Root Mean Standard Error Of Prediction Research Articles

The utility of a near-infrared spectrometer to predict the maturity of green peas (Pisum sativum)

This work evaluated the utility of an embedded diode array near-infrared reflectance spectrometer to predict the hardness (maturity) of green peas. Partial least squares, support vector machine (SVM) and artificial neural network regression models were developed to predict tenderometer reading (TR) from near-infrared spectra over the range of maturities experienced throughout a harvest season. The SVM model achieved the best performance in an independent data set collected in a second harvest season. Here, the SVM model explained 83% of the variation in TR with a root mean standard error of prediction (RMSEP) of 4.4 TR units over a range of TR from 65 to 104. The RPD of 2.3 achieved in this study exceeds the threshold for rough screening but falls short of replacing the tenderometer for quality control.

Near Infrared Reflectance Spectroscopy and Multivariate Analyses for Fast and Non-Destructive Prediction of Corn Seed Germination

The application of near-infrared reflectance spectroscopy (NIRS) and multivariate analysis for determining the seed germination rate of corn genotypes was assessed. Seed samples about 90 gr belong to commercial and local corn varieties at various ages were scanned with FT-NIRS on the reflectance mode from 1000 to 2500 nm wavelength. Filter paper technique showed the seed germination rates varied between 18-100% depending on the genotypes after 7 days at ±25°C. Partial least squares regression (PLSR) was applied to the reference values corresponding to the spectra. The best statistical results obtained from the pre-treatment combinations of Smooth Savitzky-Golay 9 Points (sg9), MSC full and normalization to unit length (nle). The regression coefficient of calibration (R2C) and prediction (R2P) of the created NIRS calibration via chemometric software NIRCal are realized 0.97 and 0.98 respectively for the property of corn germination rate. The standard error of both calibration (SEC) and prediction (SEP) were almost overlapping (4.17%, 4.61% respectively). The prediction accuracy of the final NIRS model was quite reasonable with the acceptable root mean standard error of prediction (RMSEP) as 8.88%. According to the residual predictive deviation (RPD) index (4.18), the accuracy of the NIRS model regarded as in the best category. Therefore, the NIRS model developed here is sufficient to predict the corn seed germination rate very fast and non-destructively without using any regents.

Terahertz spectroscopy and machine learning algorithm for non-destructive evaluation of protein conformation

Given the condition that protein conformation and activity are highly susceptible to environment factors such as temperature and pH, evaluation of protein conformation and activity is urgently needed in many fields. For example, most protein drugs need a stable and proper environment during production, storage and transportation, and it’s an enormous challenge to maintain protein activity throughout the whole process. Therefore, it’s necessary to ensure the safety and effectiveness of protein drugs by monitoring their activity before use. In our study, we presented an improved method for non-destructive evaluation of protein conformation and biological activity by terahertz spectroscopy combined with t-SNE-XGBoost. Firstly, bovine serum albumin (BSA) samples heated to different temperature were measured with THz-TDS. The obtained results indicated that native-conformation BSA will undergo transient states in the process of temperature induced denaturation. However, for any single given sample, it’s difficult to identify its conformation and activity directly by using the measured raw terahertz data. Therefore, we applied several different algorithms to the raw data for recognition of BSA samples with different conformation and activity induced by temperature. Finally, the models obtained by different algorithms were evaluated by calculating the root mean standard error of prediction (RMSEP) and the correlation coefficient of prediction ($$R_p$$). The THz-TDS plus t-SNE-XGBoost proved to be an effective non-destructive and label-free method for evaluation of protein conformation and activity. It can provide a new technique in many applications, such as pharmaceutical industry, clinical diagnosis and quality control.

Data Fusion Approach Improves the Prediction of Single Phenolic Compounds in Honey: A Study of NIR and Raman Spectroscopies

The combination of Near-Infrared Spectroscopy (NIR) and Raman Spectroscopy (RS) of 100 honey samples collected from different countries were used to develop the calibration model for determination of single phenolic compound. In high-performance liquid chromatography with diode-array detection analysis, 16 phenolic compounds were identified with p-hydroxybenzoic acid being the major compound detected in all honey samples. Thus, p-hydroxybenzoic acid was used for developing prediction models. Spectral data were modeled individually and using data fusion methodologies. The performance of the model based on RS spectra [\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$R_p^2 = 0.9500$$\\end{document}, Root Mean Standard Error of Prediction (RMSEP) = 6.83] was higher than that based on the NIR spectra (\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$R_p^2 = 0.8147$$\\end{document}, RMSEP = 13.80). The application of both low-level (\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$R_p^2 = 0.9553$$\\end{document}, RMSEP = 6.59) and mid-level (\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$R_p^2 = 0.9563$$\\end{document}, RMSEP = 7.95) data fusion together with Partial Least Squares (PLS) had effectively improved the prediction models of NIR but did not enhance prediction models based on RS technique. The results demonstrated that the NIR, RS, and data fusion approaches together with the PLS model could be used as alternative quantitative methods for determination of p-hydroxybenzoic acid in honey samples.

Modeling for SSC and firmness detection of persimmon based on NIR hyperspectral imaging by sample partitioning and variables selection

Nondestructive detecting for persimmon’s internal quality is meaningful for post-harvest processing. This study focused on the modeling of soluble solid content (SSC) and firmness (FM) determination of persimmon with near-infrared (NIR) hyperspectral imaging within 900–1700 nm. Sample partitioning and variables selecting were used to optimize the partial least squares (PLS) regression detective models. Three sample partitioning methods included Kennard-stone (KS) algorithm, sample set partitioning joint x-y distances algorithm (SPXY) and random selection (RS) methods were adopted. Monte Carlo uninformative variables elimination (MC-UVE), competitive adaptive reweight sampling method (CARS) and successive projection algorithm (SPA) were applied for feature variables selection. For SSC and FM detecting, the best models were SPXY-MC-UVE-CARS-PLS model with 12 feature variables and Savitzky-Golay-RS-CARS-PLS with 7 feature variables respectively. The 12 and 7 feature variables are grouped together to build PLS models for SSC and FM determination and after the evaluation the regression coefficient of each variables, finally 10 and 9 selected wavelengths were selected to SSC and FM detection respectively. The final models obtained coefficient of determination (Rp2) of 0.757, root mean standard error of prediction (RMSEP) of 1.404 OBrix and Rp2 of 0.876, RMSEP of 0.395 kg/cm2 for SSC and FM detection respectively. Meanwhile, we obtained the SSC and FM distribution maps which could give help to visual detection. The results in this study could provide reference for the development of online classification equipment with multi-indicators detection for persimmon.

Rapid and simultaneous determination of physical and chemical properties of asphalt by ATR-FTIR spectroscopy combined with a novel calibration-free method

Determining wax content, softening point, and penetration of asphalt products by traditional methods is too time-consuming to meet the instant need in practice. Although infrared spectroscopy combined with multivariate calibration may realize rapid determination of asphalt, its modeling and maintenance are very labor-intensive and challenging. In this paper, a novel method combining attenuated total reflectance-Fourier transform infrared spectroscopy (ATR-FTIR) with a fast algorithm of spectrum representation (SR) was proposed to rapidly and expediently determine the properties. The proposed method achieves calibration-free, effectively overcoming the drawback of the modeling methods. The feasibility of the new method was demonstrated using 431 asphalt samples produced by different kinds of crude oil. The root mean standard error of prediction (RMSEP) of wax content, softening point, and penetration by SR method are 0.14%, 0.55 °C and 4.71 (0.1 mm), respectively. Each of them is below the reproducibility of corresponding standard test method. Moreover, the new method is compared with other two commonly used methods, partial least squares regression (PLS) and local modeling by densification (LMD), in detail. The results show that the new method not only can solve the disadvantages of PLS modeling effectively, but also is remarkably superior to LMD method, in repeatability and speed, and robustness for predicting the sample in the region with lower density and unreasonable distribution of the samples in data bank.

Dynamic Modeling of Carbon Metabolism During the Dormant Period Accurately Predicts the Changes in Frost Hardiness in Walnut Trees Juglans regia L.

The leafless period is often considered as inactive, although trees have to actively modulate their metabolism through the cold acclimation/deacclimation processes, to cope with frost exposure during winter and to restore growth ability in spring. Carbon metabolism is a key component of these processes through the osmotic control of extracellular ice formation and the trophic control of bud growth. The influence of temperature on the inter-conversion between starch and soluble carbohydrate has been evidenced for years, but we are currently missing an operational tool to predict starch vs. soluble carbohydrate contents during this period, which should allow to better predict frost hardiness. For this purpose, we exposed 1-year-old branches of Juglans regia to constant temperature for one to 3 weeks and measured the changes in carbohydrate composition at three periods (autumn, winter, and spring). As expected, the temperature significantly affected the changes in carbohydrate composition, but the water content and the sampling period were also relevant. Higher starch hydrolysis was observed at low temperature (<5°C) for all sampling periods. Starch hydrolysis was also observed at warm temperature, but in autumn only. These data were used to compare three modeling approaches simulating the changes in carbohydrate composition through enzymatic analogy. The most empirical and the most mechanistic approach did not succeed to simulate external observations (Root Mean Standard Error of Prediction (RMSEP) > 30 mg.g DM−1, Efficiency (Eff) <0), whereas the intermediate model was more efficient (RMSEP = 15.19 mg.g DM−1, Eff = 0.205 and 16.61 mg.g DM−1, Eff = 0.366, for GFS (Glucose + Fructose + Sucrose) and starch, respectively). The accuracy of the model was further improved when using field data for calibration (RMSEP = 5.86 mg.g DM−1, Eff = 0.962; RMSEP = 10.56 mg.g DM−1, Eff = 0.752, for GFS and starch, respectively). This study provided an operative tool to simulate carbohydrate dynamics over leafless period that could predict frost hardiness with approx. 3.4°C accuracy with temperature, water content and initial starch and soluble carbohydrate measurements. It should now be tested under various meteorological conditions and biological systems.

Assessment of Robustness for a Near-Infrared Concentration Model for Real-Time Release Testing in a Continuous Manufacturing Process

This study describes the robustness test of a transmission near-infrared spectroscopic (NIRS) model for prediction of drug concentration in core tablets as part of real-time release testing (RTRt) efforts for a continuous manufacturing process. Seven calibration blend samples were prepared spanning a concentration range from 70 to 130% of the active pharmaceutical ingredient (API) target level with tablets prepared with laboratory-scale equipment, and their NIR spectra were obtained to develop a partial least squares (PLS) calibration model. The calibration model’s accuracy was assessed with an independent set of tablets obtaining a root mean standard error of prediction (RMSEP) of less than 1.5% label claim at each concentration level. However, a decrease in the prediction of API concentration of tablets exposed to the production environment was observed indicating lack of robustness as a RTRt method. The API is known to interchange solvate molecules with water molecules from the environment. Even though this exchange does not affect product quality, it affects the NIR spectra. The calibration model was then optimized by excluding O–H bands from the spectral range and shown to predict accurately the drug concentration in tablets for up to 113 h after tablet compaction. The calibration model was updated by including tablets produced in the continuous manufacturing line within the calibration model. The updated calibration model was used to predict tablets from three different continuous manufacturing runs, involving different lots of excipients and drug. The results from the updated calibration model showed excellent agreement with reference HPLC results.

Total Anthocyanin Content in Intact Açaí (Euterpe oleracea Mart.) and Juçara (Euterpe edulis Mart.) Fruit Predicted by Near Infrared Spectroscopy

Açaí (Euterpe oleraceae Mart.) and juçara (Euterpe edulis Mart.) palms are native to Brazil and these species are rich in anthocynanins. The methods applied to determine anthocyanins are time-consuming, generate chemical residues, and do not fit in modern on-line grading machines. As near infrared (NIR) spectroscopy has been used as a nondestructive method to determine anthocyanin, the objective of this study was to use NIR spectroscopy to predict total anthocyanin (TA) in intact açaí and juçara fruits. Spectra were collected using a Fourier transform (FT)-NIR spectrophotometer in the diffuse reflectance (4,000–10,000 cm−1) and TA reference data were obtained using the Association of Official Analytical Chemists (AOAC) method. Different treatments were applied to spectra and spectral data sets were correlated with TA by using partial least squares (PLSs) regression algorithm. The global-PLS model obtained with açaí and juçara spectra resulted in a root mean standard error of prediction (RMSEP) of 10.05 g·kg−1. However, this model was not adequate for TA levels found in açaí fruits, therefore individual models were developed. The açaí-PLS model proved to be more adequate, as RMSEP was reduced to 3.56 g·kg−1. On the other hand, the RMSEP obtained with the juçara-PLS model (6.59 g·kg−1) was almost the same of the global model. NIR spectroscopy can be used to adequately predict TA content in intact açaí and juçara fruits and this method could be used as an analytical procedure to monitor their quality.

Robust Ultraviolet-Visible (UV-Vis) Partial Least-Squares (PLS) Models for Tannin Quantification in Red Wine.

The validation of ultraviolet-visible (UV-vis) spectroscopy combined with partial least-squares (PLS) regression to quantify red wine tannins is reported. The methylcellulose precipitable (MCP) tannin assay and the bovine serum albumin (BSA) tannin assay were used as reference methods. To take the high variability of wine tannins into account when the calibration models were built, a diverse data set was collected from samples of South African red wines that consisted of 18 different cultivars, from regions spanning the wine grape-growing areas of South Africa with their various sites, climates, and soils, ranging in vintage from 2000 to 2012. A total of 240 wine samples were analyzed, and these were divided into a calibration set (n = 120) and a validation set (n = 120) to evaluate the predictive ability of the models. To test the robustness of the PLS calibration models, the predictive ability of the classifying variables cultivar, vintage year, and experimental versus commercial wines was also tested. In general, the statistics obtained when BSA was used as a reference method were slightly better than those obtained with MCP. Despite this, the MCP tannin assay should also be considered as a valid reference method for developing PLS calibrations. The best calibration statistics for the prediction of new samples were coefficient of correlation (R2val) = 0.89, root mean standard error of prediction (RMSEP) = 0.16, and residual predictive deviation (RPD) = 3.49 for MCP and R2val = 0.93, RMSEP = 0.08, and RPD = 4.07 for BSA, when only the UV region (260-310 nm) was selected, which also led to a faster analysis time. In addition, a difference in the results obtained when the predictive ability of the classifying variables vintage, cultivar, or commercial versus experimental wines was studied suggests that tannin composition is highly affected by many factors. This study also discusses the correlations in tannin values between the methylcellulose and protein precipitation methods.

Application of FT-NIR Spectroscopy to Apple Wine for Rapid Simultaneous Determination of Soluble Solids Content, pH, Total Acidity, and Total Ester Content

The feasibility of rapid simultaneous analysis of soluble solid content (SSC), PH, total acidity, and total ester content in apple wine by Fourier transform near-infrared (FT-NIR) spectroscopy together with partial least squares (PLS) models was studied in this work. A set of 380 apple wine samples were scanned in a quartz cuvette with a 1-mm path length in the NIR region (4,000–12,000 cm−1) in transmission mode. Calibration models were established by PLS with full cross-validation. The NIR spectra were pretreated with different methods, and the effective wave number range together with the number of PLS factors used in the model calibration was optimized. The optimal models yielded the determination coefficients of 0.91, 0.93, 0.98, and 0.92 and the root mean standard error of prediction (RMSEP) of 0.601, 0.077, 0.021, and 0.096 for SSC, pH, total acidity, and total ester content, respectively. This work demonstrated that NIR spectroscopy was a rapid and convenient technique for analyzing and predicting SSC, pH, total acidity, and total ester content simultaneously during apple wine fermentation process.

A Model System and Chemometrics to Develop near Infrared Spectroscopic Monitoring for Chinese Hamster Ovary Cell Cultivations

Near infrared (NIR) spectroscopy is an ideal tool for biopharmaceutical process monitoring, as it can be used to generate information on key bioprocess variables rapidly online. Mammalian cell cultivation is a rapidly developing field of biopharmaceutical production where one of these key process variables is the glucose concentration. However, calibrations for the NIR-based monitoring of glucose are usually not available in the early phase of process development owing to the lack of sufficient NIR data from small-scale experiments. This article demonstrates the development of calibrations based on NIR spectroscopy (11,988.0–4297.0 cm−1) for the determination of glucose concentration in a novel shake flask model system for mammalian cell cultivation. To generate a homogeneous distribution of glucose concentration in the calibration range and to reduce the correlation between glucose and other metabolites, cultivation samples with low glucose levels were spiked with glucose, and NIR measurements were subsequently performed. Biochemical and physical variability was deliberately induced in the model system to mimic possible matrix effects typically occurring in fed-batch bioreactor cultivation samples. The specificity of the calibration model to glucose was increased by variable selection, so the spectral region between the two water peaks was used for calibration. Four mathematical pretreatments were evaluated by comparing the root-mean standard error of prediction (RMSEP) values of the models; the best preprocessing method was a novel combination of baseline offset method with the deresolve function. Our models predicted the glucose concentration in two test cultivations with an RMSEP of 3.12 mmol L−1 and 5.51 mmol L−1.

Protein and Oil Composition Predictions of Single Soybeans by Transmission Raman Spectroscopy

The soybean industry requires rapid, accurate, and precise technologies for the analyses of seed/grain constituents. While the current gold standard for nondestructive quantification of economically and nutritionally important soybean components is near-infrared spectroscopy (NIRS), emerging technology may provide viable alternatives and lead to next generation instrumentation for grain compositional analysis. In principle, Raman spectroscopy provides the necessary chemical information to generate models for predicting the concentration of soybean constituents. In this communication, we explore the use of transmission Raman spectroscopy (TRS) for nondestructive soybean measurements. We show that TRS uses the light scattering properties of soybeans to effectively homogenize the heterogeneous bulk of a soybean for representative sampling. Working with over 1000 individual intact soybean seeds, we developed a simple partial least-squares model for predicting oil and protein content nondestructively. We find TRS to have a root-mean-standard error of prediction (RMSEP) of 0.89% for oil measurements and 0.92% for protein measurements. In both calibration and validation sets, the predicative capabilities of the model were similar to the error in the reference methods.

Selection of wavelength for strawberry NIR spectroscopy based on BiPLS combined with SAA

To improve and simplify the NIR prediction model of the soluble solid content(SSC) of strawberry,backward interval partial least squares(BiPLS) and simulated annealing algorithm(SAA) were combined to select the efficient wavelengths.The strawberry spectra were divided into 21 intervals,among which 4 subsets,i.e.No.8,13,16 and 17 were selected by BiPLS.Then SAA was used to select variables in these informative regions,which were used for regression variables of MLR model.Finally,7565 cm-1,7706 cm-1,8289 cm-1,8489 cm-1,8499 cm-1,8724 cm-1 and 8807 cm-1 were used to build a MLR model.The MLR model performs well with root mean standard error of prediction(RMSEP) of 0.428 for SSC,which out performs models using PLS and BiPLS.This work proved that the BiPLS-SAA could determine optimal variables in NIR spectra and improve the accuracy of model.

Determination of moisture content in lyophilized mannitol through intact glass vials using NIR micro-spectrometers

Determination of moisture content in lyophilized solids is fundamental to predict quality and stability of freeze-dried products, but conventional methods are time-consuming, invasive and destructive. The aim of this study was to develop and optimize a fast, inexpensive, noninvasive and nondestructive method for determination of moisture content in lyophilized mannitol, based on an NIR micro-spectrometer instead of a conventional NIR spectrometer. Measurements of lyophilized mannitol were performed through the bottom of rotating glass vials by means of a reflectance probe. The root mean standard error of prediction (RMSEP) and the correlation coefficient (R²pred), yielded by the pre-treatments and calibration method proposed, was 0.233% (w/w) and 0.994, respectively.

The prediction of soil chemical and physical properties from mid-infrared spectroscopy and combined partial least-squares regression and neural networks (PLS-NN) analysis

This study compares the performance of partial least-squares (PLS) regression analysis for the prediction of a wide range of soil chemical and physical properties from their mid-infrared (MIR) spectra with those from a combination of PLS regression and neural networks (NN). A combination of PLS and NN, referred to as PLS-NN, uses a relatively few PLS scores as inputs to the NN. This has the advantages of the robustness and qualitative and quantitative features of PLS and the non-linear capabilities of neural networks. In this study, the PLS-NN method outperformed the basic PLS regression for the prediction of some soil properties from MIR spectra of soils from throughout New South Wales, Australia. The coefficient of determination ( R 2) and root-mean standard error of prediction (RMSEP) for total organic carbon (TOC) were improved from an R 2 = 0.87 and RMSEP = 0.7 by PLS, to an R 2 = 0.94 and RMSEP = 0.5 by PLS-NN. Predictions appeared to be most improved where PLS regression curvature was apparent or when the analytical value distributions were heavily skewed to low values with many negative predictions, e.g. for exchangeable-Al and Mg, TOC, CEC, P-sorption and moisture content. For other soil properties there appeared to be a significant improvement in the lower 25 percentile of analyte values but the prediction errors increased at high analyte values and no net improvement was achieved, or were only marginal for the full range of data. In cases where PLS failed to give a viable model, NN also could not derive a converging model. The use of PLS-NN over the usual PLS method for routine soil analytical applications must be questioned with regard to a tradeoff between some possible limited improvement versus the added computational complexity and additional software requirement for NN.

Rapid estimation of single cell oil content of solid-state fermented mass using near-infrared spectroscopy

Calibration model using near-infrared reflectance spectroscopy (NIRS) for estimation of SCO content in solid-state fermented mass was established. The NIRS calibration model was derived by partial least-squares (PLS) regression and prediction of SCO contents of independent solid-state fermented mass samples fermented by different oleaginous fungi showed the model to be rapid and accurate, giving R 2-value higher than 0.9552 and root mean standard error of prediction (RMSEP) value lower than 0.5772%. The established NIRS calibration model could be used to estimate the SCO contents of the solid-state fermented masses and will provide much convenience to the research of SCO production in solid-state fermentation.

Preliminary investigations on the effects of ageing and cooking on the Raman spectra of porcine longissimus dorsi

The influence of ageing and cooking on the Raman spectrum of porcine longissimus dorsi was investigated. The rich information contained in the Raman spectrum was highlighted, with numerous changes attributed to changes in the environment and conformations of the myofibrillar proteins. Predictions equations for shear force and cooking loss were developed from the Raman spectra of both raw and cooked pork. Good correlations and standard errors of prediction were obtained for both WB shear force and cooking loss, with the raw and the cooked samples showing almost identical results R 2 = 0.77, root mean standard error of prediction (RMSEP)% of mean = 12% for shear force; R 2 = 0.71, RMSEP% of mean = 10% for cooking loss. The Raman spectra were also able to predict the extent of cooking that occurred within the pork ( R 2 val = 0.94, RMSEP% of range = 5.5%). Raman spectroscopy has considerable potential as a method for non-destructive and rapid determination of pork quality parameters such as tenderness. Raman spectroscopy may provide a means of determining changes during cooking and the extent to which foods have been cooked.

The use of FT-NIR for API content assay in organic solvent media: A single calibration for multiple downstream processing streams

The use of near infrared spectroscopy (NIRS) in downstream solvent based processing steps of an active pharmaceutical ingredient (API) is reported. A single quantitative method was developed for API content assessment in the organic phase of a liquid–liquid extraction process and in multiple process streams of subsequent concentration and depuration steps. A new methodology based in spectra combinations and variable selection by genetic algorithm was used with an effective improvement in calibration model prediction ability. Root mean standard error of prediction (RMSEP) of 0.05 in the range of 0.20–3.00% (w/w) was achieved. With this method, it is possible to balance the calibration data set with spectra of desired concentrations, whenever acquisition of new spectra is no longer possible or improvements in model's accuracy for a specific selected range are necessary. The inclusion of artificial spectra prior to genetic algorithms use improved RMSEP by 10%. This method gave a relative RMSEP improvement of 46% compared with a standard PLS of full spectral length.

Real-Time Moisture Measurement on a Forage Harvester Using Near-Infrared Reflectance Spectroscopy

A mobile, diode-array NIR spectrometer was integrated into the spout of a self-propelled forage harvester to measure crop moisture. Spectra and moisture reference samples were collected in 2004 and 2005 for the development of laboratory and field-based moisture calibrations. Moisture prediction models for whole-plant corn silage (WPCS) developed using laboratory data had a root mean standard error of cross-validation (RMSECV) of 1.1% using five principle components (PCs), while a calibration developed using field data had an RMSECV of 3.3% using four PCs. Alfalfa validation results produced RMSECVs of 2.5% using four PCs and 3.7% using three PCs for models using laboratory and field data, respectively. Field data were predicted with calibrations developed using laboratory data with similar error levels, but more spectral information was required. A laboratory-based alfalfa model predicted field data with a root mean standard error of prediction (RMSEP) of 3.4% using three PCs as compared to the field model's RMSECV of 3.7% using three PCs. Similar trends were found with WPCS models. Predicting data independent of type of crop resulted in the utilization of more PCs but with higher RMSEPs than the cross-validation results of the predicted dataset. The sensor and associated calibrations were able to predict forage moisture adequately, although more diverse data and further calibration development are needed to improve sensor accuracy to the desired range of 2.0 percentage units.