共查询到19条相似文献，搜索用时 156 毫秒
1.
《土壤学报》2014,(5)
对应用田间行走式设备获取的土壤红外光谱数据,通过特征变换和特征选择相结合,以提高所建立土壤碳校正模型的预测精度。首先应用独立成分分析(ICA)、主成分分析(PCA)和小波分析(WA)对土壤红外光谱数据进行特征变换,然后分别应用无信息变量消除法(UVE)、连续投影算法(SPA)、无信息变量消除结合连续投影算法(UVE-SPA)、基于遗传算法和偏最小二乘法的变量选择法(GA-PLS)来进行特征选择,基于所选择的特征建立了土壤碳校正模型。结果表明,通过ICA进行特征变换,然后进行特征选择,可以建立比直接对光谱数据进行波长选择精度更好的预测模型;而WA或PCA与特征选择方法结合,只能获得与对光谱数据直接进行波长选择相近的效果。因此,针对田间条件下通过行走式设备获得的光谱数据由于受复杂的环境条件下干扰多的情况,可以将ICA与特征选择方法结合起来对光谱数据进行特征变换和选择,以建立更可靠的土壤碳含量预测模型。 相似文献
2.
为了实现小麦蛋白质的无损检测,简化便携式小麦蛋白质检测设备的预测模型,提高模型预测精度,该文针对小麦采集波长范围为950～1690nm的近红外漫透反射光谱,结合蒙特卡罗采样(MCS,monte carlosampling)技术与特征投影图(LPG,latent projective graph)方法对波长变量进行选择。根据模型集群分析(MPA,model population analysis)思想,采用MCS技术建立样本子空间,利用主成分分析(PCA,principal componentanalysis)得到LPG,假定LPG中共线性光谱变量对建模作用相同,选出少数波长变量建立子预测模型,选出预测均方根误差(RMSEP,root-mean-square error of prediction)较小的子模型,统计分析其变量的出现频次,选取频次最高的波长变量作为影响变量(IVs,influential variables)。研究结果表明,利用IVs建模可以将RMSEP值由0.5245减小到0.2548,采用蒙特卡罗采样技术的特征投影图方法(MC-LPG,monte carlo-latent projective graph)进行变量选择,对于提高模型预测精度是可行的。 相似文献
3.
蚁群算法在土壤质地高光谱预测建模中的应用 总被引：1，自引：0，他引：1
为提高土壤质地高光谱预测模型精度,以巢湖流域177个土样光谱为基础数据源,运用蚁群算法选择特征波长,结合BP神经网络构建土壤质地光谱预测模型,并与全光谱构建的光谱预测模型进行比较。结果表明,运用蚁群算法选择特征波长构建的光谱预测模型精度优于全光谱构建的预测模型精度,土壤粉粒含量预测模型预测集决定系数R2为0.76,RPIQ为2.23,土壤砂粒含量预测模型预测集决定系数R2为0.72,RPIQ为1.94;全光谱土壤粉粒含量预测模型预测集R2为0.57,RPIQ为1.75,全光谱土壤砂粒含量预测模型预测集R2为0.48,RPIQ为1.82。运用蚁群算法选择光谱特征波长建模,减少了数据冗余,提高了预测模型精度。 相似文献
4.
连续投影算法在猪肉pH值无损检测中的应用 总被引：4，自引：4，他引：0
利用鲜肉的近红外光谱中少量特征波长对其pH值进行预测,可以大幅度降低模型复杂性和计算量,对开发无损检测装置, 实施肉品生产加工过程中pH值监测有重要意义。该文通过连续投影算法（SPA）选择特征波长建立简单多元线性回归模型（SPA-MLR）,并对比了SPA-MLR模型与全波段（5 000～10 440 cm-1）偏最小二乘回归模型（PLSR）及逐步线性回归（SMLR）、遗传算法（GA）选择特征波长所建模型的性能。结果表明经连续投影算法提取37个特征波长建立的模型,所用变量数仅占全波段的2.6%,校正集相关系数0.870,校正集均方根误差为0.094,验证集相关系数0.892,验证集均方根误差为0.085;性能与经多元散射校正预处理的PLSR模型接近,但采用变量数明显减少,优于逐步线性回归和遗传算法选择特征波长建立的模型,表明该方法可较好的选择特征波长,建立简单的预测模型。 相似文献
5.
基于热红外发射率光谱的土壤盐分预测模型的建立与验证 总被引：1，自引：0，他引：1
该研究尝试性地分析盐渍化土壤热红外发射率光谱特征,并建立土壤盐分高光谱预测模型,旨在为遥感传感器识别土壤盐分信息奠定基础。首先,采用FTIR（Fourier transform infrared spectrometer）温度与发射率分离处理软件进行土壤温度和发射率的分离。运用高斯滤波平滑法对研究区野外测的土壤样品热红外发射率光谱数据进行滤波去噪处理。其次,对不同土壤含盐量热红外发射率光谱数据进行特征分析;然后在原始热红外发射率光谱数据的基础上进行4种形式的数学变换,分析热红外发射率光谱数据的变换处理形式与土壤含盐量之间的定量相关性。最后,使用多元回归方法建立预测模型并进行精度评价。结果表明：经过数据变换处理后,发射率光谱差异性有所提高;以平方根变换后的热红外光谱数据建立的预测模型效果较好,R2达到0.82。该研究将热红外遥感独特的发射率光谱特性应用于土壤盐渍化的实际科学问题中,为定量地分析盐渍土热红外发射率光谱信息提供参考。 相似文献
6.
运用高光谱技术鉴别玉米霉变等级时,因光谱波段数多、数据量大、信息冗余度高,使鉴别工作难度加大。为了减少数据量,获得最有利于鉴别的高光谱信息特征波长,本研究提出了一种连续投影算法(SPA)融合信息熵的特征波长选择方法。首先,对霉变玉米样本高光谱数据运用多元散射校正(MSC)进行光谱预处理以消除噪声,然后利用SPA对处理过的光谱进行波长初选,得到8个初选特征波长,再通过信息熵原理处理初选特征波长下的图像信息,获得最佳特征波长。结果表明,运用SPA融合信息熵法得到有利于霉变玉米鉴别的最佳波长为819 nm,提取该波长下霉变玉米图像的纹理特征后,采用Fisher判别分析(FDA)进行鉴别,6个等级霉变玉米的鉴别正确率高达98.6%,充分证明所给出的特征波长选择方法是有效的。本研究特征波长选择方法可为更好地运用高光谱技术鉴别玉米霉变等级提供指导。 相似文献
7.
基于VIS-NIR光谱的互花米草入侵湿地土壤有机碳预测研究 总被引：1，自引：0，他引：1
为了研究可见光—近红外光谱技术预测互花米草入侵背景下滨海湿地土壤有机碳含量的潜力,以江苏省典型互花米草湿地为研究对象,利用时空替代法采集15个土壤剖面3个深度共45个土壤样品。在实验室测定土壤光谱及有机碳含量,利用偏最小二乘回归方法建立了基于6种光谱变换的土壤有机碳预测模型,并分析了互花米草入侵年限和土层深度对土壤光谱和模型预测精度的影响。结果表明,表层土壤有机碳含量随互花米草入侵而显著增加。相对于仅包含光谱信息的预测模型,加入辅助变量(土层深度和植物入侵年限)建立的混合模型预测精度更高。交叉验证结果表明,基于光谱倒数1/R建立的混合模型预测精度最高,其决定系数(R~2)为0.68,预测相对分析误差(RPD)为1.6,是预测互花米草入侵湿地土壤有机碳含量的最优模型。本研究表明,利用可见光—近红外光谱技术可以对互花米草入侵湿地的土壤有机碳含量进行有效预测,土层深度和植物入侵年限辅助变量可以在一定程度上提高模型预测精度。 相似文献
8.
基于CARS算法的不同类型土壤有机质高光谱预测 总被引：10，自引：8，他引：2
不同土壤类型的理化性质和光谱性质存在差异,以往研究多以高光谱反射率或光谱吸收特征建立模型,输入变量类型结构单一,往往导致土壤有机质(Soil Organic Matter,SOM)预测模型的精度不高。为提高SOM高光谱预测模型精度,该研究以黑龙江省海伦市为研究区,将不同类型土壤分别以竞争自适应重加权采样(Competitive Adaptive Reweighted Sampling,CARS)筛选的特征波段、数字高程模型(Digital Elevation Model,DEM)数据和光谱指数作为输入变量,结合随机森林(Random Forest,RF)算法建立SOM预测模型。结果表明:1)通过CARS算法筛选后,各土壤类型特征波段压缩至全波段数目的16%以下,在很大程度上降低土壤高光谱变量维度和计算复杂程度,从而提高了模型的预测能力,说明CARS算法在提取特征关键波段变量、优化模型结构方面起到重要作用;2)不同类型土壤的SOM预测精度存在差异,沼泽土的预测精度最高为0.768,性能与四分位间隔距离的比率(Ratio of Performance to InterQuartile distance,RPIQ)为3.568;黑土次之,草甸土的预测精度最低,仅0.674,RPIQ为1.848。3类土壤的RPIQ均达到1.8以上,模型具有较好的预测能力;3)局部回归预测精度最优,验证集的调整后决定系数为0.777,均方根误差(Root Mean Square Error,RMSE)为0.581%,模型验证RPIQ为2.689,模型稳定性高。该试验筛选的预测因子通过RF模型可实现SOM含量的快速预测,简化了传统复杂的程序,可为中尺度区域不同类型土壤的SOM预测提供依据,为输入量的选择提供参考。 相似文献
9.
基于不同光谱变换的土壤盐含量光谱特征分析 总被引：4，自引：0，他引：4
跟踪初生盐渍土壤的微生物修复实验,采用同步实测得土壤盐含量和光谱数据,详细分析了基于34种光谱变换,修复过程中盐渍土的光谱特征。对于选取的6种光谱变换,采用全波段(400~1650 nm)和分析获得的最佳敏感波段分别建立了土壤盐含量的光谱反演PLSR(partial least squares regression)模型。研究表明,光谱变换处理使土壤盐含量与平滑后的光谱反射数据的相关性明显增强,且最佳敏感波段范围进一步聚焦。本研究得到最佳光谱变换为导数变换,基于全波段的土壤盐含量预测模型以SGSD变换效果最好,与原始光谱相比,模型的r、RMSEP分别从0.537和1.928改善到0.823和1.256。而SGSD(Log R)是基于最佳波段所建立的盐含量预测模型的有效光谱变换方法,该研究为进一步实现盐渍土中盐含量快速定量分析提供了方法和数据参考。 相似文献
10.
摘要：【目的】传统的基于近红外光谱数据预测土壤全氮的方法需要对原始光谱数据做复杂的预处理，筛选出与土壤全氮含量相关性高的敏感波长之后进行模型的回归拟合。本文提出一种一维卷积神经网络（1D-CNN）模型，可以在对数据进行简单预处理甚至无处理的情况下达到非常理想的结果，实现用近红外光谱技术对土壤全氮含量的预测。【方法】于江苏无锡采集410个土壤样品，利用半微量开氏法(NY/T 53-1987)测定土壤的全氮含量，并利用NIR Quest 512光谱仪，在室内环境下对每份土壤样品做光谱检测，并用均值中心化(CT)、标准正态变换(SNV)、趋势校正(DT)对光谱进行预处理，运用偏最小二乘回归（PLS）、BP神经网络、1D-CNN方法建立土壤全氮含量的回归预测模型。每种模型在采用不同预处理方法的数据集上做十折交叉验证，记录预测模型的决定系数(R2)和均方根误差(RMSE)的平均值，并对比三种预处理方法对模型精度的影响。【结果】证明了本文提出的1D-CNN模型基于土壤近红外光谱数据预测土壤全氮含量的可靠性。使用原始数据与经均值中心化、标准正态变换、趋势校正预处理的数据训练得到的1D-CNN模型的决定系数分别为0.907、0.931、0.922、0.964，构建的PLS回归模型决定系数为0.856、0.863、0.861、0.880，训练的BP神经网络的决定系数为0.874、0.907、0.901、0.911。【结论】本文提出的1D-CNN模型在原始数据和经预处理的光谱数据上的表现都优于PLS和BP神经网络，且可以证明，对光谱数据进行预处理能够有效提高1D-CNN模型的性能，尤其是趋势校正对模型的提升效果最明显。研究表明，1D-CNN能更好地提取光谱特征并建立其与含氮量的映射关系，有效地避免过拟合，在未经过预处理的光谱数据上依然能够达到一定的精度。 相似文献
11.
Reducing large spectral datasets to parsimonious representations of wavelengths is of value for efficient storage and easing analysis, in addition to the potential to use a simpler and cheaper spectrophotometer. This study evaluated the potential of calibrating visible and near infrared (vis‐NIR) spectra to total nitrogen (N), total carbon (C), organic C and inorganic C in soil on a 15‐ha farm, with the aim of comparing several wavelength reduction algorithms and rates in terms of model prediction accuracy. We explored the uninformative variables elimination (UVE), UVE coupled with successive projections algorithm (SPA) and two uniform‐interval wavelength reduction approaches (UWR‐I and UWR‐II) with successive wavelength reduction rates (WRRs) of 2, 5, 10, 20, 50, 100, 200, 500 and 1000. The standard normal variate (SNV)‐transformed absorbance spectra of soil samples recorded from 400 to 2499 nm at 1‐nm intervals were used. The calibration sets were subjected to a partial least squares regression (PLSR) with leave‐one‐out cross‐validation. Prediction results showed that UVE can reduce wavelength variables significantly while retaining good model prediction accuracy. The UVE‐SPA produced only three or four wavelengths, with which PLSR models achieved competitive prediction performance, compared with those based on all 2100 wavelengths, with coefficient of determination (R2) of 0.91, 0.89, 0.91 and 0.53 and residual prediction deviation (RPD) of 3.53, 2.95, 3.27 and 1.53 for soil total N, total C, organic C and inorganic C, respectively. The UWR tests showed that PLSR models responded insensitively to various WRRs from 2 to 100. The models calibrated for the 100‐nm interval spectra (21 remaining wavelengths) performed almost as well as those for the 1‐nm interval spectra. Although these findings might be valid only at the farm scale, it is recommended that the proposed wavelength reduction algorithms for more soil types and soils originated from larger areas should be examined. 相似文献
12.
Several algorithms exist for the calibration procedures of near‐infrared spectra in soil‐scientific studies, but the potential of a genetic algorithm (GA) for spectral feature selection and interpretation has not yet been sufficiently explored. Objectives were (1) to test the usefulness of near‐infrared spectroscopy (NIRS) for a prediction of C and N from char and forest‐floor Oa material in soils using either a partial least squares (PLS) method or a GA‐PLS approach and (2) to discuss the mechanisms of GA feature selection for the examined constituents. Calibration and validation were carried out for measured reflectance spectra in the visible and near‐IR region (400–2500 nm) on an existing set of 432 artificial mixtures of C‐free soil, char (lignite, anthracite, charcoal, or a mixture of the three coals), and forest‐floor Oa material. For all constituents (total C and N, C and N from all coals and from the Oa material, C derived from mixed coal, charcoal, lignite, and anthracite), the GA‐PLS approach was superior over the full‐spectrum PLS method. The RPD values (ratio of standard deviation of the laboratory results to standard error of prediction) ranged from 2.4 to 5.1 in the validation and indicated a better category of prediction for three constituents: “approximate quantitative” instead of a “distinction between high and low” for C derived from mixed coal and “good” instead of “approximate quantitative” for C and N derived from all coals. Overall, this study indicates that the approach using GA may have a greater potential than the PLS method in NIRS. 相似文献
13.
Lauric Cécillon Nathalie Cassagne Sonia Czarnes Raphaël Gros Jean-Jacques Brun 《Soil biology & biochemistry》2008,40(7):1975-1979
Near infrared reflectance spectroscopy (NIRS) was used to predict six biological properties of soil and earthworm casts including extracellular soil enzymes, microbial carbon, potential nitrification and denitrification. Partial least squares regression (PLSR) models were developed with a selection of the most important near infrared wavelengths. They reached coefficients of determination ranging from 0.81 to 0.91 and ratios of performance-to-deviation above 2.3. Variable selection with the variable importance in the projection (VIP) method increased dramatically the prediction performance of all models with an important contribution from the 1750–2500 nm region. We discuss whether selected wavelengths can be attributed to macronutrient availability or to microbial biomass. Wavelength selection in NIR spectra is recommended for improving PLSR models in soil research. 相似文献
14.
应用红外光声光谱技术及支持向量机模型测定土壤有机质含量 总被引：2，自引：0，他引：2
快速测定土壤有机质含量对作物生产和土壤肥力评价具有重要意义,红外光声光谱技术的应用为土壤有机质快速测定提供了可能。本研究以江苏省南京市溧水区水稻土土样为材料,探究了红外光声光谱技术在有机质测定中的应用。采用主成分分析、偏最小二乘和独立成分分析,分别提取了土壤光谱的主成分、偏最小二乘潜变量和独立成分,并以提取的信息输入支持向量机,从而构建了三种支持向量机校正模型。同时,偏最小二乘也被用于建立校正模型,作为支持向量机模型的对照。预测结果表明,基于独立成分的支持向量机模型效果最好,预测相关系数R2、均方根误差RMSEP和实际测量值的标准差与光谱模型预测值标准差的比值即RPD值分别为0.808、0.575和2.28。F检验表明,该模型显著优于基于主成分的支持向量机模型,但与基于偏最小二乘潜变量的支持向量机模型,以及经典偏最小二乘模型没有显著差异。t检验表明,各校正模型对有机质的预测结果与化学测定结果没有显著差异。因此,红外光声光谱技术为土壤有机质的快速测定提供了新的技术手段。 相似文献
15.
土壤有机质可见光-近红外光谱预测样本优化选择 总被引：2，自引：0，他引：2
土壤有机质可见光-近红外光谱预测中建模样本的优化选择对提高有机质模型估算精度具有重要作用。本文以湟水流域土壤有机质为例,采用基于土壤单一属性信息考虑的建模样本选择方法:浓度梯度法、Kennard-Stone(KS)方法,以及基于土壤多种信息考虑的建模样本选择方法:Rank-KS(RKS)法、土壤类型结合浓度梯度法以及土壤类型结合KS法。通过偏最小二乘回归建模,探索可见光–近红外光谱预测青海湟水流域有机质的最优样本集。结果表明:不同级别样本数的最佳建模样本选择方法不同,整体表现为基于土壤多种信息挑选的建模样本集的模型精度相比土壤单一信息均较高,特别是KS方法结合土壤类型后的建模样本集模型精度明显提高且在样本数较少时更为明显。土壤类型可以优化建模样本选择方法提高模型预测精度。在保证固定验证样本模型预测精度的情况下,土壤类型参与建模样本的选择可以有效减少建模样本数,进而降低了建模成本。 相似文献
16.
Shengxiang Xu Meiyan Wang Xuezheng Shi 《Acta Agriculturae Scandinavica, Section B - Plant Soil Science》2019,69(5):452-464
This study evaluates the use of visible and near-infrared spectroscopy for rapid prediction of total carbon, total nitrogen, and total phosphorus concentrations in field crop samples. Two multivariate models (partial least squares regression and support vector machine regression) were compared. In addition, four spectral variable selection algorithms (competitive adaptive reweighted sampling, genetic algorithm, uninformative variable elimination, and variable importance for projection) were applied with support vector machine regression to determine the most accurate predictions. The results showed that support vector machine regression performed better than partial least squares regression for predicting the three chemical compositions. The combination of competitive adaptive reweighted sampling and support vector machine regression outperformed the other models for the predictions of total carbon and total nitrogen with high coefficients of determination of 0.91 and 0.90, respectively. For the determination of total phosphorus, the prediction accuracy of competitive adaptive reweighted sampling was comparable with the best result obtained from genetic algorithm with the coefficients of determination of 0.73 and 0.77, respectively. In conclusion, the support vector machine regression combined with competitive adaptive reweighted sampling has great potential to accurately determine the chemical composition of field crops using the visible and near-infrared spectroscopy. 相似文献
17.
绿原酸(chlorogenicacid,CGA)是评价金银花品质的重要指标。为了实现金银花贮藏期间CGA含量变化的快速有效检测,该文采集了500个不同贮藏时间(0～20d)的金银花高光谱图像,构建CGA含量的高光谱检测模型。为了提高模型性能,采用savizky-golay卷积平滑(SG),移动窗口平滑(moving average),标准正态变量(standard normal variable,SNV),基线校正(baseline correction,BC),多元散射校正(multiplicative scatter correction,MSC),正交信号校正(orthogonal signal correction,OSC)6种预处理方法并建立偏最小二乘回归(partial least squares regression,PLSR)模型,确定SNV方法为最佳预处理方法,其预测集的R2为0.976 6,RMSE为0.271 1%。为了简化校准模型,利用无信息变量消除(uninformative variable elimination,UVE),连续投影算法(successive projections algorithm,SPA),竞争性自适应加权算法(competitive adaptive reweighted sampling,CARS)以及UVE-CARS、UVE-SPA等方法对SNV预处理后的光谱提取特征波长。然后,分别基于全光谱数据和所选特征变量数据,建立线性偏最小二乘回归(PLSR)和非线性BP神经网络模型。结果表明:UVE-CARS算法可以有效地减少提取变量个数(共提取26个,仅占全光谱范围的3.2%),PLSR和BP模型的预测集R2分别为0.974 6和0.978 4,RMSE分别为0.286 3%和0.250 3%。非线性BP模型预测结果整体优于线性PLSR模型,在BP模型中,UVE-CARS-BP预测精度最高,预测集的R2和RMSE的值分别为0.978 4, 0.250 3%。综上,基于高光谱成像技术建立的SNV-UVE-CARS-BP模型,可以实现金银花贮藏过程中CGA含量变化的快速无损预测。 相似文献
18.
Characterizing spatial variability of soil attributes, using traditional soil sampling and laboratory analysis, is cost prohibitive. The potential benefit of managing soils on a site-specific basis is well established. High variations in glacial till soil render detailed soil mapping difficult with limited number of soil samples. To overcome this problem, this paper demonstrates the feasibility of soil carbon and clay mapping using the newly developed on-the-go near-infrared reflectance spectroscopy (NIRS). Compared with the geostatistics method, the partial least squares regression (PLSR), with NIRS measurements, could yield a more detailed map for both soil carbon and clay. Further, by using independent validation dataset, the accuracy of predicting could be improved significantly for soil clay content and only slightly for soil carbon content. Owing to the complexity of field conditions, more work on data processing and calibration modeling might be necessary for using on-the-go NIRS measurements. 相似文献
19.
Mid‐infrared spectroscopy (MIRS) is a well‐established analytical tool for qualitative and quantitative analysis of soil samples. However, effects of soil sample grinding procedures on the prediction accuracy of MIR models and on qualitative spectral information have not been well investigated and, in consequence, not standardized up to now. Further, the effects of soil sample selection on the accuracy of MIR prediction models has not been quantified yet. This study investigated these effects by using 180 well‐characterized soil samples that were ground for different times (0, 2 or 4 minutes) and then used for MIR measurements. To study the impact of sample preparation, soil spectra were subjected to principal component analyses (PCA), multiple regression and partial least square (PLS) analysis. The results indicate that the prediction accuracy of MIR models for soil organic carbon (SOC) and pH and the qualitative spectral information were better overall for lightly ground (2 minutes) soil samples compared with intensively (4 minutes) or unground soil samples. Whereas the grinding procedure did not show any effect on spectra of clay minerals, spectral information for quartz and for SOC was modified. Even though it is difficult to recommend a global standardized soil sample grinding procedure for MIR measurements because of different mill types available within laboratories, we highly recommend using an internally standardized grinding procedure. Moreover, we show that neither land use nor soil sampling depth influences the prediction of the SOC content. However, sand and clay content substantially affect the score vectors used by the PLS algorithm to predict the SOC content. Thus, we recommend using soil samples similar in texture for more precise SOC calibration models for MIR spectroscopy. 相似文献