Optical sensors combined with machine learning algorithms have led to significant advances in seed science. These advances have facilitated the development of robust approaches, providing decision-making support in the seed industry related to the marketing of seed lots. In this study, a novel approach for seed quality classification is presented. We developed classifier models using Fourier transform near-infrared (FT-NIR) spectroscopy and X-ray imaging techniques to predict seed germination and vigor. A forage grass (Urochloa brizantha) was used as a model species. FT-NIR spectroscopy data and radiographic images were obtained from individual seeds, and the models were created based on the following algorithms: linear discriminant analysis (LDA), partial least squares discriminant analysis (PLS-DA), random forest (RF), naive Bayes (NB), and support vector machine with radial basis (SVM-r) kernel. In the germination prediction, the models individually reached an accuracy of 82% using FT-NIR data, and 90% using X-ray data. For seed vigor, the models achieved 61% and 68% accuracy using FT-NIR and X-ray data, respectively. Combining the FT-NIR and X-ray data, the performance of the classification model reached an accuracy of 85% to predict germination, and 62% for seed vigor. Overall, the models developed using both NIR spectra and X-ray imaging data in machine learning algorithms are efficient in quickly, non-destructively, and accurately identifying the capacity of seed to germinate. The use of X-ray data and the LDA algorithm showed great potential to be used as a viable alternative to assist in the quality classification of U. brizantha seeds.

New computer vision solutions combined with artificial intelligence algorithms can help recognize patterns in biological images, reducing subjectivity and optimizing the analysis process. The aim of this study was to propose an approach based on interactive and traditional machine learning methods to classify soybean seeds and seedlings according to their appearance and physiological potential. In addition, we correlated the appearance of seeds to their physiological performance. Images of soybean seeds and seedlings were used to develop models using low-cost approaches and free-access software. The models developed showed high performance, with overall accuracy reaching 0.94 for seeds and seedling classification. The high precision of the models that were developed based on interactive and traditional machine learning demonstrated that the method can easily be used to classify soybean seeds according to their appearance, as well as to classify soybean seedling vigor quickly and non-subjectively. The appearance of soybean seeds is strongly correlated with their physiological performance.

Manual monitoring of animal behavior is time-consuming and prone to bias. An alternative to such limitations is using computational resources in behavioral assessments, such as tracking systems, to facilitate accurate and long-term evaluations. There is a demand for robust software that addresses analysis in heterogeneous environments (such as in field conditions) and evaluates multiple individuals in groups while maintaining their identities. The Ethoflow software was developed using computer vision and artificial intelligence (AI) tools to monitor various behavioral parameters automatically. An object detection algorithm based on instance segmentation was implemented, allowing behavior monitoring in the field under heterogeneous environments. Moreover, a convolutional neural network was implemented to assess complex behaviors expanding behavior analyses’ possibilities. The heuristics used to generate training data for the AI models automatically are described, and the models trained with these datasets exhibited high accuracy in detecting individuals in heterogeneous environments and assessing complex behavior. Ethoflow was employed for kinematic assessments and to detect trophallaxis in social bees. The software was developed in desktop applications and had a graphical user interface. In the Ethoflow algorithm, the processing with AI is separate from the other modules, facilitating measurements on an ordinary computer and complex behavior assessing on machines with graphics processing units. Ethoflow is a useful support tool for applications in biology and related fields.

Background: The use of non-destructive methods with less human interference is of great interest in agricultural industry and crop breeding. Modern imaging technologies enable the automatic visualization of multi-parameter for characterization of biological samples, reducing subjectivity and optimizing the analysis process. Furthermore, the combination of two or more imaging techniques has contributed to discovering new physicochemical tools and interpreting datasets in real time. Results: We present a new method for automatic characterization of seed quality based on the combination of multispectral and X-ray imaging technologies. We proposed an approach using X-ray images to investigate internal tissues because seed surface profile can be negatively affected, but without reaching important internal regions of seeds. An oilseed plant (Jatropha curcas) was used as a model species, which also serves as a multi-purposed crop of economic importance worldwide. Our studies included the application of a normalized canonical discriminant analyses (nCDA) algorithm as a supervised transformation building method to obtain spatial and spectral patterns on different seedlots. We developed classification models using reflectance data and X-ray classes based on linear discriminant analysis (LDA). The classification models, individually or combined, showed high accuracy (> 0.96) using reflectance at 940 nm and X-ray data to predict quality traits such as normal seedlings, abnormal seedlings and dead seeds. Conclusions: Multispectral and X-ray imaging have a strong relationship with seed physiological performance. Reflectance at 940 nm and X-ray data can efficiently predict seed quality attributes. These techniques can be alternative methods for rapid, efficient, sustainable and non-destructive characterization of seed quality in the future, overcoming the intrinsic subjectivity of the conventional seed quality analysis.

Light-based methods are being further developed to meet the growing demands for food in the agricultural industry. Optical imaging is a rapid, non-destructive, and accurate technology that can produce consistent measurements of product quality compared to conventional techniques. In this research, a novel approach for seed quality prediction is presented. In the proposed approach two advanced optical imaging techniques based on chlorophyll fluorescence and chemometric-based multispectral imaging were employed. The chemometrics encompassed principal component analysis (PCA) and quadratic discrimination analysis (QDA). Among plants that are relevant as both crops and scientific models, tomato, and carrot were selected for the experiment. We compared the optical imaging techniques to the traditional analytical methods used for quality characterization of commercial seedlots. Results showed that chlorophyll fluorescence-based technology is feasible to discriminate cultivars and to identify seedlots with lower physiological potential. The exploratory analysis of multispectral imaging data using a non-supervised approach (two-component PCA) allowed the characterization of differences between carrot cultivars, but not for tomato cultivars. A Random Forest (RF) classifier based on Gini importance was applied to multispectral data and it revealed the most meaningful bandwidths from 19 wavelengths for seed quality characterization. In order to validate the RF model, we selected the five most important wavelengths to be applied in a QDA-based model, and the model reached high accuracy to classify lots with high-and low-vigor seeds, with a correct classification from 86 to 95% in tomato and from 88 to 97% in carrot for validation set. Further analysis showed that low quality seeds resulted in seedlings with altered photosynthetic capacity and chlorophyll content. In conclusion, both chlorophyll fluorescence and chemometrics-based multispectral imaging can be applied as reliable proxies of the physiological potential in tomato and carrot seeds. From the practical point of view, such techniques/methodologies can be potentially used for screening low quality seeds in food and agricultural industries.