The quality and price of navel oranges vary depending on their geographical origin, thus providing a financial incentive for origin fraud. To prevent this phenomenon, it is necessary to explore a fast, non-destructive, and precise method for tracing the origin of navel oranges. In this study, a total of 490 Newhall navel oranges were selected from five major production regions in China, and the diffuse reflectance near-infrared spectrum in 4000–10,000 cm⁻¹ were non-invasively collected. We examined seven preprocessing techniques for the spectra, including Savitzky–Golay (SG) smoothing, first derivative (FD), multiplicative scattering correction (MSC), combinations of SG with MSC (SG+MSC), SG with FD (SG+FD), MSC with FD (MSC+FD), and three combined (SG+MSC+FD). A one-dimensional convolutional neural network (1DCNN) deep learning model for geographical origin tracing of navel orange was established, and five machine learning algorithms, i.e., partial least squares discriminant analysis (PLS-DA), linear discriminant analysis (LDA), support vector machine (SVM), random forest (RF), and back-propagation neural network (BPNN), were compared with 1DCNN. The results show that the 1DCNN model based on the SG+FD preprocessing method achieved the optimal performance for the testing set, with prediction accuracy, precision, recall, and F1-score of 97.92%, 98%, 97.95%, and 97.90%, respectively. Therefore, NIRS combined with deep learning has a significant research and application value in the rapid, nondestructive, and accurate geographical origin traceability of agricultural products.

To noninvasively and precisely discriminate among the growth stages of multiple cultivars of navel oranges simultaneously, the fusion of the technologies of near-infrared (NIR) hyperspectral imaging (HSI) combined with machine vision (MV) and deep learning is employed. NIR reflectance spectra and hyperspectral and RGB images for 740 Gannan navel oranges of five cultivars are collected. Based on preprocessed spectra, optimally selected hyperspectral images, and registered RGB images, a dual-branch multi-modal feature fusion convolutional neural network (CNN) model is established. In this model, a spectral branch is designed to extract spectral features reflecting internal compositional variations, while the image branch is utilized to extract external color and texture features from the integration of hyperspectral and RGB images. Finally, growth stages are determined via the fusion of features. To validate the availability of the proposed method, various machine-learning and deep-learning models are compared for single-modal and multi-modal data. The results demonstrate that multi-modal feature fusion of HSI and MV combined with the constructed dual-branch CNN deep-learning model yields excellent growth stage discrimination in navel oranges, achieving an accuracy, recall rate, precision, F1 score, and kappa coefficient on the testing set are 95.95%, 96.66%, 96.76%, 96.69%, and 0.9481, respectively, providing a prominent way to precisely monitor the growth stages of fruits.