The feasibility of using hyperspectral imaging with convolutional neural network (CNN) to identify rice seed varieties was studied. Hyperspectral images of 4 rice seed varieties at two different spectral ranges (380–1030 nm and 874–1734 nm) were acquired. The spectral data at the ranges of 441–948 nm (Spectral range 1) and 975–1646 nm (Spectral range 2) were extracted. K nearest neighbors (KNN), support vector machine (SVM) and CNN models were built using different number of training samples (100, 200, 300, 400, 500, 600, 700, 800, 900, 1000, 1500, 2000, 2500 and 3000). KNN, SVM and CNN models in the Spectral range 2 performed slightly better than those in the Spectral range 1. The model performances improved with the increase in the number of training samples. The improvements were not significant when the number of training samples was large. CNN model performed better than the corresponding KNN and SVM models in most cases, which indicated the effectiveness of using CNN to analyze spectral data. The results of this study showed that CNN could be adopted in spectral data analysis with promising results. More varieties of rice need to be studied in future research to extend the use of CNNs in spectral data analysis.

There are possible environmental risks related to gene flow from genetically engineered organisms. It is important to find accurate, fast, and inexpensive methods to detect and monitor the presence of genetically modified (GM) organisms in crops and derived crop products. In the present study, GM maize kernels containing both cry1Ab/cry2Aj-G10evo proteins and their non-GM parents were examined by using hyperspectral imaging in the near-infrared (NIR) range (874.41–1733.91 nm) combined with chemometric data analysis. The hypercubes data were analyzed by applying principal component analysis (PCA) for exploratory purposes, and support vector machine (SVM) and partial least squares discriminant analysis (PLS–DA) to build the discriminant models to class the GM maize kernels from their contrast. The results indicate that clear differences between GM and non-GM maize kernels can be easily visualized with a nondestructive determination method developed in this study, and excellent classification could be achieved, with calculation and prediction accuracy of almost 100%. This study also demonstrates that SVM and PLS–DA models can obtain good performance with 54 wavelengths, selected by the competitive adaptive reweighted sampling method (CARS), making the classification processing for online application more rapid. Finally, GM maize kernels were visually identified on the prediction maps by predicting the features of each pixel on individual hyperspectral images. It was concluded that hyperspectral imaging together with chemometric data analysis is a promising technique to identify GM maize kernels, since it overcomes some disadvantages of the traditional analytical methods, such as complex and monotonous sampling.

Hyperspectral images in the spectral range of 874–1734 nm were collected for 14,015, 14,300 and 15,042 grape seeds of three varieties, respectively. Pixel-wise spectra were preprocessed by wavelet transform, and then, spectra of each single grape seed were extracted. Principal component analysis (PCA) was conducted on the hyperspectral images. Scores for images of the first six principal components (PCs) were used to qualitatively recognize the patterns among different varieties. Loadings of the first six PCs were used to identify the effective wavelengths (EWs). Support vector machine (SVM) was used to build the discriminant model using the spectra based on the EWs. The results indicated that the variety of each single grape seed was accurately identified with a calibration accuracy of 94.3% and a prediction accuracy of 88.7%. An external validation image of each variety was used to evaluate the proposed model and to form the classification maps where each single grape seed was explicitly identified as belonging to a distinct variety. The overall results indicated that a hyperspectral imaging (HSI) technique combined with multivariate analysis could be used as an effective tool for non-destructive and rapid variety discrimination and visualization of grape seeds. The proposed method showed great potential for developing a multi-spectral imaging system for practical application in the future.

The wine-making industry generates a considerable amount of grape pomace. Grape seeds, as an important part of pomace, are rich in bioactive compounds and can be reutilized to produce useful derivatives. The nutritional properties of grape seeds are largely influenced by the cultivar, which calls for effective identification. In the present work, the spectral profiles of grape seeds belonging to three different cultivars were collected by laser-induced breakdown spectroscopy (LIBS). Three conventional supervised classification methods and a deep learning method, a one-dimensional convolutional neural network (CNN), were applied to establish discriminant models to explore the relationship between spectral responses and cultivar information. Interval partial least squares (iPLS) algorithm was successfully used to extract the spectral region (402.74–426.87 nm) relevant for elemental composition in grape seeds. By comparing the discriminant models based on the full spectra and the selected spectral regions, the CNN model based on the full spectra achieved the optimal overall performance, with classification accuracy of 100% and 96.7% for the calibration and prediction sets, respectively. This work demonstrated the reliability of LIBS as a rapid and accurate approach for identifying grape seeds and will assist in the utilization of certain genotypes with desirable nutritional properties essential for production rather than their being discarded as waste.

Mildew damage is a major reason for chestnut poor quality and yield loss. In this study, a near-infrared hyperspectral imaging system in the 874–1734 nm spectral range was applied to detect the mildew damage to chestnuts caused by blue mold. Principal component analysis (PCA) scored images were firstly employed to qualitatively and intuitively distinguish moldy chestnuts from healthy chestnuts. Spectral data were extracted from the hyperspectral images. A successive projections algorithm (SPA) was used to select 12 optimal wavelengths. Artificial neural networks, including back propagation neural network (BPNN), evolutionary neural network (ENN), extreme learning machine (ELM), general regression neural network (GRNN) and radial basis neural network (RBNN) were used to build models using the full spectra and optimal wavelengths to distinguish moldy chestnuts. BPNN and ENN models using full spectra and optimal wavelengths obtained satisfactory performances, with classification accuracies all surpassing 99%. The results indicate the potential for the rapid and non-destructive detection of moldy chestnuts by hyperspectral imaging, which would help to develop online detection system for healthy and blue mold infected chestnuts.

Unmanned aerial vehicles (UAV) are widely used as remote sensing platforms to effectively monitor agricultural conditions. The wind field generated by the rotors in low-altitude operations will cause the deformation of rice crops, and may affect the acquisition of the true spectral information. In this study, a low-altitude UAV remote sensing simulation platform and a triple-direction wind field wireless sensor network system were built to explore the wind field distribution law. Combined with the multi-spectral images of the rice canopy, the influence of wind field on the spectral information acquisition was analyzed through variance and regression analysis. The results showed that the Z-direction wind field of UAV rotors dominated along three directions (X, Y, and Z). The coefficient of determination (R²) of three linear regression models for Normalized Difference Vegetation Index (NDVI), Ratio Vegetation Index (RVI), and Canopy Coverage Rate (CCR) was 0.782, 0.749, and 0.527, respectively. Therefore, the multi-rotor UAV wind field had an impact on the spectral information acquisition of rice canopy, and this influence could eventually affect the assessment of rice growth status. The models established in this study could provide a reference for the revised model of spectral indices, and offer guidance for the actual operations of low-altitude multi-rotor UAV.