Loading...
9 results
Search Results
Now showing 1 - 9 of 9
- Estimation of soluble solids content and fruit temperature in 'rocha' pear using Vis-NIR spectroscopy and the spectraNet–32 deep learning architecturePublication . Martins, J. A; Rodrigues, Daniela; Cavaco, A. M.; Antunes, Maria Dulce; Guerra, Rui Manuel Farinha das NevesSpectra-based methods are becoming increasingly important in Precision Agriculture as they offer non-destructive, quick tools for measuring the quality of produce. This study introduces a novel approach for esti-mating the soluble solids content (SSC) of 'Rocha' pears using the SpectraNet-32 deep learning architecture, which operates on 1D fruit spectra in the visible to near-infrared region (Vis-NIRS). This method was also able to estimate fruit temperatures, which improved the SSC prediction performance. The dataset consisted of 3300 spectra from 1650 'Rocha' pears collected from local markets over several weeks during the 2010 and 2011 seasons, which had varying edaphoclimatic conditions. Two types of partial least squares (PLS) feature selection methods, under various configurations, were applied to the input spectra to identify the most significant wavelengths for training SpectraNet-32. The model's robustness was also compared to a similar state-of-the-art deep learning architecture, DeepSpectra, as well as four other classical machine learning algorithms: PLS, multiple linear regression (MLR), support vector machine (SVM), and multi-layer perceptron (MLP). In total, 23 different experimental method configurations were assessed, with 150 neural networks each. SpectraNet-32 consistently outperformed other methods in several metrics. On average, it was 6.1% better than PLS in terms of the root mean square error of prediction (RMSEP, 1.08 vs. 1.15%), 7.7% better in prediction gain (PG, 1.67 vs. 1.55), 3.6% better in the coefficient of determination (R2, 0.58 vs. 0.56) and 5.8% better in the coefficient of variation (CV%, 8.35 vs. 8.86).
- A TSS classification study of 'Rocha' pear (Pyrus communis L.) based on non-invasive visible/near infra-red reflectance spectraPublication . Bexiga, Florentino; Rodrigues, Daniela; Guerra, Rui Manuel Farinha das Neves; Brazio, António; Balegas, Tiago; Cavaco, A. M.; Antunes, Maria Dulce; Valente de Oliveira, JOSÉThe study focuses on the application of machine learning techniques for classifying the internal quality of 'Rocha' Pear (Pyrus communis L.), i.e., the total soluble solids (TSS), using the non-invasive technique of visible/near infra-red reflectance spectroscopy. Six representative classifiers were evaluated under realistic experimental conditions. The classifiers include representatives of classic parametric (logistic and multiple linear regression), non-parametric distance based methods (K-nearest neighbors), correlation-based (partial least squares), ensemble methods (random forests) and maximum margin classifiers (support vector machines). The classifiers were assessed against metrics such as accuracy, Cohen's Kappa, F-Measure, and the area under the precision recall curve (AUC) in a 10 x 10-fold cross-validation plan. For result analysis non-parametric statistical test of hypotheses were employed. A total of 4880 fruit samples from different origins, maturation states, and harvest years were considered. The main conclusion is that the maximum margin classifier outperforms all the others studied ones, including the commonly used partial least squares. The conclusion holds for both a reflectance spectrum with 1024 features and for a 128 subsample of these. An estimate of the out-of-sample performance for the best classifier is also provided.
- On the application of spatially resolved reflectance and diffuse light backscattering goniometry to the prediction of firmness in apple ‘bravo de esmolfe’Publication . Guerra, Rui Manuel Farinha das Neves; Almeida, Sandro; Cavaco, A. M.; Antunes, Maria Dulce; Nunes, CarlaIn this study we have made exploratory tests on a set of 40 apples (Malus domestica Borkh.) ‘Bravo de Esmolfe’, using spatially resolved reflectance (SRR) and diffuse light backscattering goniometry (DLBG). The objective was to test the potential of DLBG for firmness prediction, as compared with SRR, whose potential has been already proved in the literature. SRR is performed with a red diode laser and a CMOS camera. DLBG uses the same laser shining on the apple and a photomultiplier tube collecting the light reemitted from a small area, at angles ranging from 90 deg (tangent to the surface) to 180 deg (normal to the surface). From the measurements several parameters have been calculated (e.g. decay exponent for SRR profiles, anisotropy factor for the DLBG angular distributions) and Partial Least squares (PLS) models for the prediction of firmness were build. The model based on DLBG variables (only) and on SRR variables (only) gave similar results. From here we conclude that, within the obvious statistical limitations of the test, DLBG seems to match the potential of SRR for firmness prediction. The possibility of combining both measures in one model is also discussed.
- Non-destructive soluble solids content determination for ‘Rocha’ Pear Based on VIS-SWNIR spectroscopy under ‘Real World’ sorting facility conditionsPublication . Passos, Dário; Rodrigues, Daniela; Cavaco, Ana M.; Antunes, Maria Dulce; Guerra, Rui Manuel Farinha das NevesIn this paper we report a method to determine the soluble solids content (SSC) of 'Rocha' pear (Pyrus communis L. cv. Rocha) based on their short-wave NIR reflectance spectra (500-1100 nm) measured in conditions similar to those found in packinghouse fruit sorting facilities. We obtained 3300 reflectance spectra from pears acquired from different lots, producers and with diverse storage times and ripening stages. The macroscopic properties of the pears, such as size, temperature and SSC were measured under controlled laboratory conditions. For the spectral analysis, we implemented a computational pipeline that incorporates multiple pre-processing techniques including a feature selection procedure, various multivariate regression models and three different validation strategies. This benchmark allowed us to find the best model/preproccesing procedure for SSC prediction from our data. From the several calibration models tested, we have found that Support Vector Machines provides the best predictions metrics with an RMSEP of around 0.82 ∘ Brix and 1.09 ∘ Brix for internal and external validation strategies respectively. The latter validation was implemented to assess the prediction accuracy of this calibration method under more 'real world-like' conditions. We also show that incorporating information about the fruit temperature and size to the calibration models improves SSC predictability. Our results indicate that the methodology presented here could be implemented in existing packinghouse facilities for single fruit SSC characterization.
- Preliminary Results on the Non-Destructive Determination of Pear (Pyrus communis L.) cv. Rocha Ripeness by Visible/Near Infrared Reflectance SpectroscopyPublication . Cavaco, A. M.; Antunes, Maria Dulce; da Silva, J. M.; Guerra, Rui Manuel Farinha das NevesPear (Pyrus communis L.), cv. Rocha was rapidly adopted by consumers due to its inherent quality and currently has great acceptance in both national and international markets, being mainly produced in the west region of Portugal. We report here a first approach to the use of the non-intrusive method of Visible/Near Infrared Reflectance Spectroscopy (Vis/NIRS) to estimate the ripeness of pear cv. Rocha. Mature unripe pears obtained from Frutoeste (Mafra, Portugal) after a six-month cold-storage, were maintained in a dark room at circa 20 degrees C during three weeks. They were followed using the Vis/NIRS in the wavelength band between 400 and 950 nm with two different configurations for the spectra acquisition, namely the Integrating Sphere (IS) and the Partial Transmittance (PT). The diffuse reflectance spectra obtained by the two configurations were compared with the respective fruit ripening parameters (colour, firmness, soluble solids content and % dry matter), determined through the standard techniques. Concerning the rough estimation of ripening parameters, data suggested an increase in both the intensity in the green to red band and pulp %dry matter, but a decreasing firmness. All other parameters remained constant. Relatively to the optical results, we have observed that the PT spectra exhibited clearer features than the IS spectra, especially from 700 nm onwards. This is probably due to the fact that the PT configuration probes more deeply into the fruit pulp. Three peaks at 600 (circa 30%), 725 and 812 nm (both at circa 50%) and a minimum at 675 nm, were identified in both IS and PT spectra. The values of reflectance peaks were approximately constant during ripening, but they moved to slightly lower wavelengths in the second week. A significant increase (circa 3-fold) in the minimal diffuse reflectance was observed in the second week, most probably associated partially, to a decrease in the fruit peel chlorophyll content.
- Ripening assessment of ‘Ortanique’ (Citrus reticulata Blanco x Citrus sinensis (L) Osbeck) on tree by SW-NIR reflectance spectroscopy-based calibration modelsPublication . Pires, Rosa; Guerra, Rui Manuel Farinha das Neves; Cruz, Sandra; Antunes, MDC; Brazio, António; Afonso, Andreia M.; Daniel, Mariana; Panagopoulos, Thomas; Gonçalves, Isabel; Cavaco, Ana M.The aim of this study was the non-destructive assessment of ‘Ortanique’ (Citrus reticulata Blanco x Citrus sinensis (L) Osbeck) ripening, based on the prediction of internal quality attributes (IQA) by short-wave near-infrared reflectance spectroscopy (SW-NIRS) calibration models. Spectra from fruit of 50 trees located in two different orchards, were acquired on tree using a customized portable visible near-infrared (vis-NIR) system. Partial least squares (PLS) was used to build the various IQA calibration models. The models were tested through internal validation (IV) and external validation (EV). Generally, the IV results were always superior to those of EV: regarding IV, a high regression coefficient (R2) and low root mean square error of prediction (RMSEP) were achieved, revealing a good predictive performance for juice pH (R2 = 0.80; RMSEP = 0.10; SDR = 2.23), soluble solids content (SSC) (R2 = 0.79; RMSEP = 0.75 %; SDR = 2.27), titratable acidity (TA) (R2 = 0.73; RMSEP = 0.24 % citric acid; SDR = 1.94) and the maturation index (MI) (R2 = 0.80; RMSEP = 1.38; SDR = 2.2). The best EV predictions were obtained for TA (R2 = 0.69; RMSEP = 0.38 % citric acid; SDR = 1.24), and MI (R2 = 0.69; RMSEP = 2.07; SDR = 1.49). Calibration models for glucose, fructose and sucrose showed medium-coarse predictions for both validation strategies. A detailed investigation of MI models was performed, to understand the causes of their poor EV results. In the context of EV, model updating strategies were explored by using some validation samples to improve the calibration model. The methods of bias correction and spiking were tested, showing a clear improvement in the predictions.
- Feature discovery in NIR spectroscopy based Rocha pear classificationPublication . Daniel, Mariana; Guerra, Rui Manuel Farinha das Neves; Brazio, António; Rodrigues, Daniela; M. Cavaco, A.; Antunes, Maria Dulce; Valente de Oliveira, JOSÉNon-invasive techniques for automatic fruit classification are gaining importance in the global agro-industry as they allow for optimizing harvesting, storage, management, and distribution decisions. Visible, near infra-red (NIR) diffuse reflectance spectroscopy is one of the most employed techniques in such fruit classification. Typically, after the acquisition of a fruit reflectance spectrum the wavelength domain signal is preprocessed and a classifier is designed. Up to now, little or no work considered the problem of feature generation and selection of the reflectance spectrum. This work aims at filling this gap, by exploiting a feature engineering phase before the classifier. The usual approach where the classifier is fed directly with the reflectances measured at each wavelength is contrasted with the proposed division of the spectra into bands and their characterization in wavelength, frequency, and wavelength-frequency domains. Feature selection is also applied for optimizing efficiency, predictive accuracy, and for mitigating over-training. A total of 3050 Rocha pear samples from different origins and harvest years are considered. Statistical tests of hypotheses on classification results of soluble solids content - a predictor of both fruit sweetness and ripeness - show that the proposed preliminary phase of feature engineering outperforms the usual direct approach both in terms of accuracy and in the number of necessary features. Moreover, the method allows for the identification of features that are physical chemistry meaningful.
- Non-destructive follow-up of ‘Jintao’ kiwifruit ripening through VIS-NIR spectroscopy – individual vs. average calibration model’s predictionsPublication . Afonso, Andreia M.; Antunes, Maria Dulce; Cruz, Sandra; Cavaco, A. M.; Guerra, Rui Manuel Farinha das NevesVisible/near infrared spectroscopy (Vis-NIRS) was used to monitor the yellow-fleshed kiwifruit (Actinidia chinensis Planch 'Jintao') ripening on two selected orchards along 13 weeks, from pre-harvest to the late harvest. Calibration models for several Internal Quality Attibutes (IQA) were built from the spectral data of 375 individual kiwifruit. The analyzed IQA were L*, a* and b* from the CIELAB color space, hue angle, chroma, firmness, dry matter (DM), soluble solids content (SSC), juice pH and titratable acidity (TA). Different pre-processing methods were tested for the construction of PLS calibration models. SSC and Hue were the best performing models with a correlation coefficient of 0.81 and 0.88, and root mean square error of prediction (RMSEP) of 1.27% and 1.95 degrees, respectively. The interpretation of the models in terms of the known absorption bands and the impact of signal to noise ratio (SNR) in them is discussed. The calibration models were used to perform average predictions of the IQA on orchard subareas, for each day of the experiment. These average predictions were compared with the IQA's average reference values on the same subareas and days. The model's metrics improved significantly through the averaging procedure, with RMSEP = 0.26-0.36% and R-2 = 0.99 for SSC; and RMSEP = 0.42 degrees - 0.56 degrees and R-2 = 1 for Hue. Since orchard management is done essentially through averages and not individual values, this result reinforces the applicability of the NIR technology for follow-up of fruit ripening in the tree.
- 'Rocha' pear firmness predicted by a Vis/NIR segmented modelPublication . Cavaco, A. M.; Pinto, Patricia IS; Antunes, Maria Dulce; da Silva, J. M.; Guerra, Rui Manuel Farinha das NevesWe present a segmented partial least squares (PLS) prediction model for firmness of 'Rocha' pear (Pyres communis L) during fruit ripening under shelf-life conditions. Pears were collected from three different orchards. Orchard I provided the pears for model calibration and internal validation (set 1). These were transferred to shelf-life in the dark at 20 +/- 2 degrees C and 70% RH, immediately after harvest. External validation was performed on the pears from the other two orchards (sets 2 and 3), which were stored under different conditions before shelf-life. Fruit was followed in the shelf-life period by visible/near infrared reflectance spectroscopy (Vis/NIRS) in the range 400-950 nm. The correlation between firmness and the reflectance at some wavelength bands was markedly different depending on ripening stage. A segmented partial least squares model was then constructed to predict firmness. This PLS model has two segments: (1) unripe and ripening/ripe pears (high firmness); (2) over-ripe pears (low firmness). The prediction is done in two steps. First, a full range model (full model) is applied. When the full model prediction gives a low firmness value, then the over-ripe model is applied to refine the prediction. The full model is reasonably significant in regression terms, robust, but allows only a coarse quantitative prediction (standard deviation ratio, SDR = 2.48, 1.50 and 2.40 for sets 1, 2 and 3, respectively). Also, RMSEP% = 139%, 91% and 56%, indicating large relative errors at low firmness values. The segmented model improved moderately the correlation, and the values of RMSEC, RMSEP and SDR: it improved significantly the RMSEP% (29%, 55% and 31%), providing an improvement of the relative prediction errors at low firmness values. This method improves the ordinary PLS models. Finally, we tested whether chlorophyll alone was enough for a predictive model for firmness, but the results showed that the absorption of chlorophyll alone does not explain the performance of the PLS models. (C) 2008 Elsevier B.V. All rights reserved.