Determination of apple varieties by near infrared reflectance spectroscopy coupled with improved possibilistic Gath–Geva clustering algorithm

Application of Hyperspectral Imaging as a Nondestructive Technology for Identifying Tomato Maturity and Quantitatively Predicting Lycopene Content

Maturity is a crucial indicator in assessing the quality of tomatoes, and it is closely related to lycopene content. Using hyperspectral imaging, this study aimed to monitor tomato maturity and predict its lycopene content at different maturity stages. Standard normal variable (SNV) transformation was applied to preprocess the hyperspectral data. Then, using competitive adaptive reweighted sampling (CARS), the characteristic wavelengths were selected to simplify the calibration models. Based on the full and characteristic wavelengths, a support vector classifier (SVC) model was developed to determine tomato maturity qualitatively. The results demonstrated that the classification accuracy using the characteristic wavelength led to the obtention of better results with an accuracy of 95.83%. In addition, the support vector regression (SVR) and partial least squares regression (PLSR) models were utilized to predict lycopene content. With a coefficient of determination for prediction (R2P) of 0.9652 and a root mean square error for prediction (RMSEP) of 0.0166 mg/kg, the SVR model exhibited the best quantitative prediction capacity based on the characteristic wavelengths. Following this, a visual distribution map was created to evaluate the lycopene content in tomato fruit intuitively. The results demonstrated the viability of hyperspectral imaging for detecting tomato maturity and quantitatively predicting the lycopene content during storage.

Deep Learning-Based Intelligent Apple Variety Classification System and Model Interpretability Analysis

In this study, series networks (AlexNet and VGG-19) and directed acyclic graph (DAG) networks (ResNet-18, ResNet-50, and ResNet-101) with transfer learning were employed to identify and classify 13 classes of apples from 7439 images. Two training datasets, model evaluation metrics, and three visualization methods were used to objectively assess, compare, and interpret five Convolutional Neural Network (CNN)-based models. The results show that the dataset configuration had a significant impact on the classification results, as all models achieved over 96.1% accuracy on dataset A (training-to-testing = 2.4:1.0) compared to 89.4-93.9% accuracy on dataset B (training-to-testing = 1.0:3.7). VGG-19 achieved the highest accuracy of 100.0% on dataset A and 93.9% on dataset B. Moreover, for networks of the same framework, the model size, accuracy, and training and testing times increased as the model depth (number of layers) increased. Furthermore, feature visualization, strongest activations, and local interpretable model-agnostic explanations techniques were used to show the understanding of apple images by different trained models, as well as to reveal how and why the models make classification decisions. These results improve the interpretability and credibility of CNN-based models, which provides guidance for future applications of deep learning methods in agriculture.

Discrimination of the Red Jujube Varieties Using a Portable NIR Spectrometer and Fuzzy Improved Linear Discriminant Analysis

In order to quickly, nondestructively, and effectively distinguish red jujube varieties, based on the combination of fuzzy theory and improved LDA (iLDA), fuzzy improved linear discriminant analysis (FiLDA) algorithm was proposed to classify near-infrared reflectance (NIR) spectra of red jujube samples. FiLDA shows performs better than iLDA in dealing with NIR spectra containing noise. Firstly, the portable NIR spectrometer was employed to gather the NIR spectra of five kinds of red jujube, and the initial NIR spectra were pretreated by standard normal variate transformation (SNV), multiplicative scatter correction (MSC), Savitzky-Golay smoothing (S-G smoothing), mean centering (MC) and Savitzky-Golay filter (S-G filter). Secondly, the high-dimensional spectra were processed for dimension reduction by principal component analysis (PCA). Then, linear discriminant analysis (LDA), iLDA and FiLDA were applied to extract features from the NIR spectra, respectively. Finally, K nearest neighbor (KNN) served as a classifier for the classification of red jujube samples. The highest classification accuracy of this identification system for red jujube, by using FiLDA and KNN, was 94.4%. These results indicated that FiLDA combined with NIR spectroscopy was an available method for identifying the red jujube varieties and this method has wide application prospects.