Variety identification of seeds is critical for assessing variety purity and ensuring crop yield. In this paper, a novel method based on hyperspectral imaging (HSI) and deep convolutional neural network (DCNN) was proposed to discriminate the varieties of oat seeds. The representation ability of DCNN was also investigated. The hyperspectral images with a spectral range of 874–1734 nm were primarily processed by principal component analysis (PCA) for exploratory visual distinguishing. Then a DCNN trained in an end-to-end manner was developed. The deep spectral features automatically learnt by DCNN were extracted and combined with traditional classifiers (logistic regression (LR), support vector machine with RBF kernel (RBF_SVM) and linear kernel (LINEAR_SVM)) to construct discriminant models. Contrast models were built based on the traditional classifiers using full wavelengths and optimal wavelengths selected by the second derivative (2nd derivative) method. The comparison results showed that all DCNN-based models outperformed the contrast models. DCNN trained in an end-to-end manner achieved the highest accuracy of 99.19% on the testing set, which was finally employed to visualize the variety classification. The results demonstrated that the deep spectral features with outstanding representation ability enabled HSI together with DCNN to be a reliable tool for rapid and accurate variety identification, which would help to develop an on-line system for quality detection of oat seeds as well as other grain seeds.

Variety classification is an important step in seed quality testing. This study introduces t-distributed stochastic neighbourhood embedding (t-SNE), a manifold learning algorithm, into the field of hyperspectral imaging (HSI) and proposes a method for classifying seed varieties. Images of 800 maize kernels of eight varieties (100 kernels per variety, 50 kernels for each side of the seed) were imaged in the visible- near infrared (386.7–1016.7 nm) wavelength range. The images were pre-processed by Procrustes analysis (PA) to improve the classification accuracy, and then these data were reduced to low-dimensional space using t-SNE. Finally, Fisher’s discriminant analysis (FDA) was used for classification of the low-dimensional data. To compare the effect of t-SNE, principal component analysis (PCA), kernel principal component analysis (KPCA) and locally linear embedding (LLE) were used as comparative methods in this study, and the results demonstrated that the t-SNE model with PA pre-processing has obtained better classification results. The highest classification accuracy of the t-SNE model was up to 97.5%, which was much more satisfactory than the results of the other models (up to 75% for PCA, 85% for KPCA, 76.25% for LLE). The overall results indicated that the t-SNE model with PA pre-processing can be used for variety classification of waxy maize seeds and be considered as a new method for hyperspectral image analysis.

This paper proposes the use of Near Infrared Hyperspectral Imaging (NIR-HSI) as a new strategy for fast and non-destructive classification of cotton seeds with respect to variety. A total of 807 seeds of four different cotton varieties are employed in this study. For classification purposes, each seed is represented by an average spectrum obtained by coaveraging the pixel spectra of the NIR-HSI image. Conventional NIR and VIS-NIR spectra are also employed for comparison. By using Partial-Least-Squares Discriminant Analysis (PLS-DA), correct classification rates of 98.0%, 89.7% and 91.7% were achieved in the NIR-HSI, conventional NIR and conventional VIS-NIR datasets. The superiority of the NIR-HSI system can be ascribed to a more comprehensive scan of the seed area, as compared to the conventional VIS-NIR spectrometer.

This paper presents a system for automated classification of rice variety for rice seed production using computer vision and image processing techniques. Rice seeds of different varieties are visually very similar in color, shape and texture that make the classification of rice seed varieties at high accuracy challenging. We investigated various feature extraction techniques for efficient rice seed image representation. We analyzed the performance of powerful classifiers on the extracted features for finding the robust one. Images of six different rice seed varieties in northern Vietnam were acquired and analyzed. Our experiments have demonstrated that the average accuracy of our classification system can reach 90.54% using Random Forest method with a simple feature extraction technique. This result can be used for developing a computer-aided machine vision system for automated assessment of rice seeds purity.

This paper investigated the use of hyperspectral imaging (HSI) to discriminate the variety and quality of rice. Hyperspectral images (400–1,000 nm) of paddy rice samples were acquired to extract both spectral and image information. Dimension reduction was carried out on the region of interest (ROI) of the images by principal component analysis (PCA). The first principal components (PCs) explained over 98 % of variances of all spectral bands. Chalkiness degree and shape feature (‘MajorAxisLength’, ‘MinorAxisLength’, length-width ratio, ‘Perimeter’ and ‘Eccentricity’) were further extracted and used for subsequent rice variety discrimination by PCA and back propagation neural network (BPNN). An integration of spectral and image data was used for BPNN classification. The BPNN model based on spectral data (seven optimal wavelengths) achieved better results than PCA based on spectral data (seven optimal wavelengths) in variety discrimination with classification accuracy of 89.18 and 89.91 % for PCA and BPNN model, respectively. The BPNN model based on data fusion achieved the best results (94.45 %), which was superior to the results based on spectral data (89.91 %) or image data (88.09 %) alone. Finally, the resulting classification maps were able to visualize different rice varieties. The results demonstrated that discrimination of rice variety and quality with HSI technology was feasible and could be utilized for quality control purposes and/or for innovative sorting of rice.

This current study was carried out to investigate the ability of hyperspectral imaging (HSI) technique and multivariate classification for the differentiation of lychee varieties. A total of 122 lychee samples from three varieties (“Baila,” “Jizhui,” and “Guiwei”) were used. The relationship between reflectance spectra and lychee varieties were established. Principal component analysis (PCA) was implemented on the region of interest (ROI) image to reduce data dimensionality and visualize the cluster trend. The first two principal components (PCs) explained over 97 % of variances of all spectral bands. Linear (soft independent modeling of class analogy (SIMCA) and partial least square discriminant analysis (PLS-DA)) and nonlinear (back propagation neural network (BPNN) and support vector machine (SVM)) multivariate classification methods were used to develop discrimination models. The results revealed that SVM model achieved the best result, with the identification rate of 100 % in the calibration set and 87.81 % in the prediction set. BPNN had a discrimination rate of 100 % for the training set and 85.37 % for prediction set, while PSL-DA and SIMCA model had a discrimination rate of 78.05 % and 60.98 % for prediction sets, respectively. The nonlinear classification methods used were superior to the linear ones. The overall results showed that HSI system with SVM classification tool could be used in identification of lychee varieties.

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.

The purpose of this study was to highlight the use of multispectral imaging in seed quality testing of castor seeds. Visually, 120 seeds were divided into three classes: yellow, grey and black seeds. Thereafter, images at 19 different wavelengths ranging from 375–970 nm were captured of all the seeds. Mean intensity for each single seed was extracted from the images, and a significant difference between the three colour classes was observed, with the best separation in the near-infrared wavelengths. A specified feature (RegionMSI mean) based on normalized canonical discriminant analysis, were employed and viable seeds were distinguished from dead seeds with 92% accuracy. The same model was tested on a validation set of seeds. These seeds were divided into two groups depending on germination ability, 241 were predicted as viable and expected to germinate and 59 were predicted as dead or non-germinated seeds. This validation of the model resulted in 96% correct classification of the seeds. The results illustrate how multispectral imaging technology can be employed for prediction of viable castor seeds, based on seed coat colour.

The pre-harvest sprouting of wheat have significant influence for its quality and yield, therefore the fast detection of sprouting extent of wheat is very important for breeding and producing. In this study, the hyperspectral images of these seeds were collected by a near infrared hyperspectral imaging system, the wavelength of which was 850-1700 nm after wheat germination experiment at 0h, 12h, 24h, and 48h. The original light intensity of embryo and endosperm were extracted, and were then changed to reflectivity for later analysis. The image and spectral information of wheat with different parts, different varieties and different sprouting extent were compared. The results showed that after 12h sprouting, the reflectivity of embryo was lower than that of endosperm for the same seed, this is mainly due to the water and fat content of embryo was higher than the endosperm portions. For the same varieties of wheat seed at the germination of 12h, 24h and 48h,in the wavelength range of 870-1300 nm, the reflectivity increased with the increase of sprouting time, it was related to the changes of its internal content of fat in the seed germination process. At 1400nm, the reflectivity of sprouted wheat seeds were all lower than that of dry seeds, it was related to the rise of internal water content in the process of seed germination. Due to differences in seed water absorption and sprouting resistance, for different varieties of wheat seeds, its spectral characteristics are also different. The presented indicated that hyperspectral imaging could reflect the characteristics of sprouted wheat seeds, which provide some basis for explore the sprouting index by hyperspectral imaging.

The objectives of this research are to present the automatic organization of agricultural seeds with the explosion of digital information through compute image vision processing. In this research paper CVIP (computer vision image processing) tool has been applied on different varieties and categorized of canola seeds. We had the 4 different varieties of canola seeds which were named as Gobhi Sarson (A), Barassica comp (B), Sathri (C) and Rocket Herbof (D). Each variety had 10 images and total 10*4 =40 images of canola seeds. We took the train data results of all kinds of canola seeds. After that the train data results were compared for pattern classification and apply the nearest neighbor and k-nearest neighbor algorithms for final classification. We achieved in nearest neighbor 85% and 76% average and k-nearest neighbor 90% and 73% average as a final pattern classification results. These are the best percentage for classification and provide more accuracy. The formers can select best healthy seed variety with the help of the results of this research. Keywords:  Features, Pattern classification, nearest neighbor, k-nearest neighbor

The internal quality of nectarines (Prunus persica L. Batsch var. nucipersica) cv. ‘Big Top’ (yellow flesh) and ‘Magique’ (white flesh) has been inspected using hyperspectral transmittance imaging. Hyperspectral images of intact fruits were acquired in the spectral range of 630–900 nm using transmittance mode during their ripening under controlled conditions. The detection of split pit disorder and classification according to an established firmness threshold were performed using PLS-DA. The prediction of the Internal Quality Index (IQI) related to ripeness was performed using PLS-R. The most important variables were selected using interval-PLS. As a result, an accuracy of 94.7% was obtained in the detection of fruits with split pit of the ‘Big Top’ cultivar. Accuracies of 95.7% and 94.6% were achieved in the classification of the ‘Big Top’ and ‘Magique’ cultivars, respectively, according to the firmness threshold. The internal quality was predicted through the IQI with R2 values of 0.88 and 0.86 for the two cultivars. The results obtained indicate the great potential of hyperspectral transmittance imaging for the assessment of the internal quality of intact nectarines.

The increasing market interest in coffee beverage, lead coffee growers around the world to adopt more efficient methods to select the best-quality coffee beans. Currently, coffee beans selection is carried out either manually, which is a costly and unreliable process, or using electronic sorting machines, which are often inefficient because some coffee beans defects, such as sour and immature beans, have similar spectral response patterns. In this sense, the present work aimed to assess coffee beans quality using both computer vision and machine learning techniques, such as Support Vector Machine (SVM), Deep Neural Network (DNN) and Random Forest (RF). For this purpose, an algorithm written in Python language was developed to extract shape and color features from coffee beans images. The obtained dataset was then used as input to the machine learning algorithms. The data reported in this study pointed to the importance of color descriptors for classifying coffee beans defects. Among the variables used, the components from RGB (Red, Green and Blue) and HSV (Hue, Saturation and Value) color spaces presented the most relevant contribution for the classification models. Also, the results reported in this study provides evidence that computer vision along with machine learning algorithms can be used to identify and classify coffee beans with a very high accuracy (> 90%). Key words: Deep neural network; classification; artificial intelligence; image processing; granulometry.

The identification of different grapevine varieties, currently attended using visual ampelometry, DNA analysis and very recently, by hyperspectral analysis under laboratory conditions, is an issue of great importance in the wine industry. This work presents support vector machine and artificial neural network’s modelling for grapevine varietal classification from in-field leaf spectroscopy. Modelling was attempted at two scales: site-specific and a global scale. Spectral measurements were obtained on the near-infrared (NIR) spectral range between 1600 to 2400 nm under field conditions in a non-destructive way using a portable spectrophotometer. For the site specific approach, spectra were collected from the adaxial side of 400 individual leaves of 20 grapevine (Vitis vinifera L.) varieties one week after veraison. For the global model, two additional sets of spectra were collected one week before harvest from two different vineyards in another vintage, each one consisting on 48 measurement from individual leaves of six varieties. Several combinations of spectra scatter correction and smoothing filtering were studied. For the training of the models, support vector machines and artificial neural networks were employed using the pre-processed spectra as input and the varieties as the classes of the models. The results from the pre-processing study showed that there was no influence whether using scatter correction or not. Also, a second-degree derivative with a window size of 5 Savitzky-Golay filtering yielded the highest outcomes. For the site-specific model, with 20 classes, the best results from the classifiers thrown an overall score of 87.25% of correctly classified samples. These results were compared under the same conditions with a model trained using partial least squares discriminant analysis, which showed a worse performance in every case. For the global model, a 6-class dataset involving samples from three different vineyards, two years and leaves monitored at post-veraison and harvest was also built up, reaching a 77.08% of correctly classified samples. The outcomes obtained demonstrate the capability of using a reliable method for fast, in-field, non-destructive grapevine varietal classification that could be very useful in viticulture and wine industry, either global or site-specific.

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.

The feasibility of a color machine vision technique with the one-class classification method was investigated for the quality assessment of tomato seeds. The health of seeds is an important quality factor that affects their germination rate, which may be affected by seed contamination. Hence, segregation of healthy seeds from diseased and infected seeds, along with foreign materials and broken seeds, is important to improve the final yield. In this study, a custom-built machine vision system containing a color camera with a white light emitting diode (LED) light source was adopted for image acquisition. The one-class classification method was used to identify healthy seeds after extracting the features of the samples. A significant difference was observed between the features of healthy and infected seeds, and foreign materials, implying a certain threshold. The results indicated that tomato seeds can be classified with an accuracy exceeding 97%. The infected tomato seeds indicated a lower germination rate (<10%) compared to healthy seeds, as confirmed by the organic growing media germination test. Thus, identification through image analysis and rapid measurement were observed as useful in discriminating between the quality of tomato seeds in real time.

The expansion of the global rice marketplace ultimately raises concerns about authenticity control. Several analytical methods for differentiating the geographical origin of rice have been developed, yet a high-throughput method is still in demand. In this study, we developed a rapid approach using direct infusion-mass spectrometry (DI-MS) to distinguish rice products from different countries. Specifically, the elimination of the matrix effect by a polytetrafluoroethylene (PTFE) filter, a mixed-mode cation exchange (MCX) solid-phase extraction (SPE) with 20% methanol, and an MCX SPE with 100% methanol were measured. Afterward, partial least squares discriminant analysis and random forests were applied to seek the optimal discrimination method. The results revealed that the combination of MCX SPE with 100% methanol and DI-MS in positive ion mode (accuracy=1.000, R2=0.916, Q2=0.720, B/W-based p-value=0.015) or the combination of MCX SPE with 20% methanol and targeted DI-MS/MS in positive ion mode (accuracy=1.000, R2=0.931, Q2=0.849, B/W-based p-value=0.002) showed the excellent discriminatory ability. Furthermore, differentially expressed metabolites including sodiated lysophosphatidylcholine, lysophosphatidylcholine, lysophosphatidylethanolamines and lysophosphatidylglycerol classes were found. In conclusion, our study provides a rapid and reliable platform for geographical discrimination of white rice and will contribute to the authenticity control of rice products.

The area of remote sensing techniques in agriculture has reached a significant degree of development and maturity, with numerous journals, conferences, and organizations specialized in it. Moreover, many review papers are available in the literature. The present work describes a literature review that adopts the form of a systematic mapping study, following a formal methodology. Eight mapping questions were defined, analyzing the main types of research, techniques, platforms, topics, and spectral information. A predefined search string was applied in the Scopus database, obtaining 1590 candidate papers. Afterwards, the most relevant 106 papers were selected, considering those with more than six citations per year. These are analyzed in more detail, answering the mapping questions for each paper. In this way, the current trends and new opportunities are discovered. As a result, increasing interest in the area has been observed since 2000; the most frequently addressed problems are those related to parameter estimation, growth vigor, and water usage, using classification techniques, that are mostly applied on RGB and hyperspectral images, captured from drones and satellites. A general recommendation that emerges from this study is to build on existing resources, such as agricultural image datasets, public satellite imagery, and deep learning toolkits.

The aim of this work was to analyze the variety of soil by laser-induced breakdown spectroscopy (LIBS) coupled with chemometrics methods. 6 certified reference materials (CRMs) of soil samples were selected and their LIBS spectra were captured. Characteristic emission lines of main elements were identified based on the LIBS curves and corresponding contents. From the identified emission lines, LIBS spectra in 7 lines with high signal-to-noise ratio (SNR) were chosen for further analysis. Principal component analysis (PCA) was carried out using the LIBS spectra at 7 selected lines and an obvious cluster of 6 soils was observed. Soft independent modeling of class analogy (SIMCA) and least-squares support vector machine (LS-SVM) were introduced to establish discriminant models for classifying the 6 types of soils, and they offered the correct discrimination rates of 90% and 100%, respectively. Receiver operating characteristic (ROC) curve was used to evaluate the performance of models and the results demonstrated that the LS-SVM model was promising. Lastly, 8 types of soils from different places were gathered to conduct the same experiments for verifying the selected 7 emission lines and LS-SVM model. The research revealed that LIBS technology coupled with chemometrics could conduct the variety discrimination of soil.

The adulteration of rice in the food industry is a very serious problem nowadays. To realize the rapid and stable identification of adulterated Wuchang rice, a hyperspectral imaging system (380–1000 nm) has been introduced in this study. Piece-wise multiplicative scatter correction (PMSC) was first used to correct the non-linear additive and multiplicative scatter effects. Then, the adulterated rice samples were identified via support vector machines (SVM). The PMSC-SVM model was attained over the whole spectral range, with the correct classification rate (CCR) increased from 95.47% to 99.20%, the kappa coefficient increased from 0.95 to 0.99, and the prediction CV (coefficient of variation) decreased from 3.04% to 1.56%. Furthermore, a simplified PMSC-SVM model was established, where 13 principal components were selected using 5-fold cross-validation. The CCR was increased from 95.40% to 99.08%, the kappa coefficient was increased from 0.94 to 0.99, and the prediction CV was decreased from 3.02% to 1.72%. The results demonstrated that the accuracy and stability for identifying adulterated rice has been improved by PMSC in the hyperspectral imaging system.

The adaptation and use of advanced technologies is an effective and encouraging way to efficiently and reliably characterise crops and plants. Additionally advances in these technologies will improve the information available for agronomists, breeders and plant physiologists in order to develop best management practices in the process and commercialization of agricultural products and commodities. Methods based on vibrational spectroscopy such as near infrared (NIR) spectroscopy using either single spot or hyperspectral measurements are now more available and ready to use than ever before. The main characteristics of these methodologies (high-throughput, non-destructive) have determined a growth in basic and applied research using NIR spectroscopy in many disciplines related with crop and plant sciences. A wide range of studies have demonstrated the ability of NIR spectroscopy to analyse different parameters in crops. Recently the use of hyperspectral imaging techniques have expanded the range of applications in crop and plant sciences. This article provides an overview of applications and developments of NIR hyperspectral image for the analysis, monitoring and characterisation of crops and plants.