Optimal Partial Least Squares Regression Model Research Articles

Quantification and Detection of Ground Garlic Adulteration Using Fourier-Transform Near-Infrared Reflectance Spectra.

This study demonstrates the rapid and cost-effective possibility of quantifying adulterant amounts (corn flour or corn starch) in ground and dried garlic samples. Prepared mixtures with different concentrations of selected adulterant were effectively characterized using Fourier-transform near-infrared reflectance spectra (FT-NIR), and multivariate calibration models were developed using two methods: principal component regression (PCR) and partial least squares regression (PLSR). They were constructed for optimally preprocessed FT-NIR spectra, and PLSR models generally performed better regarding model fit and predictions than PCR. The optimal PLSR model, built to estimate the amount of corn flour present in the ground and dried garlic samples, was constructed for the first derivative spectra obtained after Savitzky-Golay smoothing (fifteen sampling points and polynomial of the second degree). It demonstrated root mean squared errors for calibration and validation samples equal to 1.8841 and 1.8844 (i.e., 1.88% concerning the calibration range), respectively, and coefficients of determination equal to 0.9955 and 0.9858. The optimal PLSR model constructed for spectra after inverse scattering correction to assess the amount of corn starch had root mean squared errors for calibration and validation samples equal to 1.7679 and 1.7812 (i.e., 1.77% and 1.78% concerning the calibration range), respectively, and coefficients of determination equal to 0.9961 and 0.9873. It was also possible to discriminate samples adulterated with corn flour or corn starch using partial least squares discriminant analysis (PLS-DA). The optimal PLS-DA model had a very high correct classification rate (99.66%), sensitivity (99.96%), and specificity (99.36%), calculated for external validation samples. Uncertainties of these figures of merit, estimated using the Monte Carlo validation approach, were relatively small. One-class classification partial least squares models, developed to detect the adulterant type, presented very optimistic sensitivity for validation samples (above 99%) but low specificity (64% and 45.33% for models recognizing corn flour or corn starch adulterants, respectively). Through experimental investigation, chemometric data analysis, and modeling, we have verified that the FT-NIR technique exhibits the required sensitivity to quantify adulteration in dried ground garlic, whether it involves corn flour or corn starch.

Development of Miconazole-Loaded Microemulsions for Enhanced Topical Delivery and Non-Destructive Analysis by Near-Infrared Spectroscopy.

The antifungal drug miconazole nitrate has a low solubility in water, leading to reduced therapeutic efficacy. To address this limitation, miconazole-loaded microemulsions were developed and assessed for topical skin delivery, prepared through spontaneous emulsification with oleic acid and water. The surfactant phase included a mixture of polyoxyethylene sorbitan monooleate (PSM) and various cosurfactants (ethanol, 2-(2-ethoxyethoxy) ethanol, or 2-propanol). The optimal miconazole-loaded microemulsion containing PSM and ethanol at a ratio of 1:1 showed a mean cumulative drug permeation of 87.6 ± 5.8 μg/cm2 across pig skin. The formulation exhibited higher cumulative permeation, permeation flux, and drug deposition than conventional cream and significantly increased the in vitro inhibition of Candida albicans compared with cream (p < 0.05). Over the course of a 3-month study conducted at a temperature of 30 ± 2 °C, the microemulsion exhibited favorable physicochemical stability. This outcome signifies its potential suitability as a carrier for effectively administering miconazole through topical administration. Additionally, a non-destructive technique employing near-infrared spectroscopy coupled with a partial least-squares regression (PLSR) model was developed to quantitatively analyze microemulsions containing miconazole nitrate. This approach eliminates the need for sample preparation. The optimal PLSR model was derived by utilizing orthogonal signal correction pretreated data with one latent factor. This model exhibited a remarkable R2 value of 0.9919 and a root mean square error of calibration of 0.0488. Consequently, this methodology holds potential for effectively monitoring the quantity of miconazole nitrate in various formulations, including both conventional and innovative ones.

Quantitative analysis of amorphous form in indomethacin binary polymorphic mixtures using infrared spectroscopy analytical techniques combined with chemometrics methods

Pharmaceutical corporations prefer amorphous active pharmaceutical ingredients (APIs) because of their high apparent solubility and bioavailability, which are caused by long-range disorder, large unit surface free energy, and easily wetted particle surfaces. APIs in amorphous form tend to recrystallize into crystalline form during production, transportation, and storage, leading to differences in safety and clinical efficacy, which has been a major obstacle to development of amorphous drugs. In this study, rapid and nondestructive methods for quantitative analysis of amorphous indomethacin (A-INDO) content in A-INDO and γ-indomethacin (γ-INDO) binary mixtures were established based on Attenuated Total Reflection Fourier Transform Infrared spectroscopy (ATR-FTIR) and Near-Infrared spectroscopy (NIR) combined with chemometrics methods. Partial least squares regression (PLSR) was used to establish quantitative analysis models of A-INDO content ranging from 0.000 % to 10.000 % w/w %. A variety of spectral pretreatment methods were used to pretreat the spectral, reducing the influence of inconsistent particle size and uneven mixing, and highlighting the sample component information. The two analytical techniques’ best PLSR models were selected, and the performance of the best models was confirmed by the validation samples. The most effective PLSR model based on ATR-FTIR was Y = 0.99456 X + 0.21526, R2 = 0.99456, the limit of detection (LOD) was 0.714 % and the limit of quantification (LOQ) was 2.164 %. The optimal PLSR model based on NIR was Y = 0.99950 X + 0.07465, R2 = 0.99950, with LOD = 0.246 % and LOQ = 0.747 %. The results showed that the PLSR models established by ATR-FTIR and NIR solid-state analytical techniques combined with stoichiometric method could be used to detect the content of amorphous APIs and to control the production quality of APIs.

Estimation of Anthocyanins in Leaves of Trees with Apple Mosaic Disease Based on Hyperspectral Data

Anthocyanins are severity indicators for apple mosaic disease and can be used to monitor tree health. However, most of the current studies have focused on healthy leaves, and few studies have estimated the anthocyanin content in diseased leaves. In this study, we obtained the hyperspectral data of apple leaves with mosaic disease, analyzed the spectral characteristics of leaves with different degrees of Mosaic disease, constructed and screened the spectral index sensitive to anthocyanin content, and improved the estimation model. To improve the conciseness of the model, we integrated Variable Importance in Projection (VIP), Partial Least Squares Regression (PLSR), and Akaike Information Criterion (AIC) to select the optimal PLSR model and its independent variables. Sparrow Search Algorithm-Random Forest (SSA-RF) was used to improve accuracy. Results showed the following: (1) anthocyanin content increased gradually with the aggravation of disease. The reflectance of the blade spectrum in the visible band increased, the red edge moved to short wave, and the phenomenon of “blue shift of spectrum” occurred. (2) The VIP-PLSR-AIC selected 17 independent variables from 21 spectral indices. (3) Variables were used to construct PLSR, Back Propagation (BP), Support Vector Machine (SVM), Random Forest (RF), and SSA-RF to estimate anthocyanin content. Results showed the estimation accuracy and stability of the SSA-RF model were better than other models. The model set determination coefficient (R2) was up to 0.955, which is 0.047 higher than that of the RF model and 0.138 higher than that of the SVM model with the lowest accuracy. The model was constructed at the leaf scale and can provide a reference for other scale studies, including a theoretical basis for large-area, high-efficiency, high-precision anthocyanin estimation and monitoring of apple mosaics using remote sensing technology.

An efficient methodology for modeling to predict wine aroma expression based on quantitative data of volatile compounds: A case study of oak barrel-aged red wines

Correlating aroma expression with volatile compounds has long been an ambition in researches of flavor chemistry. To propose a reliable methodology to depict wine aroma, 76 oak barrel-aged dry red wines were investigated through the combination of machine learning algorithm and multivariate analysis. Aromatic characteristic was evaluated by quantitative descriptive analysis (QDA), while non– or oak derived volatiles were detected by HS-SPME-GC–MS and targeted SPE-GC-QqQ-MS/MS, respectively. Results showed that variable importance for projection values (VIPs) from partial least-squares regression (PLSR) and mean decrease accuracy (MDA) from random forest were efficient parameters for feature selection. The correlating accuracy of the optimal PLSR model to predict intensities of different aroma characteristics through selected volatile compounds could achieve 0.754 to 0.943, representing potential application to manage wine aroma by chemical assay in winemaking. From the perspective of mathematical modeling in the real wine matrix, the network analysis between aroma characteristics and key volatile compounds indicated that the expression of oak aroma was not only directly contributed by volatiles derived from oak wood, but also influenced by ethyl esters, including ethyl acetate, ethyl butanoate, ethyl hexanoate, ethyl decanoate, and ethyl nonanoate.

Prediction of oil content in single maize kernel based on hyperspectral imaging and attention convolution neural network

An attention (A) based convolutional neural network regression (CNNR) model, namely ACNNR, was proposed to combine hyperspectral imaging to predict oil content in single maize kernel. During the period, a reflectance HSI system was used to collect hyperspectral images of embryo side and non-embryo side of single maize kernel, and the performances of CNNR (without attention mechanism), ACNNR and partial least squares regression (PLSR) were compared. For PLSR, a series of spectral preprocessing and dimensionality reduction methods were used to finally determine the optimal hybrid PLSR model. Whereas for CNNR and ACNNR, only raw spectra were used as their inputs. The results showed that embryo side was more suitable for developing regression models; the attentional mechanism was helpful to reduce the error of prediction, making ACNNR performed best (coefficient of determination of prediction = 0.9198). Overall, the proposed method did not require additional processing on raw spectra, and performed well.

Rapid Geographical Origin Identification and Quality Assessment of Angelicae Sinensis Radix by FT-NIR Spectroscopy.

Angelicae Sinensis Radix is a widely used traditional Chinese medicine and spice in China. The purpose of this study was to develop a methodology for geographical classification of Angelicae Sinensis Radix and determine the contents of ferulic acid and Z-ligustilide in the samples using near-infrared spectroscopy. A qualitative model was established to identify the geographical origin of Angelicae Sinensis Radix using Fourier transform near-infrared (FT-NIR) spectroscopy. Support vector machine (SVM) algorithms were used for the establishment of a qualitative model. The optimum SVM model had a recognition rate of 100% for the calibration set and 83.72% for the prediction set. In addition, a quantitative model was established to predict the content of ferulic acid and Z-ligustilide using FT-NIR. Partial least squares regression (PLSR) algorithms were used for the establishment of a quantitative model. Synergy interval-PLS (Si-PLS) was used to screen the characteristic spectral interval to obtain the best PLSR model. The coefficient of determination for calibration (R2C) for the best PLSR models established with the optimal spectral preprocessing method and selected important spectral regions for the quantitative determination of ferulic acid and Z-ligustilide was 0.9659 and 0.9611, respectively, while the coefficient of determination for prediction (R2P) was 0.9118 and 0.9206, respectively. The values of the ratio of prediction to deviation (RPD) of the two final optimized PLSR models were greater than 2. The results suggested that NIR spectroscopy combined with SVM and PLSR algorithms could be exploited in the discrimination of Angelicae Sinensis Radix from different geographical locations for quality assurance and monitoring. This study might serve as a reference for quality evaluation of agricultural, pharmaceutical, and food products.

Possibility of Zhuhai-1 Hyperspectral Imagery for Monitoring Salinized Soil Moisture Content Using Fractional Order Differentially Optimized Spectral Indices

The possibility of quantitative inversion of salinized soil moisture content (SMC) from Zhuhai-1 hyperspectral imagery and the application effect of fractional order differentially optimized spectral indices were discussed, which provided new research ideas for improving the accuracy of hyperspectral remote sensing inversion. The hyperspectral data from indoor and Zhuhai-1 remote sensing imagery were resampled to the same spectral scale. The soil hyperspectral data were processed by fractional order differential preprocessing method and optimized spectral indices method, and the Pearson correlation coefficient (PCC/r) analysis was made with SMC data. The sensitive optimized spectral indices were used to establish the ground hyperspectral estimation model, and a variety of modeling methods were used to select the best SMC inversion model. The results were as follows: the maximum one-dimensional r between SMC and the 466–938 nm band was −0.635, the maximum one-dimensional r with the 0.5-order absorbance spectrum was 0.665, and the maximum two-dimensional r with the difference index (DI) calculated by the 0.5-order absorbance spectrum was ±0.72. The maximum three-dimensional r with the triangle vegetation index (TVI) calculated from the 0.5-order absorbance spectrum reached 0.755, which exceeded the one-dimensional r extreme value of 400–2400 nm. The TreeNet gradient boosting machine (TGBM) regression model had the highest modeling accuracy, with a calibration coefficient of determination (R2C) = 0.887, calibration root mean square error (RMSEC) = 2.488%, standard deviation (SD) = 6.733%, and r = 0.942. However, the partial least squares regression (PLSR) model had the strongest predictive ability, with validation coefficient of determination (R2V) = 0.787, validation root mean square error (RMSEV) = 3.247%, and relative prediction deviation (RPD) = 2.071. The variable importance in projection (VIP) method could not only improve model efficiency but also increased model accuracy. R2C of the optimal PLSR model was 0.733, RMSEC was 3.028%, R2V was 0.805, RMSEV was 3.100%, RPD was 1.976, and Akaike information criterion (AIC) was 151.050. The three-band optimized spectral indices with fractional differential pretreatment could to a certain extent break through the limitation of visible near-infrared spectrum in SMC estimation due to the lack of shortwave infrared spectra, which made it possible to quantitatively retrieve saline SMC on the basis of Zhuhai-1 hyperspectral imagery.

Assessment of the human albumin in acid precipitation process using NIRS and multi-variable selection methods combined with SPA

In this study, human albumin (HA) content was determined by near infrared spectroscopy (NIRS), which was an process analytical technology (PAT) tool, to realize process monitoring during acid precipitation. Eight batches of acid precipitation process samples were simulated in laboratory, and HA contents were determined by bromocresol green (BCG) method. Five batches were selected as calibration set and three batches were selected as validation set to build partial least squares regression (PLSR) model of HA content. Before the establishment of PLSR model, partial samples and different variable selection methods were compared in details and 35 variables were finally selected. The results showed that the successive projection algorithm (SPA) combined with the multi-variable selection methods (CC, UVE, CARS, SCARS, FiPLS, MWPLS, BiPLS) could obtain more useful information than the single method. The RMSECV and RMSEP values of the optimal PLSR model were 0.7038 g/L and 0.5893 g/L, respectively. Meanwhile, the results indicated that NIRS was an effective tool and could be successfully used for quantitative monitoring during acid precipitation process. To some extent, it could provide theoretical supports for quality control with a fast, nondestructive and low-cost NIRS in monitoring the complex biological production process.

Effect of long‐term application of manure and nitrogen fertilizer on infiltration for a wheat–maize rotation system

AbstractModelling soil infiltration capacity and identifying controlling factors under long‐term fertilizer application could help in developing better fertilization treatments to increase infiltration, thereby preventing land degradation. This study was performed at Laiyang long‐term fertilization experimental field in Shandong Province, China. Three application rates of manure (0, 30,000, and 60,000 kg ha−2) and nitrogen (0, 138, and 276 kg ha−2) were full designed in a nine‐treatment randomized block design with three replicates per treatment. The infiltration processes were detected using an automatic infiltration measurement system and represented by three indices, that is, initial infiltration rate for the first 3 min (Ii), the average infiltration rate for the whole 30‐min period (Ia), and the final infiltration rate for the last 3 min (If). Manure application could increase Ia and If significantly (p &lt; 0.01). Nitrogen application had no significant effect on infiltration. The linkage between potential influencing factors and infiltration capacity indices was analysed by partial least squares regression (PLSR), an efficient method that can deal with high colinear variables. The optimal PLSR model could explain 95% of the variability in potential influencing factors and 92% of the variability in soil infiltration capacity indices and could predict 82% of the variability in soil infiltration capacity. Three soil properties, that is, organic matter content, bulk density, and weight ratio of soil aggregate with diameter greater than 0.25 mm measured using wet sieving method, were identified as main influencing factors. The results would be informative to assess infiltration and then land degradation after a long‐term cultivation.

Discrimination and prediction of cultivation age and parts of Panax ginseng by Fourier-transform infrared spectroscopy combined with multivariate statistical analysis.

Panax ginseng C.A. Meyer is a herb used for medicinal purposes, and its discrimination according to cultivation age has been an important and practical issue. This study employed Fourier-transform infrared (FT-IR) spectroscopy with multivariate statistical analysis to obtain a prediction model for discriminating cultivation ages (5 and 6 years) and three different parts (rhizome, tap root, and lateral root) of P. ginseng. The optimal partial-least-squares regression (PLSR) models for discriminating ginseng samples were determined by selecting normalization methods, number of partial-least-squares (PLS) components, and variable influence on projection (VIP) cutoff values. The best prediction model for discriminating 5- and 6-year-old ginseng was developed using tap root, vector normalization applied after the second differentiation, one PLS component, and a VIP cutoff of 1.0 (based on the lowest root-mean-square error of prediction value). In addition, for discriminating among the three parts of P. ginseng, optimized PLSR models were established using data sets obtained from vector normalization, two PLS components, and VIP cutoff values of 1.5 (for 5-year-old ginseng) and 1.3 (for 6-year-old ginseng). To our knowledge, this is the first study to provide a novel strategy for rapidly discriminating the cultivation ages and parts of P. ginseng using FT-IR by selected normalization methods, number of PLS components, and VIP cutoff values.

Characterization of moisture content in dehydrated scallops using spectral images

Herein, a hyperspectral imaging system in the 380–1030 nm range was used to rapidly determine the moisture content of scallops in different dehydration periods. Mean spectral values of scallops were extracted from hyperspectral images. Only eight optimal wavelengths were selected using the regression coefficient method. Spectra of full wavebands and selected wavelengths were used as independent variables for modeling. Partial least squares regression (PLSR) and least-squares support vector machines (LSSVM) were employed to establish multispectral calibration models to correlate spectral features with moisture content. The best results, with correlation coefficients of prediction (RP), root mean square error of prediction (RMSEP), and residual predictive deviation (RPD) of 0.9673, 3.5584%, and 3.7150, respectively, were achieved using the optimal wavelength-based PLSR model. To visualize moisture content in scallops, a visualization map was generated using the selected wavelength-based PLSR model. These results highlight the potential of hyperspectral imaging for non-destructive prediction of moisture content in scallops.

Utilizing a PLSR-Based Band-Selection Procedure for Spectral Feature Characterization of Floristic Gradients

The study introduces a new approach for the characterization of floristic gradients by hyperspectral features in a partial least squares regression (PLSR) framework. As ecological factors influence the composition of vegetation, our study is aimed to reveal related effects on spectral signatures. For this purpose, the variation of plant species in an open dryland area was projected into a three-dimensional ordination space using nonmetric multidimensional scaling (NMDS). Subsequently, ordination axes score rotations were performed in $\text{180}^\circ$ semicircles and the waveband-specific correlation to spectral field measurements of reflectance, continuum removed, and first-derivative spectra were extracted. A bootstrapped PLSR modeling was applied over the entire rotation space using varying numbers of correlated spectral variables as input samples. On that basis, a new PLSR model suitability term was defined by isosurfaces that are spanned over ordination regions where PLSR latent vector (LV) number and PLSR $\text{R}^2$ variance is minimized. It incorporates model performance evaluation with feature characterization using weighted frequencies of spectral variable input in suitable ordination areas. Final PLSR suitability surfaces were transferred to image spectra to prove feature stability and model performance. Our investigation supports the assumption that spectral features are separable to distinct ordination space regions that can be related to individual species gradients. Thereby, the selection of an optimal PLSR model crucially depends on the spectral transformation technique. We further show that stable PLSR models can be derived in multiple ordination directions whereby an appropriate variable selection using suitability surface optimization reduces feature mismatch between field and image spectra.

Feasibility of combining spectra with texture data of multispectral imaging to predict heme and non-heme iron contents in pork sausages

To precisely determine heme and non-heme iron contents in meat product, the feasibility of combining spectral with texture features extracted from multispectral imaging data (405–970nm) was assessed. In our study, spectra and textures of 120 pork sausages (PSs) treated by different temperatures (30–80°C) were analyzed using different calibration models including partial least squares regression (PLSR) and LIB support vector machine (Lib-SVM) for predicting heme and non-heme iron contents in PSs. Based on a combination of spectral and textural features, optimized PLSR models were obtained with determination coefficient (R2) of 0.912 for heme and of 0.901 for non-heme iron prediction, which demonstrated the superiority of combining spectra with texture data. Results of satisfactory determination and visualization of heme and non-heme iron contents indicated that multispectral imaging could serve as a feasible approach for online industrial applications in the future.

Near-Infrared Spectroscopy as an Analytical Process Technology for the On-Line Quantification of Water Precipitation Processes during Danhong Injection.

This paper used near-infrared (NIR) spectroscopy for the on-line quantitative monitoring of water precipitation during Danhong injection. For these NIR measurements, two fiber optic probes designed to transmit NIR radiation through a 2 mm flow cell were used to collect spectra in real-time. Partial least squares regression (PLSR) was developed as the preferred chemometrics quantitative analysis of the critical intermediate qualities: the danshensu (DSS, (R)-3, 4-dihydroxyphenyllactic acid), protocatechuic aldehyde (PA), rosmarinic acid (RA), and salvianolic acid B (SAB) concentrations. Optimized PLSR models were successfully built and used for on-line detecting of the concentrations of DSS, PA, RA, and SAB of water precipitation during Danhong injection. Besides, the information of DSS, PA, RA, and SAB concentrations would be instantly fed back to site technical personnel for control and adjustment timely. The verification experiments determined that the predicted values agreed with the actual homologic value.

Hyperspectral Imaging Coupled with Random Frog and Calibration Models for Assessment of Total Soluble Solids in Mulberries.

Chemometrics methods coupled with hyperspectral imaging technology in visible and near infrared (Vis/NIR) region (380–1030 nm) were introduced to assess total soluble solids (TSS) in mulberries. Hyperspectral images of 310 mulberries were acquired by hyperspectral reflectance imaging system (512 bands) and their corresponding TSS contents were measured by a Brix meter. Random frog (RF) method was used to select important wavelengths from the full wavelengths. TSS values in mulberry fruits were predicted by partial least squares regression (PLSR) and least-square support vector machine (LS-SVM) models based on full wavelengths and the selected important wavelengths. The optimal PLSR model with 23 important wavelengths was employed to visualise the spatial distribution of TSS in tested samples, and TSS concentrations in mulberries were revealed through the TSS spatial distribution. The results declared that hyperspectral imaging is promising for determining the spatial distribution of TSS content in mulberry fruits, which provides a reference for detecting the internal quality of fruits.

라만 분광법과 부분최소자승법을 이용한 불량 분말식품 비파괴검사 기술 개발

This study was conducted to develop a non-destructive detection method for adulterated powder products using Raman spectroscopy and partial least squares regression(PLSR). Garlic and ginger powder, which are used as natural seasoning and in health supplement foods, were selected for this experiment. Samples were adulterated with corn starch in concentrations of 5-35%. PLSR models for adulterated garlic and ginger powders were developed and their performances evaluated using cross validation. The and SEC of an optimal PLSR model were 0.99 and 2.16 for the garlic powder samples, and 0.99 and 0.84 for the ginger samples, respectively. The variable importance in projection (VIP) score is a useful and simple tool for the evaluation of the importance of each variable in a PLSR model. After the VIP scores were taken pre-selection, the Raman spectrum data was reduced by one third. New PLSR models, based on a reduced number of wavelengths selected by the VIP scores technique, gave good predictions for the adulterated garlic and ginger powder samples.

Monitoring arsenic contamination in agricultural soils with reflectance spectroscopy of rice plants.

The objective of this study was to explore the feasibility and to investigate the mechanism for rapidly monitoring arsenic (As) contamination in agricultural soils with the reflectance spectra of rice plants. Several data pretreatment methods were applied to improve the prediction accuracy. The prediction of soil As contents was achieved by partial least-squares regression (PLSR) using laboratory and field spectra of rice plants, as well as linear regression employing normalized difference spectral index (NDSI) calculated from fild spectra. For laboratory spectra, the optimal PLSR model for predicting soil As contents was achieved using Savitzky-Golay smoothing (SG), first derivative and mean center (MC) (root-mean-square error of prediction (RMSEP)=14.7 mg kg(-1); r=0.64; residual predictive deviation (RPD)=1.31). For field spectra, the optimal PLSR model was also achieved using SG, first derivative and MC (RMSEP=13.7 mg kg(-1); r=0.71; RPD=1.43). In addition, the NDSI with 812 and 782 nm obtained a prediction accuracy with r=0.68, RMSEP=13.7 mg kg(-1), and RPD=1.36. These results indicated that it was feasible to monitor the As contamination in agricultural soils using the reflectance spectra of rice plants. The prediction mechanism might be the relationship between the As contents in soils and the chlorophyll-a/-b contents and cell structure in leaves or canopies of rice plants.

Remote Sensing Technique for Predicting Harvest Time of Tomatoes

Fast determination of growing stages and harvest time of fruits and vegetables is necessary to implement robotic operation for horticulture automation. This study evaluates the feasibility of using visible-near-infrared (Vis-NIR) spectroscopy to nondestructively determine the harvest time of tomatoes. A mobile, fibre-type, AgroSpec VIS-NIR spectrophotometer (Tec5, Germany) with a spectral range of 350-2200 nm, was used for spectral acquisition of tomatoes in reflection mode. A new index was used to measure the growing stages of tomatoes. Tomato plants were provided by Silsoe Horticultural Center, Bedfordshire, United Kingdom. Spectra were divided into a calibration set (70%) and an independent validation set (30%). Calibration set were subjected to a partial least squares regression (PLSR) with leave-one-out cross validation to establish calibration models respectively based on different spectral ranges, e.g., VIS(400-760 nm), NIR(760-2100 nm) and VIS-NIR(400-2100 nm). Prediction performance of these models on the independent validation set indicates that PLSR models based on entire spectral range (VIS-NIR) outperform those based on partial spectral ranges (VIS or NIR). Coupled with appropriate spectral transformation, the PLSR models can achieve excellent prediction performance of harvest time of tomatoes with coefficient of determination (R2) of 0.89 and RPD of 3.00. It is concluded that VIS-NIR spectroscopy combined with optimized PLSR models for GS prediction can be successfully adopted as a remote sensing technique for predicting harvest time of tomatoes, which allows for implementing autonomous fruit-picking robots.

Quantitative analysis of volatile organic compounds using ion mobility spectrometry and cascade correlation neural networks

Ion mobility spectrometry (IMS) has a limited linear range. Nonlinear calibration methods, such as neural networks are ideally suited for IMS due to their capability of modeling complex systems. Many neural networks suffer from long training times and overfitting. Cascade correlation neural networks (CCN) are interesting, because they train at fast rates. Another advantage of CCNs is that they automatically configure their own topology (number of layers and number of units in each layer). By using a the decay parameter in training neural networks, reproducible and general models may be obtained at the cost of longer training times. CCN networks were trained to furnish both quantitative and qualitative prediction for a complex IMS data set (229 spectra, 200 input points, and 15 output classes). The advantage of rapid training is that replicate neural networks may be obtained. The precision of replicated network predictions appears to provide a measure of accuracy. Partial least-squares regression (PLS) is used as a comparative method. The CCN with decay rates an order of magnitude larger than learning rate achieves significantly better results than those obtained from an optimal PLS model.