Comparison of detection modes in terms of the necessity of visible region (VIS) and influence of the peel on soluble solids content (SSC) determination of navel orange using VIS–SWNIR spectroscopy

Integrating deep learning and data fusion for enhanced oranges soluble solids content prediction using machine vision and Vis/NIR spectroscopy

The visible/near infrared (Vis/NIR) spectrum will become distorted due to variations in sample color, thereby reducing the prediction accuracy of fruit composition. In this study, we aimed to develop a deep learning model with color correction capability to predict oranges soluble solids content (SSC) based on multi-source data fusion. Initially, a machine vision and Vis/NIR spectroscopy online acquisition device was designed to collect and analyze color images and transmission spectra. Subsequently, data fusion methods were proposed for color features and spectral data. Finally, color-correction one-dimensional convolutional neural network (1D-CNN) models base on multi-source data were constructed. The results showed that, the RMSEP of optimal color-correction model was decreased by 36.4 % and 16.1 % compared to partial least squares model and conventional 1D-CNN model, respectively. The multi-source data fusion of machine vision and Vis/NIR spectroscopy has the potential to improve the accuracy of food composition prediction.

Comparison of color spectrophotometer and Vis/NIR spectroscopy on assessing natural pigments of cucumber applied with different ethephon concentrations

In recent decades, the development of non-destructive measurement methods for agricultural commodities has gained a lot of attention among scientists, but these techniques have different levels of accuracy for each instrument used. Therefore, this study aimed to compare the prediction accuracy of natural pigments, such as Total Carotenoid Content (TCC) and Total Flavonoid Content (TFC) using a color spectrophotometer and Visible/Near-Infrared (Vis/NIR) spectroscopy (381-1065 nm). The effect of ethephon concentration on the spectral characteristics and the accuracy of predicting pigments was studied. The samples used include cucumber fruit, which consisted of the 'Mars', 'Vanesa', and 'Roberto' varieties. During the planting of the fruit, ethephon was applied at different concentrations of 0 ppm, 150 ppm, and 300 ppm. The results showed that the best accuracy for predicting TCC was obtained through a color spectrophotometer (Rcal = 0.89, Rpred = 0.90, RPD = 2.44), while the best prediction for TFC was the Vis/NIR spectroscopy (Rcal = 0.86, Rpred = 0.83, RPD = 1.78). Furthermore, the ethephon affects the spectral characteristics of cucumber fruit. Ethephon concentration of 150 ppm produced the highest accuracy value compared to others. This study proved that the use of non-destructive measurement methods with a color spectrophotometer and Vis/NIR spectroscopy has good performance in predicting TCC and TFC. The techniques are also easy to use, do not require chemicals, and have high accuracy.

Advancement of non-destructive spectral measurements for the quality of major tropical fruits and vegetables: a review

The quality of tropical fruits and vegetables and the expanding global interest in eating healthy foods have resulted in the continual development of reliable, quick, and cost-effective quality assurance methods. The present review discusses the advancement of non-destructive spectral measurements for evaluating the quality of major tropical fruits and vegetables. Fourier transform infrared (FTIR), Near-infrared (NIR), Raman spectroscopy, and hyperspectral imaging (HSI) were used to monitor the external and internal parameters of papaya, pineapple, avocado, mango, and banana. The ability of HSI to detect both spectral and spatial dimensions proved its efficiency in measuring external qualities such as grading 516 bananas, and defects in 10 mangoes and 10 avocados with 98.45%, 97.95%, and 99.9%, respectively. All of the techniques effectively assessed internal characteristics such as total soluble solids (TSS), soluble solid content (SSC), and moisture content (MC), with the exception of NIR, which was found to have limited penetration depth for fruits and vegetables with thick rinds or skins, including avocado, pineapple, and banana. The appropriate selection of NIR optical geometry and wavelength range can help to improve the prediction accuracy of these crops. The advancement of spectral measurements combined with machine learning and deep learning technologies have increased the efficiency of estimating the six maturity stages of papaya fruit, from the unripe to the overripe stages, with F1 scores of up to 0.90 by feature concatenation of data developed by HSI and visible light. The presented findings in the technological advancements of non-destructive spectral measurements offer promising quality assurance for tropical fruits and vegetables.

Effect of potato peel on the determination of soluble solid content by visible near-infrared spectroscopy and model optimization

The quantitative determination of the soluble solid content (SSC) of potatoes using NIR spectroscopy is useful for predicting the internal and external quality of potato products, especially fried products. In this study, the effect of peel on the partial least squares regression (PLSR) quantitative prediction of potato SSC was investigated by transmission and reflection. The results show that the variable sorting for normalization (VSN) pre-processing method improved model accuracy. Additive multiplicative scattering effects and intensity drift interference of the peels were reduced. The model accuracy reached a correlation coefficient of prediction (RP) of 0.85. The selection algorithm using variable combination population analysis and iterative retention of information variables (VCPA-IRIV) demonstrated that peel increases unnecessary information. When the effect of irrelevant variables was reduced, the results reached RP = 0.88 and the root mean square error of prediction (RMSEP) = 0.25 in the transmission mode was close to that of the full-wavelength peeled PLSR model (RP = 0.89 and RMSEP = 0.25). This indicates that the use of the combined algorithm (VSN-VCPA-IRIV) reduces the effect of the peel and enables samples with a peel to still be predicted accurately in the full-wavelength model. It also improves detection efficiency through the extraction of the necessary variables and optimizes the stability and accuracy of the model.

Hyperspectral Imaging with Machine Learning Approaches for Assessing Soluble Solids Content of Tribute Citru

Tribute Citru is a natural citrus hybrid with plenty of vitamins and nutrients. Fruits' soluble solids content (SSC) is a critical quality index. This study used hyperspectral imaging at two spectral ranges (400-1000 nm and 900-1700 nm) to determine SSC in Tribute Citru. Partial least squares regression (PLSR) and support vector regression (SVR) models were established in order to determine SSC using the spectral information of the calyx and blossom ends. The average spectra of both ends as well as their fusion was studied. The successive projections algorithm (SPA) and the correlation coefficient analysis (CCA) were used to examine the differences in characteristic wavelengths between the two ends. Most models achieved performances with the correlation coefficient of the training, validation, and testing sets over 0.6. Results showed that differences in the performances among the models using the one-sided and two-sided spectral information. No particular regulation could be found for the differences in model performances and characteristic wavelengths. The results illustrated that the sampling side was an influencing factor but not the determinant factor for SSC determination. These results would help with the development of real-world applications for citrus quality inspection without concerning the sampling sides and the spectral ranges.

Prediction of Total Soluble Solids and pH of Strawberry Fruits Using RGB, HSV and HSL Colour Spaces and Machine Learning Models

Determination of internal qualities such as total soluble solids (TSS) and pH is a paramount concern in strawberry cultivation. Therefore, the main objective of the current study was to develop a non-destructive approach with machine learning algorithms for predicting TSS and pH of strawberries. Six hundred samples (100 samples in each ripening stage) in six ripening stages were collected randomly for measuring the biometrical characteristics, i.e., length, diameters, weight and TSS and pH values. An image of each strawberry fruit was captured for colour feature extraction using an image processing technique. Channels of each colour space (RGB, HSV and HSL) were used as input variables for developing multiple linear regression (MLR) and support vector machine regression (SVM-R) models. The result of the study indicated that SVM-R model with HSV colour space performed slightly better than MLR model for TSS and pH prediction. The HSV based SVM-R model could explain a maximum of 84.1% and 79.2% for TSS and 78.8% and 72.6% for pH of the variations in measured and predicted data in training and testing stages, respectively. Further experiments need to be conducted with different strawberry cultivars for the prediction of more internal qualities along with the improvement of model performance.

NIR Spectroscopy Oranges Origin Identification Framework Based on Machine Learning

Research on the identification model of orange origin based on machine learning in Near infrared (NIR) spectroscopy. According to the characteristics of NIR spectral data, a complete general framework for origin identification is proposed. It includes steps such as data preprocessing, feature selection, model building and cross validation. Compare multiple preprocessing algorithms and multiple machine learning algorithms under the framework. Based on NIR spectroscopy to identify the origin of orange, a good identification result was obtained. Improve the accuracy of orange origin identification and obtained the best origin identification accuracy of 92.8%.

An Accuracy Improvement Method Based on Multi-Source Information Fusion and Deep Learning for TSSC and Water Content Nondestructive Detection in “Luogang” Orange

The objective of this study was to find an efficient method for measuring the total soluble solid content (TSSC) and water content of “Luogang” orange. Quick, accurate, and nondestructive detection tools (VIS/NIR spectroscopy, NIR spectroscopy, machine vision, and electronic nose), four data processing methods (Savitzky–Golay (SG), genetic algorithm (GA), multi-source information fusion (MIF), convolutional neural network (CNN) as the deep learning method, and a partial least squares regression (PLSR) modeling method) were compared and investigated. The results showed that the optimal TSSC detection method was based on VIS/NIR and machine vision data fusion and processing and modeling by SG + GA + CNN + PLSR. The R2 and RMSE of the TSSC detection results were 0.8580 and 0.4276, respectively. The optimal water content detection result was based on VIS/NIR data and processing and modeling by SG + GA + CNN + PLSR. The R2 and RMSE of the water content detection results were 0.7013 and 0.0063, respectively. This optimized method largely improved the internal quality detection accuracy of “Luogang” orange when compared to the data from a single detection tool with traditional data processing method, and provides a reference for the accuracy improvement of internal quality detection of other fruits.

Optimizing genetic algorithm-partial least squares model of soluble solids content in Fukumoto navel orange based on visible-near-infrared transmittance spectroscopy using discrete wavelet transform

BACKGROUND

The thick rind of Fukumoto navel orange is a great barrier to light penetration, which makes it difficult to evaluate the internal quality of Fukumoto navel orange accurately by visible-near-infrared (Vis-NIR) transmittance spectroscopy. The information carried by the transmission spectrum is limited. Thus, the application of genetic algorithm (GA) for variable selection may not reach the expected results, and selected variables may contain redundancy. In this paper, we present the use of discrete wavelet transforms for optimizing a GA-partial least squares (PLS) model based on Vis-NIR transmission spectra of Fukumoto navel orange. Haar, Db, Sym, Coif and Bior wavelets were used to compress the spectral data selected by GA. Then a PLS model was established based on the variables compressed by each wavelet function.

RESULTS

The use of Db4, Sym4, Coif2 and Bior3.5 succeeded in further simplification of the GA-PLS model by reducing the number of variables by 40-44% without decreasing the prediction accuracy. The application of Bior3.5 not only could reduce the number of variables in the GA-PLS model by 40%, but also increase the value of correlation coefficient of prediction by 1% and decrease the value of root mean square error of prediction by 3%.

CONCLUSIONS

The results indicated that the combination of GA and discrete wavelet transforms for variable selection in the internal quality assessment of Fukumoto navel orange by Vis-NIR transmittance spectroscopy was feasible. © 2019 Society of Chemical Industry.

Noncontact and Wide-Field Characterization of the Absorption and Scattering Properties of Apple Fruit Using Spatial-Frequency Domain Imaging

Spatial-frequency domain imaging (SFDI), as a noncontact, low-cost and wide-field optical imaging technique, offers great potential for agro-product safety and quality assessment through optical absorption (μ_a) and scattering (μ) property measurements. In this study, a laboratory-based SFDI system was constructed and developed for optical property measurement of fruits and vegetables. The system utilized a digital light projector to generate structured, periodic light patterns and illuminate test samples. The diffuse reflected light was captured by a charge coupled device (CCD) camera with the resolution of 1280 × 960 pixels. Three wavelengths (460, 527, and 630 nm) were selected for image acquisition using bandpass filters in the system. The μ_a and μ were calculated in a region of interest (ROI, 200 × 300 pixels) via nonlinear least-square fitting. Performance of the system was demonstrated through optical property measurement of ‘Redstar’ apples. Results showed that the system was able to acquire spatial-frequency domain images for demodulation and calculation of the μ_a and μ. The calculated μ_a of apple tissue experiencing internal browning (IB) were much higher than healthy apple tissue, indicating that the SFDI technique had potential for IB tissue characterization.