• Proceedings of the Korean Society of Near Infrared Spectroscopy Conference
• /
• 2001.06a
• /
• pp.1620-1620
• /
• 2001

Forages, both grazed and conserved, provide the basis of ruminant production systems throughout the world. More than 90 per cent of the feed energy consumed by herbivorous animals world - wide were provided by forages. With such world - wide dependence on forages, the economic and nutritional necessity of been able to characterize them in a meaningful way is vital. The characterization of forages for productive animals is becoming important for several reasons. Relative to conventional laboratory procedures, Near Infrared Reflectance Spectroscopy (NIRS) offers advantages of simplicity, speed, reduced chemical waste, and more cost-effective prediction of product functionality. NIR spectroscopy represents a radical departure from conventional analytical methods, in that entire sample of forage is characterized in terms of its absorption properties in the near infrared region, rather than separate subsamples being treated with various chemicals to isolate specific components. This forces the analyst to abandon his/her traditional narrow focus on the sample (one analyte at a time) and to take a broader view of the relationship between components within the sample and between the sample and the population from which it comes. forage is usually analysed by NIRS in dry and ground presentation. Initial success of NIRS analysis of coarse forages suggest a need to better understand the potential for analysis of minimally processed samples. Preparation costs and possible compositional alterations could be reduced by samples presented to the instrument in undried and unground conditions. NIRS has gained widespread acceptance for the analysis of forage quality constituents on dry material, however little attention has been given to the use of NIRS for chemical determinations on undried and unground forages. Relatively few works reported the use of NIRS to determine quality parameters on undried materials, most of them on both grass and corn silage. Only two works have been found on the determination of quality parameters on fresh forages. The objectives of this paper were (1) to evaluate the use of NIRS for determination of nitrogen and moisture on undried and unground forage samples and (2) to explore two mathematical treatments and two NIR regions to predict chemical parameters on fresh forage. Four hundred forage samples (n: 400) were analysed in a NIRS 6500 instrument (NIR Systems, PA, USA) in reflectance mode. Two mathematical treatments were applied: 1,4,4,1 and 2,5,5,2. Predictive equations were developed using modified partial least squares (MPLS) with internal cross - validation. Coefficient of determination in calibration (${R^2}_{CAL}$) and standard error in cross-validation (SECV) for moisture were 0.92 (12.4) and 0.92 (12.4) for 1,4,4,1 and 2,5,5,2 respectively, on g $kg^{-1}$ dry weight. For crude protein NIRS calibration statistics yield a (${R^2}_{CAL}$) and (SECV) of 0.85 (19.8) and 0.85 (19.6) for 1,4,4,1 and 2,5,5,2 respectively, on a dry weight. It was concluded that NIRS is a suitable method to predict moisture and nitrogen on fresh forage without samples preparation.

• Proceedings of the Korean Society of Near Infrared Spectroscopy Conference
• /
• 2001.06a
• /
• pp.1266-1266
• /
• 2001

Pasture composition, an important attribute determining sward condition and value, is normally assessed by hand separation, drying and measuring weight contribution of each species in the mixture. This is a tedious, time and labour consuming procedure. NIRS has demonstrated the potential for predicting botanical composition of swards, but most of the work has been carried out on dry samples. The aim of this work was to evaluate the feasibility of developing NIR models for predicting the white clover and ryegrass content in fresh or dry mixtures artificially prepared from pure samples of both species. Mixtures from pure stands of white clover(Trifolium repens) and perennial ryegrass (Lolium perenne) were prepared with different proportions (0 to 100%) of each species (fresh weight). A total of 55 samples were made (11 mixtures,5 cuts). Spectra (400 to 2500 nm) were taken from fresh chopped (rectangular cuvettes, transport sample module) samples, in a NIR Systems 6500 scanning monochromator controlled by the software NIRS 3 (Infrasoft International), which was also utilized for calibration development. Different math treatments (derivative order, subtraction gap and smooth segment) and a scatter correction treatment of the spectra (SNV and Detrend) were tested. Equations were developed by modified partial least squares. Prediction accuracy evaluated by cross-validation, showed that percentage of clover or ryegrass, as contribution in dry weight, can be successfully percentage of clover or ryegrass, as contribution in dry weight, can be successfully predicted either on fresh or dried samples, with equations developed by different math treatments. Best equations for fresh samples were developed including a first, second, or third derivative, whereas for dry samples best equations included a second or third derivative. Standard errors of ross validation were about 6% for fresh and 3.6% for dry samples, Coefficient of determination of cross validation (1-VR) were over 0.95 times the value of SECV for fresh samples and over 8 times the value of SECV for dry samples. Scatter correction (SNV and Detrend) in general improved prediction accuracy. It is concluded more precise on dried and ground samples, it can be used with an acceptable error level and less time and labour, on fresh samples.

• Proceedings of the Korean Society of Near Infrared Spectroscopy Conference
• /
• 2001.06a
• /
• pp.1244-1244
• /
• 2001

In linear discriminant analysis there are two important properties concerning the effectiveness of discriminant function modeling. The first is the separability of the discriminant function for different classes. The separability reaches its optimum by maximizing the ratio of between-class to within-class variance. The second is the stability of the discriminant function against noises present in the measurement variables. One can optimize the stability by exploring the discriminant variates in a principal variation subspace, i. e., the directions that account for a majority of the total variation of the data. An unstable discriminant function will exhibit inflated variance in the prediction of future unclassified objects, exposed to a significantly increased risk of erroneous prediction. Therefore, an ideal discriminant function should not only separate different classes with a minimum misclassification rate for the training set, but also possess a good stability such that the prediction variance for unclassified objects can be as small as possible. In other words, an optimal classifier should find a balance between the separability and the stability. This is of special significance for multivariate spectroscopy-based classification where multicollinearity always leads to discriminant directions located in low-spread subspaces. A new regularized discriminant analysis technique, the principal discriminant variate (PDV) method, has been developed for handling effectively multicollinear data commonly encountered in multivariate spectroscopy-based classification. The motivation behind this method is to seek a sequence of discriminant directions that not only optimize the separability between different classes, but also account for a maximized variation present in the data. Three different formulations for the PDV methods are suggested, and an effective computing procedure is proposed for a PDV method. Near-infrared (NIR) spectra of blood plasma samples from mastitic and healthy cows have been used to evaluate the behavior of the PDV method in comparison with principal component analysis (PCA), discriminant partial least squares (DPLS), soft independent modeling of class analogies (SIMCA) and Fisher linear discriminant analysis (FLDA). Results obtained demonstrate that the PDV method exhibits improved stability in prediction without significant loss of separability. The NIR spectra of blood plasma samples from mastitic and healthy cows are clearly discriminated between by the PDV method. Moreover, the proposed method provides superior performance to PCA, DPLS, SIMCA and FLDA, indicating that PDV is a promising tool in discriminant analysis of spectra-characterized samples with only small compositional difference, thereby providing a useful means for spectroscopy-based clinic applications.

• Proceedings of the Korean Society of Near Infrared Spectroscopy Conference
• /
• 2001.06a
• /
• pp.1527-1527
• /
• 2001

Glucose, fructose and sucrose being the main sugars that can be found in natural fruit juice. Many instrumental methods, such as GC, LC, electrochemical or spectrometric methods provide information about both the total content of sugars and the specific concentration of each carbohydrate[1]. The simplicity of sample handling and measurement in the near IR(NIR) wavelength region, which allows the use of long pathlength, optical glass cells and optical fibers, makes NIR a good alternative for sugar determination [2]. In the present study, six varieties of persian grapes were harvested at intervals through august to october and analysed for sugars by NIR. The results were processed by principal component regression (PCR) and partial least squares (PLS) analysis. Sample juice was prepared by squeezing through gauze from crashed grape. This solution was treated by zinc ferrocyanide prior to analysis in order to eliminate colored compounds and all optically active nonsugar substances. For glucose and fructose the most characteristic wavelengths were 1456nm corresponding to the first harmonic O-H stretching and the second at 2062nm corresponding to O-H stretching and deformation; secondary characteristic combination bands were also seen at 2265 nm (O-H and C-C stretching) and at 2240 nm (C-H and C-C stretching). However these spectra were taken over a wavelength range from 1100-2500nm at room temperature of 25-$30^{\circ}C$. To test the accuracy of the described procedure, samples of six varieties of grape were analysed by the proposed NIR and a standard method[2]. Good agreement were found between these two sets of the results. To perform the recovery studies , samples of grape juices previously analysed by the proposed method, were spiked with known amounts of each individual sugars and then analysed again. Relative standard deviations varied from 1.4 to 1.8% for six independent measurements of individual and total sugar concentration. In the analysis of real and synthetic samples, precise and accurate results were obtained , providing accuracy errors lower than 1.9% in all cases. Average recoveries of ${97}{\pm}{4%}$ for total sugar and between ${95}{\pm}{5%}$ and ${99}{\pm}{2%}$ for sing1e sugars demonstrate the applicability of the methodology developed to the direct analysis of grape Juice.

• Proceedings of the Korean Society of Near Infrared Spectroscopy Conference
• /
• 2001.06a
• /
• pp.1042-1042
• /
• 2001

In linear discriminant analysis there are two important properties concerning the effectiveness of discriminant function modeling. The first is the separability of the discriminant function for different classes. The separability reaches its optimum by maximizing the ratio of between-class to within-class variance. The second is the stability of the discriminant function against noises present in the measurement variables. One can optimize the stability by exploring the discriminant variates in a principal variation subspace, i. e., the directions that account for a majority of the total variation of the data. An unstable discriminant function will exhibit inflated variance in the prediction of future unclassified objects, exposed to a significantly increased risk of erroneous prediction. Therefore, an ideal discriminant function should not only separate different classes with a minimum misclassification rate for the training set, but also possess a good stability such that the prediction variance for unclassified objects can be as small as possible. In other words, an optimal classifier should find a balance between the separability and the stability. This is of special significance for multivariate spectroscopy-based classification where multicollinearity always leads to discriminant directions located in low-spread subspaces. A new regularized discriminant analysis technique, the principal discriminant variate (PDV) method, has been developed for handling effectively multicollinear data commonly encountered in multivariate spectroscopy-based classification. The motivation behind this method is to seek a sequence of discriminant directions that not only optimize the separability between different classes, but also account for a maximized variation present in the data. Three different formulations for the PDV methods are suggested, and an effective computing procedure is proposed for a PDV method. Near-infrared (NIR) spectra of blood plasma samples from daily monitoring of two Japanese cows have been used to evaluate the behavior of the PDV method in comparison with principal component analysis (PCA), discriminant partial least squares (DPLS), soft independent modeling of class analogies (SIMCA) and Fisher linear discriminant analysis (FLDA). Results obtained demonstrate that the PDV method exhibits improved stability in prediction without significant loss of separability. The NIR spectra of blood plasma samples from two cows are clearly discriminated between by the PDV method. Moreover, the proposed method provides superior performance to PCA, DPLS, SIMCA md FLDA, indicating that PDV is a promising tool in discriminant analysis of spectra-characterized samples with only small compositional difference.

• Proceedings of the Korean Society of Near Infrared Spectroscopy Conference
• /
• 2001.06a
• /
• pp.1141-1141
• /
• 2001

The aim of the study was to evaluate the performance of 3 calibration methods, modified partial least squares (MPLS), local PLS (LOCAL) and artificial neural network (ANN) on the prediction of chemical composition of forages, using a large NIR database. The study used forage samples (n=25,977) from Australia, Europe (Belgium, Germany, Italy and Sweden) and North America (Canada and U.S.A) with information relative to moisture, crude protein and neutral detergent fibre content. The spectra of the samples were collected with 10 different Foss NIR Systems instruments, which were either standardized or not standardized to one master instrument. The spectra were trimmed to a wavelength range between 1100 and 2498 nm. Two data sets, one standardized (IVAL) and the other not standardized (SVAL) were used as independent validation sets, but 10% of both sets were omitted and kept for later expansion of the calibration database. The remaining samples were combined into one database (n=21,696), which was split into 75% calibration (CALBASE) and 25% validation (VALBASE). The chemical components in the 3 validation data sets were predicted with each model derived from CALBASE using the calibration database before and after it was expanded with 10% of the samples from IVAL and SVAL data sets. Calibration performance was evaluated using standard error of prediction corrected for bias (SEP(C)), bias, slope and R2. None of the models appeared to be consistently better across all validation sets. VALBASE was predicted well by all models, with smaller SEP(C) and bias values than for IVAL and SVAL. This was not surprising as VALBASE was selected from the calibration database and it had a sample population similar to CALBASE, whereas IVAL and SVAL were completely independent validation sets. In most cases, Local and ANN models, but not modified PLS, showed considerable improvement in the prediction of IVAL and SVAL after the calibration database had been expanded with the 10% samples of IVAL and SVAL reserved for calibration expansion. The effects of sample processing, instrument standardization and differences in reference procedure were partially confounded in the validation sets, so it was not possible to determine which factors were most important. Further work on the development of large databases must address the problems of standardization of instruments, harmonization and standardization of laboratory procedures and even more importantly, the definition of the database population.

• Proceedings of the Korean Society of Near Infrared Spectroscopy Conference
• /
• 2001.06a
• /
• pp.1032-1032
• /
• 2001

Artificial Neural Network (ANN) calibration techniques have been used commercially for agricultural applications since the mid-nineties. Global models, based on transmission data from 850 to 1050 nm, are used routinely to measure protein and moisture in wheat and barley and also moisture in triticale, rye, and oats. These models are currently used commercially in approx. 15 countries throughout the world. Results concerning earlier European ANN models are being published elsewhere. Some of the findings from that study will be discussed here. ANN models have also been developed for coarsely ground samples of compound feed and feed ingredients, again measured in transmission mode from 850 to 1050 nm. The performance of models for pig- and poultry feed will be discussed briefly. These models were developed from a very large data set (more than 20,000 records), and cover a very broad range of finished products. The prediction curves are linear over the entire range for protein, fat moisture, fibre, and starch (measured only on poultry feed), and accuracy is in line with the performance of smaller models based on Partial Least Squares (PLS). A simple bias adjustment is sufficient for calibration transfer across instruments. Recently, we have investigated the possible use of ANN for a different type of NIR spectrometer, based on reflectance data from 1100 to 2500 nm. In one study, based on data for protein, fat, and moisture measured on unground compound feed samples, dedicated ANN models for specific product classes (cattle feed, pig feed, broiler feed, and layers feed) gave moderately better Standard Errors of Prediction (SEP) compared to modified PLS (MPLS). However, if the four product classes were combined into one general calibration model, the performance of the ANN model deteriorated only slightly compared to the class-specific models, while the SEP values for the MPLS predictions doubled. Brix value in molasses is a measure of sugar content. Even with a huge dataset, PLS models were not sufficiently accurate for commercial use. In contrast an ANN model based on the same data improved the accuracy considerably and straightened out non-linearity in the prediction plot. The work of Mr. David Funk (GIPSA, U. S. Department of Agriculture) who has studied the influence of various types of spectral distortions on ANN- and PLS models, thereby providing comparative information on the robustness of these models towards instrument differences, will be discussed. This study was based on data from different classes of North American wheat measured in transmission from 850 to 1050 nm. The distortions studied included the effect of absorbance offset pathlength variation, presence of stray light bandwidth, and wavelength stretch and offset (either individually or combined). It was shown that a global ANN model was much less sensitive to most perturbations than class-specific GIPSA PLS calibrations. It is concluded that ANN models based on large data sets offer substantial advantages over PLS models with respect to accuracy, range of materials that can be handled by a single calibration, stability, transferability, and sensitivity to perturbations.

• Proceedings of the Korean Society of Near Infrared Spectroscopy Conference
• /
• 2001.06a
• /
• pp.1121-1121
• /
• 2001

A previous study (Berzaghi et al., 2001) evaluated the performance of 3 calibration methods, modified partial least squares (MPLS), local PLS (LOCAL) and artificial neural networks (ANN) on the prediction of the chemical composition of forages, using a large NIR database. The study used forage samples (n=25,977) from Australia, Europe (Belgium, Germany, Italy and Sweden) and North America (Canada and U.S.A) with reference values for moisture, crude protein and neutral detergent fibre content. The spectra of the samples were collected using 10 different Foss NIR Systems instruments, only some of which had been standardized to one master instrument. The aim of the present study was to evaluate the behaviour of these different calibration methods when predicting the same samples measured on different instruments. Twenty-two sealed samples of different kind of forages were measured in duplicate on seven instruments (one master and six slaves). Three sets of near infrared spectra (1100 to 2500nm) were created. The first set consisted of the spectra in their original form (unstandardized); the second set was created using a single sample standardization (Clone1); the third was created using a multiple sample procedure (Clone6). WinISI software (Infrasoft International Inc., Port Mathilda, PA, USA) was used to perform both types of standardization, Clone1 is just a photometric offset between a “master” instrument and the “slave” instrument. Clone6 modifies both the X-axis through a wavelength adjustment and the Y-axis through a simple regression wavelength by wavelength. The Clone1 procedure used one sample spectrally close to the centre of the population. The six samples used in Clone 6 were selected to cover the range of spectral variation in the sample set. The remaining fifteen samples were used to evaluate the performances of the different models. The predicted values for dry matter, protein and neutral detergent fibre from the master Instrument were considered as “reference Y values” when computing the statistics RMSEP, SEPC, R, Bias, Slope, mean GH (global Mahalanobis distance) and mean NH (neighbourhood Mahalanobis distance) for the 6 slave instruments. From the results we conclude that i) all the calibration techniques gave satisfactory results after standardization. Without standardization the predicted data from the slaves would have required slope and bias correction to produce acceptable statistics. ii) Standardization reduced the errors for all calibration methods and parameters tested, reducing not only systematic biases but also random errors. iii) Standardization removed slope effects that were significantly different from 1.0 in most of the cases. iv) Clone1 and Clone6 gave similar results except for NDF where Clone6 gave better RMSEP values than Clone1. v) GH and NH were reduced by half even with very large data sets including unstandardized spectra.

• Proceedings of the Korean Society of Near Infrared Spectroscopy Conference
• /
• 2001.06a
• /
• pp.1510-1510
• /
• 2001

NIRS uses reflectance signals resulting from bending and stretching vibrations in chemical bonds between carbon, nitrogen, hydrogen, sulfur and oxygen. These reflectance signals are used to measure the concentration of major chemical composition and other descriptors of homogenized and freeze-dried whole broiler carcasses. Six strains of chicken were analyzed and the NIRS model predictions compared to reference data. The results of this comparison indicate that NIRS is a rapid tool for predicting dry matter (DM), fat, crude protein (CP) and ash content in the broiler carcass. Males and females of six commercial strain crosses of broiler chicken (Gallus domesticus) were used in this study (6$\times$2 factorial design). Each strain was grown to 16 weeks of age, and duplicate serial samples were taken for body composition analysis. Each whole carcass was pressure-cooked, homogenized, and a representative sample was freeze-dried. Body composition determined as follows: DM by oven dried method at 105$^{\circ}C$ for 3 hours, fat by Mojonnier diethyl ether extraction, CP by measuring nitrogen content using an auto-analyzer with Kjeldhal digest and ash by combustion in a muffle furnace for 24 hour at 55$0^{\circ}C$. These homogenized and freeze-dried carcass samples were then scanned with a Foss NIR Systems 6500 visible-NIR spectrophotometer (400-2500nm) (Foss NIR Systems, Silver Spring, MD., US) using Infra-Soft-International, ISI, WinISl software (ISI, Port Matilda, US). The NIRS spectra were analyzed using principal component (PC) analysis. This data was corrected for scatter using standard normal “Variate” and “Detrend” technique. The accuracy of the NIRS calibration equations developed using Partial Least Squares (PLS) for predicting major chemical composition and carcass descriptors- such as body mass (BM), bird dry matter and moisture content was tested using cross validation. Discrimination analysis was also used for sex and strain identification. According to Dr John Shenk, the creator of the ISI software, the calibration equations with the correlation coefficient, $R^2$, between reference data and NIRS predicted results of above 0.90 is excellent and between 0.70 to 0.89 is a good quantifying guideline. The excellent calibration equations for DM ($R^2$= 0.99), fat (0.98) and CP (0.92) and a good quantifying guideline equation for ash (0.80) were developed in this study. The results of cross validation statistics for carcass descriptors, body composition using reference methods, inter-correlation between carcass descriptors and NIRS calibration, and the results of discrimination analysis for sex and strain identification will also be presented in the poster. The NIRS predicted daily gain and calculated daily gain from this experiment, and true daily gain (using data from another experiment with closely related broiler chicken from each of the six strains) will also be discussed in the paper.

• Journal of Biosystems Engineering
• /
• v.43 no.2
• /
• pp.138-147
• /
• 2018

Purpose: A narrowband hyperspectral imaging sensor of high-dimensional spectral bands is advantageous for identifying the reflectance by selecting the significant spectral bands for predicting crop yield over the broadband multispectral imaging sensor for each wavelength range of the crop canopy. The images acquired by each imaging sensor were used to develop the models for predicting the Chinese cabbage yield. Methods: The models for predicting the Chinese cabbage (Brassica campestris L.) yield, with multispectral images based on unmanned aerial vehicle (UAV), were developed by simple linear regression (SLR) using vegetation indices, and forward stepwise multiple linear regression (MLR) using four spectral bands. The model with hyperspectral images based on the ground were developed using forward stepwise MLR from the significant spectral bands selected by dimension reduction methods based on a partial least squares regression (PLSR) model of high precision and accuracy. Results: The SLR model by the multispectral image cannot predict the yield well because of its low sensitivity in high fresh weight. Despite improved sensitivity in high fresh weight of the MLR model, its precision and accuracy was unsuitable for predicting the yield as its $R^2$ is 0.697, root-mean-square error (RMSE) is 1170 g/plant, relative error (RE) is 67.1%. When selecting the significant spectral bands for predicting the yield using hyperspectral images, the MLR model using four spectral bands show high precision and accuracy, with 0.891 for $R^2$, 616 g/plant for the RMSE, and 35.3% for the RE. Conclusions: Little difference was observed in the precision and accuracy of the PLSR model of 0.896 for $R^2$, 576.7 g/plant for the RMSE, and 33.1% for the RE, compared with the MLR model. If the multispectral imaging sensor composed of the significant spectral bands is produced, the crop yield of a wide area can be predicted using a UAV.