基于小波去噪与SVR的小麦冠层含氮率高光谱测定

为改进小麦冠层含氮率的高光谱测定模型,以正交试验筛选出小波去噪的最优参数组合（小波类型取haar,分解层数为5,阈值方案选择Fixed form threshold,噪声结构定为Unscaled white noise）,并利用去噪后的小麦冠层光谱建立偏最小二乘回归（PLS）模型,对不同预处理方法进行比较分析。发现采用小波去噪结合一阶导数能最有效消除原始光谱的背景信息,此时PLS模型校正集均方根误差（RMSEC）为0.260,预测集均方根误差（RMSEP）为0.288。对经一阶导数结合小波去噪后的光谱用主成分分析（PCA）进行降维,以前6个主成份为输入变量,建立最小二乘支撑向量机回归模型（LS-SVR）,其RMSEC与RMSEP分别为0.154与0.259,具有比PLS模型更高的精度。结果表明：以小波去噪结合一阶导数去除小麦冠层反射光谱中的土壤背景信息以提高模型的精度是可行的,且LS-SVR是建模的优选方法。     

Quality studies on wheat grown in different cropping systems: a holistic perspective

Spring wheat from a conventional (CONV) and an organic (ORG1) cropping system, both with animals, and from an organic system without animals (ORG2) was evaluated with respect to baking quality for the years 1995–2002. Amino acid (AA) composition was studied in both spring and winter wheat in 1993 and 2000–2002. The data were combined in multivariate analysis for exploration of the main factors responsible for the variation in quality. The most important factor for baking quality was weather conditions. High rainfall in May favoured baking quality in both cropping systems with animals, as did high temperature in May and high rainfall in July in the ORG1 system, and low rainfall in August in the CONV system. The only significant difference between the cropping systems was falling number, which was higher in ORG1 (252 s) than in CONV (205 s), probably due to a heavier CONV crop stand causing more difficult drying conditions. AA composition differed more between years than between cropping systems for both winter and spring wheat. The content of essential amino acids was high under the weather conditions associated with poor baking quality. The contents of threonine and leucine in spring wheat were significantly higher in ORG1, 1.76 and 8.11 g/100 g crude protein than in CONV, 1.63 and 7.72, respectively. In the interaction between AA and baking quality in spring wheat, it was possible to determine a correlation between phenylalanine, histidine, lysine and good baking properties. The primary effect was associated with weather conditions, but there was also an effect of differences between the cropping systems.     

Rapid prediction of wood crystallinity in <Emphasis Type="Italic">Pinus elliotii</Emphasis> plantation wood by near-infrared spectroscopy

Crystallinity is an important property of woody materials; it responds to tree growth traits, structure, and chemical composition, and has a significant effect on Young’s modulus, dimensional stability, density, and hardness, etc. The ability of near-infrared (NIR) spectroscopy coupled with multivariate analysis to rapidly predict the crystallinity of slash pine (Pinus elliotii) plantation wood was investigated. The results showed that the NIR data could be correlated with the X-ray diffraction (XRD)-determined crystallinity of slash pine wood by use of partial least squares (PLS) regression, producing excellent coefficients of determination, r ², and root mean square error of calibration, RMSEC. The use of either reduced spectral ranges or the selection of certain wavelengths consistent with known chemical absorptions did not have any detrimental effect on the quality of PLS models allowing the use of inexpensive, small, and portable spectrometers. These studies show that NIR spectroscopy can be used to rapidly predict the crystallinity of slash pine wood.     

基于多光谱图像的幼龄白木香含锌量无损诊断

通过探讨白木香冠层光谱和形状特征与叶片含锌量的模型关系,实现幼龄白木香冠层含锌量的快速无损诊断,为实现白木香智能化培育经营提供新思路。以幼龄白木香为研究对象,通过多光谱相机获取白木香冠层图像,结合相位相关法及贝叶斯分割法精确提取白木香冠层,在应用偏最小二乘（partial least squares, PLS）算法对图像光谱和形状特征进行降维的基础上,分别构建偏最小二乘回归（partial least squares regression, PLSR）模型和偏最小二乘-广义可加模型（partial least squares- generalized additive models, PLS-GAM）以图像特征对含锌量进行估测和分析,并通过比较模型评价指标与常用套索回归（lasso regression, LassoR）和多元逐步回归（multiple stepwise regression, MSR）模型进行对比,确定适用于白木香锌含量估测的最佳模型。研究表明：（1）结合相位相关法和贝叶斯算法能够较好地分割出白木香冠层图像,效果显著优于对各波段图像进行直接分割的方法;（2）基于多光谱图像特征提取6个主成分CF₁、CF₂、CF₃、CF₄、CF₅和CF₆,PLSR建模分析结果表明CF₁和CF₂与白木香冠层含锌量具有显著的线性关系,模型调整后R²_adj为0.475;（3）PLS-GAM建模分析结果表明,CF₁、CF₂和CF₄与白木香冠层含锌量均存在显著的非线性关系,模型调整后R²_adj为0.679,显著高于基于线性关系构建的PLSR模型;（3）经过模型评价对比,PLS-GAM模型估测精度最高,RMSE为0.095,较PLSR、LassoR、MSR模型分别降低了26.4%、43.1%和34.9%,为适用于估测白木香冠层含锌量的最优模型。因此,结合相位相关法及贝叶斯分割法能够实现对白木香冠层多光谱图像的精准分割,基于光谱和形状特征构建的PLS-GAM模型对白木香冠层含锌量具有良好的估测效果,有利于推动白木香微量元素诊断的研究进程,对幼龄白木香的智能化作业有重要意义。     

A novel method for ash analysis in wheat milling fractions by using laser-induced breakdown spectroscopy

Ash content is an important quality control parameter in milling industry. Measurement of ash content is routinely performed using standard ash analysis method in which the sample is burned at 500–600 °C for 5–6 h. However, this method is not convenient for industrial applications, and thus, rapid and reliable methods are needed to be developed. The aim of this study was to develop a new method for ash analysis to be used in wheat milling fractions by using laser induced breakdown spectroscopy (LIBS). LIBS is an optic based multi-elemental, spectroscopic method which can analyze high number of samples in a considerably short time. In the study, wheat flour, whole wheat meal and semolina samples with different ash contents were analyzed using LIBS, and the spectra were evaluated with partial least squares (PLS) method. The results were correlated with the ones taken from standard ash analysis method. Calibration graph showed good linearity with the ash content between 0.48 and 2.44%, and 0.997 coefficient of determination (R²). Limit of detection for ash analysis was calculated as 0.11%. The results indicated that LIBS is a promising and reliable method with high sensitivity for routine ash analysis in milling industry.     

基于近红外光谱技术的淡水鱼品种快速鉴别

为探索淡水鱼品种的快速鉴别方法,该文应用近红外光谱分析技术,结合化学计量学方法,对7种淡水鱼品种的判别分类进行了研究。采集了青、草、鲢、鳙、鲤、鲫、鲂等7种淡水鱼,共665个鱼肉样品的近红外光谱数据,经过多元散射校正(multiplicative scatter correction,MSC)、正交信号校正(orthogonal signal correction,OSC)、数据标准化(standardization,S)等20种方法预处理,在1 000~1 799 nm范围内分别采用偏最小二乘法(partial least square,PLS)、主成分分析(principal component analysis,PCA)和BP人工神经网络技术(back propagation artificial neural network,BP-ANN)、偏最小二乘法和BP人工神经网络技术对7种淡水鱼原始光谱数据进行了鉴别分析。结果表明,近红外光谱数据,结合主成分分析和BP人工神经网络技术建立的淡水鱼品种鉴别模型最优,模型的鉴别准确率达96.4%,对未知样本的鉴别准确率达95.5%。模型具有较好的鉴别能力,采用该方法能较为准确、快速地鉴别出淡水鱼的品种。     

Automatic grading of sawlogs: A comparison between X-ray scanning,optical three-dimensional scanning and combinations of both methods

As sawmills become increasingly efficient, the importance of focusing on value recovery becomes obvious. To maximize value recovery, sawmills require the ability to sort logs according to quality. This study compares four different combinations of three-dimensional (3D) and X-ray scanning that can be used to grade logs automatically. The study was based on 135 Scots pine (Pinus sylvestris L.) logs that had been scanned with both a 3D scanner and an X-ray scanner with two X-ray sources. The percentage of boards with correct grade sawn from automatically graded logs varied from 57% when using only 3D scanning to 66% when using a combination of 3D scanning and X-ray scanning in two directions. The highest possible result, with ideal log grading, was 81%. The result also shows that the combination of a 3D scanner and one X-ray direction results in higher accuracy than a scanner based on two X-ray directions.     

基于样本挑选和不同偏最小二乘方法的近红外光谱玉米淀粉组分校正模型的研究

[目的]获得精度高、鲁棒性强的玉米近红外光谱淀粉组分检测模型。[方法]用一阶导数和Savitzky．Golay平滑对玉米1300～2298nlTl近红外光谱进行预处理,而后分别以RS（random sampling）、KS（Kennard Stone）、Duplex、SPXY（sample set partitioning based on joint x-y distance）方法选取最佳校正集样本集合,最后分别用PLS（Partial Least Squares）、iPLS（intervalPLS）和siPLS（synergy interval PLS）方法建立校正模型。[结果]采用sPXY方法选取有代表性的校正集合样本,以siPLS方法所建立的近红外光谱玉米淀粉组分校正模型最优,校正样本集合中r为0．9917,RMSECV为n1073,预测样本集合中r达到了0．9944,RMSEP为0．0814。[结论]SPXY-siPLS方法建立的近红外光谱玉米淀粉组分校正模型,不但可以减小参与建模的数据规模．而且缩短了运算时间．预测能力和精度也均得到提高。     

偏最小二乘分光光度法同时测定茶叶中良量铁、钴和镍

痕量Fe(Ⅱ）、Co（Ⅱ）和Ni（Ⅱ）与4－（2－吡啶偶氮）－间茶二酚（PAR）在pH＝8.7时发生显色反应，其吸收光谱相互严重重叠。本用偏最小二乘法（PLS）分光光度法成功地同时测定了模拟试样及茶叶是痕量铁、钴和镍。结果表明，PLS法是化学计量学中一种可适用于基体较复杂的实际试样中痕量组分同时分光光度测定的优良的多元计算方法。     

基于NIRS和Local PLS算法的堆肥关键参数实时动态分析

为对不同堆肥工艺堆肥全过程关键参数进行实时动态分析,该研究以牛粪便和玉米秸秆为原料,进行规模化槽式和膜覆盖好氧堆肥,采集堆肥全过程样本,分析了2种堆肥技术堆肥全过程中含水率、有机质含量和碳氮比等关键参数的变化,并结合Local PLS算法建立了2种堆肥技术堆肥全过程中上述参数的通用速测模型,得出以下结果:1)2种主要工艺关键参数数值及变化规律均不同,且在整个堆肥过程中有显著性变化(P0.05);2)所建立的Local PLS模型的RPD(Ratio of Prediction to Deviation)为4.47,RSD(Relative Standard Deviation)为3.37%,可达到很好的预测效果;有机质含量和碳氮比的R_P~2分别为0.74和0.77,RPD大于1.5,RSD小于10%,模型可用于定量预测;近红外预测值与实测值随堆肥时间的变化趋势具有较好的一致性,可实现规模化堆肥过程中关键参数的实时分析。