra in the range of 1 800-700 cm-1 were selected to perform continuous wavelet transform （CWT） and discrete wavelet transform （DWT）. The 15th-Ievel decomposition coefficients of CWT and the 5=-level detail coefficients of DWT were classified by PCA and HCA. The cumulative contri- bution rates of the first three principal components of CWT and DWT were 93.12% and 89.78%, respectively. The accurate recognition rates of PCA and HCA were all 100%. It is proved that FTIR combined with WT can be used to distinguish different kinds of rhizome crops.展开更多
基金Supported by National Natural Science Foundation of China(30960179)Natural Science Foundation of Yunnan Province(2007A048M)~~
文摘[Objective] The aim was to indentify diseased leaves of broad bean by vibra- tional spectroscopy. [Method] In this paper, broad bean rust, fusarium rhizome rot, broad bean zonate spot, yellow leaf curl virus and normal leaves were studied using Fourier transform infrared spectroscopy combined with chemometrics. [Result] The spectra of the samples were similar, only with minor differences in absorption inten- sity of several peaks. Second derivative analyses show that the significant difference of all samples was in the range of 1 200-700 cm2. The data in the range of 1 200- 700 cm＇ were selected to evaluate correlation coefficients, hierarchical cluster analy- sis （HCA） and principal component analysis （PCA）. Results showed that the correla- tion coefficients are larger than 0.928 not only between the healthy leaves, but also between the same diseased leaves. The values between healthy and diseased leaves, and among diseased leaves, are all declined. HCA and PCA yielded about 73.3% and 82.2% accuracy, respectively. [Conclusion] This study demonstrated that FTIR techniques might be used to detect crop diseases.
基金Supported by National Natural Science Foundation of China(30960179)Program for Innovative Research Team in Science and Technology in University of Yunnan Province~~
文摘Fourier transform infrared （FTIR） spectroscopy was used to study diseased leaves in broad bean. Results showed that the infrared spectra of different broad bean diseased leaves were similar, which were mainly made up of the vibrational absorption bands of protein,lipid and polysaccharide.There were minor differences in-cluding the spectral peak position, peak shape and the absorption intensity in the range of 1 800-1 300 cm-1. There were obvious differences among their second derivative spectra in the range of 1 800-1 300 cm-1. After the procedure of the Fourier self-deconvolution and curve fitting of health bean leaves and broad bean diseased leaves in the range of 1 700-1 500 cm-1, three sub-peaks were obtained at 1 550 cm-1 （protein amide Ⅱ band）, 1 605 cm-1 （lignin） and 1 650 cm-1 （protein amide I band）.The ratios of relative areas of the bands of amide Ⅱ, lignin, and amide I were 38.86%, 28.68% and 32.47% in the spectra of healthy leaves, respec-tively. It was distinguished from the diseased leaves （chocolate spot leaf： 15.42%, 42.98% and 41.61%, ring spot leaf：32.39%, 35.63% and 31.98%, rust leaf： 13.97%, 46.40% and 39.65%, yel owing leaf curl disease leaf： 24.01%,36.55% and 39.44%）. For sub-peak area ratios （A1 563/A1 605, A1 650/A1 605 and A1 563/A1 654）, those of four kinds of diseased leaves were smal er than that of healthy leaves, and there were also differences among four kinds of diseased leaves. The results proved that FTIR combining with curve fitting might be a potential y useful tool for detecting different kinds of broad bean diseases.
基金Supported by National Natural Science Foundation of China(30960179)Program for Innovative Research Team in Science and Technology in University of Yunnan Province~~
文摘In order to distinguish 8 kinds of rhizome crops, the 40 samples were studied by Fourier transform infrared spectroscopy （FTIR） combined with wavelet transform （WT）, principal component analysis （PCA） and hieramhical cluster analysis （HCA）. The results showed that the infrared spectra were similar on the whole, but there were differences in peak position, peak shape and peak absorption intensity in the range of 1 800-700 cm-1. The infrared spectra in the range of 1 800-700 cm-1 were selected to perform continuous wavelet transform （CWT） and discrete wavelet transform （DWT）. The 15th-Ievel decomposition coefficients of CWT and the 5=-level detail coefficients of DWT were classified by PCA and HCA. The cumulative contri- bution rates of the first three principal components of CWT and DWT were 93.12% and 89.78%, respectively. The accurate recognition rates of PCA and HCA were all 100%. It is proved that FTIR combined with WT can be used to distinguish different kinds of rhizome crops.
