Hyperspectral Imaging for Evaluating Impact Damage to Mango According to Changes in Quality Attributes

Research on Defect Detection in Kubo Peach Based on Hyperspectral Imaging Technology Combined with CARS-MIV-GA-SVM Method

Due to the dark red surface of ripe fresh peaches, their internal injury defects cannot be detected using the naked eye and conventional images. The rapid and accurate detection of fresh peach defects can improve the efficiency of fresh peach classification. The goal of this paper was to develop a nondestructive approach to simultaneously detecting internal injury defects and external injuries in fresh peaches. First, we collected spectral data from 347 Kubo peach samples using hyperspectral imaging technology (900-1700 nm) and carried out pretreatment. Four methods (the competitive adaptive reweighting algorithm (CARS), the combination of CARS and the average influence value algorithm (CARS-MIV), the combination of CARS and the successive projections algorithm (CARS-SPA), and the combination of CARS and uninformative variable elimination (CARS-UVE)) were used to extract the characteristic wavelength. Based on the characteristic wavelength extracted using the above methods, a genetic algorithm optimization support vector machine (GA-SVM) model and a least-squares support vector machine (LS-SVM) model were used to establish classification models. The results show that the combination of CARS and other feature wavelength extraction methods can effectively improve the prediction accuracy of the model when the number of wavelengths is small. Among them, the discriminant accuracy of the CARS-MIV-GA-SVM model reaches 93.15%. In summary, hyperspectral imaging technology can accomplish the accurate detection of Kubo peaches defects, and provides feasible ideas for the automatic classification of Kubo peaches.

Classification modeling method for hyperspectral stamp-pad ink data based on one-dimensional convolutional neural network

In the questioned document, the examination of stamp-pad ink is crucial scientific evidence to discern the difference between genuine and forged documents. In this study, a new method for rapid and non-destructive identification of types of stamp-pad inks by combining hyperspectral imaging (HSI) technology and deep learning was developed. Twenty stamp-pad inks of different brands and models were collected and numbered in turn, and then, each of them was sealed six times repeatedly on the A4 printing paper for the test. After that, the hyperspectral imager was used to collect the hyperspectral images and the reflectance spectral data were obtained after pixel fusion. Principal component analysis (PCA) and non-negative matrix factorization (NMF) were used to deal with the dataset, but visual results were not good. Then, back propagation neural network (BPNN) and one-dimensional convolutional neural network (1D-CNN) were constructed and their merits and drawbacks were compared. The final loss function of the BPNN of training set and validation set was stable at 0.27 and 0.42, and the classification accuracy of the training set and validation set reached 90.02% and 83.99%, respectively. Compared with the BPNN, the 1D-CNN had better stability and efficiency for the classification. The loss function of the training set and validation set was as low as 0.068 and 0.075, and the final classification accuracy reached 98.30% and 97.94%, respectively. Therefore, the combination of hyperspectral imaging technology and 1D-CNN represents a potentially simple, non-destructive, and rapid method for stamp-pad inks detection and classification.

Hyperspectral Imaging for the Nondestructive Quality Assessment of the Firmness of Nanguo Pears Under Different Freezing/Thawing Conditions

Firmness changes in Nanguo pears under different freezing/thawing conditions have been characterized by hyperspectral imaging (HSI). Four different freezing/thawing conditions (the critical temperatures, numbers of cycles, holding time and cooling rates) were set in this experiment. Four different pretreatment methods were used: multivariate scattering correction (MSC), standard normal variate (SNV), Savitzky-Golay standard normal variate (S-G-SNV) and Savitzky-Golay multiplicative scattering correction (S-G-MSC). Combined with competitive adaptive reweighted sampling (CARS) to identify characteristic wavelengths, firmness prediction models of Nanguo pears under different freezing/thawing conditions were established by partial least squares (PLS) regression. The performance of the firmness model was analyzed quantitatively by the correlation coefficient (R), the root mean square error of calibration (RMSEC), the root mean square error of prediction (RMSEP) and the root mean square error of cross validation (RMSECV). The results showed that the MSC-PLS model has the highest accuracy at different cooling rates and holding times; the correlation coefficients of the calibration set (Rc) were 0.899 and 0.927, respectively, and the correlation coefficients of the validation set (Rp) were 0.911 and 0.948, respectively. The accuracy of the SNV-PLS model was the highest at different numbers of cycles, and the Rc and the Rp were 0.861 and 0.848, respectively. The RMSEC was 65.189, and the RMSEP was 65.404. The accuracy of the S-G-SNV-PLS model was the highest at different critical temperatures, with Rc and Rp values of 0.854 and 0.819, respectively, and RMSEC and RMSEP values of 74.567 and 79.158, respectively.