Soluble solids content (SSC) is one of the important components for evaluating fruit quality. The rapid development of hyperspectral imagery provides an efficient method for non-destructive detection ...of SSC. Previous studies have shown that the internal quality evaluation of fruits based on spectral information features achieves better results. However, the lack of comprehensive features limits the accurate estimation of fruit quality. Therefore, the deep learning theory is applied to the estimation of the soluble solid content of peaches, a method for estimating the SSC of fresh peaches based on the deep features of the hyperspectral image fusion information is proposed, and the estimation models of different neural network structures are designed based on the stack autoencoder–random forest (SAE-RF). The results show that the accuracy of the model based on the deep features of the fusion information of hyperspectral imagery is higher than that of the model based on spectral features or image features alone. In addition, the SAE-RF model based on the 1237-650-310-130 network structure has the best prediction effect (R2 = 0.9184, RMSE = 0.6693). Our research shows that the proposed method can improve the estimation accuracy of the soluble solid content of fresh peaches, which provides a theoretical basis for the non-destructive detection of other components of fresh peaches.
•The SSC of tomatoes online assessment based on Vis/NIRS diffuse transmission system.•The determination of sample orientation and light settings for online detection.•The online model optimization by ...the spectral pretreatment and key wavelength selection.•The online model compensation by the physiological traits of tomatoes.
This study developed an efficient and compensable model for predicting soluble solids content (SSC) of tomato, based on the self-developed online Vis/NIRS diffuse transmission system. The fruit stem-calyx axis horizontal coupled with suitable light settings (path and intensity) was determined as the best measurement parameters, which significantly reduce the stray light and also optimize the light propagation inside tomato. The pretreatment method of Savitzky-Golay smoothing (SGS) combined with multiplicative scatter correction (MSC) could eliminate the spectral difference between samples and the inherent system noise in the raw spectral of tomato. The partial least squares regression (PLSR) model based on the 22 key wavelengths selected by competitive adaptive reweighted sampling (CARS) had better model performance than the full wavelength model. Finally, the CARS-PLSR model was further optimized by compensating physiological traits of height and weight with Rp of 0.91 and RMSEp of 0.17%. Our results in the present study demonstrated the potential of using online Vis/NIRS diffuse transmission spectra combined with model optimization and compensation as a valuable method for tomato SSC assessment.
A partial least squares regression (PLSR) model to map internal soluble solids content (SSC) of apples using visible/near-infrared (VNIR) hyperspectral imaging was developed. The reflectance spectra ...of sliced apples were extracted from hyperspectral absorbance images obtained in the 400–1000 nm range. Prediction models for SSC mapping were developed for three different measurement/sampling designs that varied in the number and size of the regions of interest (ROIs) used for apple SSC measurement and spectral averaging. Case 1 used 29 small ROIs per apple, Case II used 9 moderate-size ROIs per apple, and Case III used 5 large ROIs per apple. The optimal pre-treatment of the spectra extracted from the hyperspectral images was investigated to enhance the performance of the prediction models. The coefficients of determination and root mean square errors of the best-performing models were, respectively, 0.802 and ±0.674 °Brix for Case I, 0.871 and ±0.524 °Brix for Case II, and 0.876 and ±0.514 °Brix for Case III. The accuracy of the PLSR models was enhanced by using the spectra and SSC measured/averaged from the fewer but larger areas of the apples rather than from more numerous but smaller areas. PLS images of SSC showed the predicted internal distribution of SSC within the apples. The overall results demonstrate that hyperspectral absorbance imaging techniques may be useful for mapping internal soluble solids content of apples.
•PLSR models to map internal SSC of apples were developed.•The SSC models of three different sampling regions were investigated.•The internal SSC distribution within the apples was shown by PLS images of SSC.•The performance of the SSC models was enhanced as the measurement area increased.
Display omitted
•The effects of maturity variation on SSC and firmness detection of apples were studied.•There were significant differences in the spectrum and quality of apples at different maturity ...stages.•The effect of maturity variation could be eliminated by establishing the multi-maturity model.•The long-term performance of the multi-maturity model could be improved using S/B correction.
In this study, the effect of maturity variation on the prediction of the soluble solids content (SSC) and firmness of apples was determined using visible and near-infrared spectroscopy. In 2018, 520 apples from six ripening stages were collected. The single maturity model and multi-maturity model of SSC and firmness were established using partial least-squares regression. Apples at the same and different maturity stages were used to verify the developed model. Whereas the single maturity model was affected by maturity variation, the multi-maturity model could accurately predict the SSC and firmness of apples at different maturity stages. The multi-maturity model developed based on six maturity calibration sets had the best predictive performance. The root mean square error of prediction (RMSEP) of SSC and firmness was 0.614–0.802 °Brix and 0.402–0.650 kg/cm2, respectively. The long-term performance of the optimal multi-maturity model was evaluated using validation sets. The predictive performance was decreased and the RMSEP increased when the model was used to predict the SSC and firmness of apples in different seasons. The predictive performance of the model was improved after slope/bias (S/B) correction, and the RMSEP of SSC and firmness decreased to 0.405–0.587°Brix and 0.518–0.628 kg/cm2 respectively. Overall, the multi-maturity model eliminated the effect of maturity variation, and the multi-maturity model coupled with S/B correction permitted the rapid and accurate detection of the SSC and firmness of apples at different maturity stages and in different seasons.
Changes of temperature and sunlight disturb the spectra and furtherly influence the prediction outcome of near infrared spectroscopy (NIRS) in assessment of fruit quality during field measurement. ...The hand-held NIRS device combination of chemometrics method was investigated for correcting the influence of temperature and sunlight to the grape spectra. The distribution of the soluble solids content (SSC) within one bunch was analyzed, and the middle region of the bunch was recommended for NIRS sampling position. Chemometrics algorithms of global model, external parameter orthogonalisation (EPO) and generalised least square weighting (GLSW) were employed to correct the influence of temperature and sunlight. Comparison of global model and GLSW, EPO improved the performance of the partial least square regression (PLSR) models with coefficient of determination (Rp2) of 0.88–0.90, root mean square error of prediction (RMSEPiv) in the internal validation set of 0.89–0.94% and ratio of standard deviation (SD) to RMSEPiv (RPD) of 3.27–3.14 in predicting the samples at three temperatures and Rp2 of 0.98, RMSEPiv of 0.50% and RPD of 7.00 outdoor measurement. The results suggested that it was feasibility for correcting the influence of temperature and sunlight to the hand-held NIRS device to assess the grape bunch quality for outdoor use.
•Middle position within one bunch was recommended for instrument sampling.•EPO was recommended for temperature and sunlight correction.•Results support use of portable NIRS with varying fruit temperature and sunlight.
•SSC is the major characteristic for assessing quality of fresh peaches.•Hyperspectral imaging was used to measure the SSC.•Different types of PLS models were established and compared, ...respectively.•The multi-region combination model was superior to local region models.•A robust and accurate model could be built using the data of multiple regions.
Soluble solids content (SSC) is one of the most important factors determining the quality and price of fresh fruits. This study was carried out to compare the different models for more robust and accurate evaluation of SSC in ‘Pinggu’ peaches using hyperspectral imaging in the visible and near infrared spectral range (400–1000 nm). The local region models and multi-region combination model based on full wavelengths were established and compared by partial least squares, respectively. The results of model analysis showed that multi-region combination model was superior to local region models due to insensitivity to the variation of the local regions of interest. Then, four typical wavelength selection algorithms including Monte Carlo cross-validation, successive projections algorithm, competitive adaptive reweighted sampling and random frog were utilized to select the effective wavelengths (EWs) for rapid quantitative determination of SSC, respectively. It was found that random frog was the most effective algorithm for local region models and multi-region combination model in SSC analyses. Considering the practical industrial application, finally, different type of models developed based on EWs selected by random frog were applied to predict the SSC of whole ‘Pinggu’ peaches from new test samples. EW-multi-region combination model achieved the optimal prediction results with rp of 0.86 and RMSEP of 0.67. The overall results of this study revealed that the hyperspectral imaging coupled with effective wavelength selection can be used to non-invasively and fast measure the SSC of ‘Pinggu’ peaches and a more robust and accurate model could be established based on multi-region information rather than any local region. These results can provide a useful reference for global evaluation of the internal quality attributes of fresh fruits.
•A Hg (Ar) lamp was used to correct wavelength of two developed Vis/NIR devices.•Model transfer methods were compared for apple SSC detection.•PDS was recommended for PLS model transfer after ...wavelength correction.•S/B was further used to correct SSC predictions for the independent validation set.
Calibration transfer is an important step for practical applications of Visible and Near-infrared (Vis/NIR) instruments, making the developed model transferable and avoiding recalibration. A calibration transfer method between two developed portable Vis/NIR devices (master and slave devices) for predicting soluble solids content (SSC) of apples was investigated in this study. The partial least squares (PLS) calibration models based on the spectra of the master and the slave devices in the range of 550–930 nm yielded high prediction performance, with the correlation coefficient (Rp) and the root mean square error of the prediction set (RMSEP) of 0.918, 0.552 % and 0.881, 0.666 %, respectively. However, the direct use of the PLS model built by the master instrument to the slave instrument was impracticable. A Hg (Ar) lamp was used to correct the spectral dimension for the two devices, followed by the transfer performance comparison of three methods including piecewise direct standardization (PDS), spectral space transformation (SST), and calibration model transformation based on canonical correlation analysis (CTCCA). The prediction results indicated that PDS yielded better performance when the window size was 3 and the number of the transfer samples was 25, with Rp and RMSEP of 0.874 and 0.713 %, respectively. Lower spectral angle θ¯ and higher spectral correlation coefficient r¯ also illustrated that PDS had a preferable performance compared with SST and CTCCA.After PDS and slope/bias (S/B), the SSC was successfully predicted, achieving high accuracy of Rp = 0.926 and RMSEP = 0.778 %. The above results illustrated that the proposed algorithm was a promising calibration transfer method from the master device to the slave device, and could effectively compensate for the differences of spectral response between the developed Vis/NIR devices and different batches of samples.
•The μs' of strawberry generally decreased with increased maturity.•The best correlations between μa and SSC and moisture content were found at 1411 nm.•The models based on μa performed best in ...determining SSC and moisture content.•Absorption property has great potential in determining internal quality of strawberry.
To understand the relationship of optical properties with soluble solids content (SSC) and moisture content of strawberry during ripening, a single integrating sphere system was built to estimate the absorption coefficient (μa) and reduced scattering coefficient (μs') of strawberry in white, color turning and red ripening stages over the wavelength range of 550−850 nm and 950−1650 nm. The relationship between optical properties and SSC and moisture content was analyzed, and the determination models for SSC and moisture content were established by using partial least squares regression and support vector machine methods based on the spectra of μa, μs', and μa together with μs'. The results showed that the absorption peaks of strawberry were at 675, 975, 1197 and 1411 nm, and the μs' generally decreased with increased maturity of strawberry. The μa was positively correlated with SSC and negatively correlated with moisture content, while the μs'was positively correlated with moisture content and negatively with SSC. The best correlations of μa with SSC and moisture content were found at 1411 nm with the correlation coefficients of 0.72 and -0.74, respectively. The established support vector machine models based on the μa spectra in 950−1650 nm and 550−850 nm had the smallest root-mean-squares error of calibration set of 0.98 % and 0.89 % for SSC and moisture content, respectively. This study indicates that SSC and moisture content mainly affect the absorption property of strawberry, and μa has greater potential than μs' and μa together with μs' in determining the internal quality of strawberry.
•Soluble solids content (SSC) is an important quality attribute of fresh fruit.•The multi-cultivar model was developed to assess SSC in three cultivars of pears.•Multi-cultivar model was superiorto ...individual-cultivar model for SSC analysis.•CARS was a powerful tool to select the most effective variables.•CARS-MLR model is optimal for SSC prediction of three cultivars of pears.
Soluble solids content (SSC) is one of the most important quality attributes affecting the price of fresh fruit. The individual-cultivar model is the mostcommon SSC analysis model. However, this type of model is not the optimal for assessment of SSC in the different cultivars of fruit. In this study, the feasibility of using multi-cultivar model for quantitatively determining SSC in three cultivars of pears was observed based on visible-NIR spectroscopy. The multi-cultivar and individual-cultivar models were developed and different variable selection algorithms were used to optimize models. Results showed that the multi-cultivar model was superior to individual-cultivar models for SSC prediction of all samples and competitive adaptive reweighted sampling (CARS) did better than Monte Carlo-uninformative variable elimination (MC-UVE) and successive projections algorithm (SPA) for selection of effective variables. Based on the selected variables, CARS-PLS and CARS-MLR multi-cultivar models can achieve effective prediction for SSC of three cultivars of pears with similar detection accuracy. The coefficients of determination for prediction set (RP2) and root mean square error of prediction (RMSEP) obtained by these two types of models are 0.90–0.92 and 0.23–0.30 for three cultivars of pears. The overall results demonstrated that it was feasible to accurately determine SSC of different cultivars of pears using the multi-cultivar model, CARS was a powerful tool to select the efficient variables, and CARS-PLS and CARS-MLR were simple and excellent for the spectral calibration.