Search

• Make a quantitative damage device for Lingwu long jujube. • Spectral of samples was scanned by VIS/NIR hyperspectral system. • Using SPA, UVE, CARS, iVISSA and 2D-COS to select the characteristic wavelength. • Establish the PLS-DA classification models. Early detection of internal bruise is one of the major challenges in postharvest quality sorting processes in Lingwu long jujube. In this study, the visible/near infrared (VIS/NIR) hyperspectral imaging system (400–1000 nm) was used to rapidly detect the intact and damaged jujube at five time points after mechanical damage (2 h, 4 h, 8 h, 12 h and 24 h). The region of interest of samples was selected by ENVI software, and the average spectrum was calculated for modelling. Different preprocessing methods were used to transform and enhance the spectral signal. Partial least squares-discriminant analysis (PLS-DA) classification models of the original and preprocessed spectra were established. The de-trending-PLS-DA model had the best effect, the accuracy of the calibration set and prediction set were 85.56% and 92.22%, respectively. The spectra were pre-processed by the de-trending algorithm, and the successive projection algorithm (SPA), uninformative variable elimination (UVE), competitive adaptive reweighted sampling (CARS), interval variable iterative space shrinkage approach (iVISSA), cluster analysis (CA), two-dimensional correlation spectra (2D-COS), UVE-SPA, CARS-SPA, iVISSA-SPA, CA-SPA and 2D-COS-SPA method were used to select characteristic variables. The PLS-DA model based on feature variables was established. The de-trending-CARS-PLS-DA model with 63 variables was found to be the optimal model. In the de-trending-CARS-PLS-DA model, the accuracy of calibration set and prediction set were 86.67% and 91.11%, respectively. It was found that the model accurately detected bruising in jujube at 8 h after bruising, and the accuracy of calibration set and prediction set were 100%. The results showed that VIS/NIR hyperspectral imaging technology combined with PLS-DA could discriminate different stages of bruising in Lingwu long jujube. This study may help develop an online detection system of bruising in Lingwu long jujube. [ABSTRACT FROM AUTHOR]

A hyperspectral imaging technique was used for acquiring reflectance images to identify common defects (bruise, insect-infestetation and cracks) on jujube fruit. Hyperspectral images of jujubes were evaluated from the regions of interest through principal component analysis (PCA) to select five optimal wavelengths (420,521,636,670,679 nm) from 300 samples in the spectral region of 400–1000 nm and four important wavelength (1028,1118,1359,1466 nm) in the region of 978–1586 nm. Compared with support vector machine (SVM) models, the soft independent modeling of class analogy (SIMCA) models of intact, cracked, bruised, and insect-infested jujubes based on five wavelengths in NIR showed good performance with high classification rates of 96%, 96%, 93.9% and 95.6%, respectively. This research demonstrates the feasibility of implementing hyperspectral imaging for identifying common defects and enhancing the product quality and marketability. [ABSTRACT FROM AUTHOR]

[Display omitted] • BOSS exhibited better superiority in wavelength selection compared with SCARS. • Texture features entropy, inertia, and gradient mean square error were highly correlated with the bruise over time (|r|>0.7). • Spectral data combined with textural features with the highest correlations improved prediction accuracy. • The 1D-CNN model had the advantage of handling large datasets. The timely detection and classification of bruised Lingwu long jujube is crucial for reducing economic losses in the agricultural industry. Traditional machine learning methods had been effectively used for bruise time classification with limited samples while the challenges were presented when the amount of data expanded. Deep learning methods can offer a solution. The hyperspectral image of the jujube sample was obtained using near-infrared hyperspectral imaging system (900–1700 nm). The optimal input variables were obtained by comparing the performance of the partial least squares-discriminant analysis, artificial neural network, and one-dimensional convolutional neural network (1D-CNN) models based on the spectral data and the fusion data of spectral and textural features. The optimal variable was the fusion data that could improve the model performance (from 92.68% to 96.10%). The above-mentioned three classifiers were used to establish models based on fusion data for two different dataset sizes consisting of large datasets (LD) (820) and small datasets (SD) (200) to explore the influence of dataset size on model performance. As the sample size increased, the 1D-CNN had good robustness with a classification rate of 94.00% for SD and 96.10% for LD in the prediction sets for all samples. The classification rate of prediction sets for 0 h after bruising increased from 90.00% to 92.68%, and from 80.00% to 90.24% in the class of 24 h after bruising. The findings provided a new approach to building a large spectral library and developing an online classification and detection system, thereby providing new ideas for the development of grading systems for Lingwu long jujube. [ABSTRACT FROM AUTHOR]