The incorporation of autonomous computation and artificial intelligence (AI) technologies into smart agriculture concepts is becoming an expected scientific procedure. The airborne hyperspectral system with its vast area coverage, high spectral resolution, and varied narrow-band selection is an excellent tool for crop physiological characteristics and yield prediction. However, the extensive and redundant three-dimensional (3D) cube data processing and computation have made the popularization of this tool a challenging task. This research integrated two important open-sourced systems (R and Python) combined with automated hyperspectral narrowband vegetation index calculation and the state-of-the-art AI-based automated machine learning (AutoML) technology to estimate yield and biomass, based on three crop categories (spring wheat, pea and oat mixture, and spring barley with red clover) with multifunctional cultivation practices in northern Europe and Estonia. Our study showed the estimated capacity of the empirical AutoML regression model was significant. The best coefficient of determination (R2) and normalized root mean square error (NRMSE) for single variety planting wheat were 0.96 and 0.12 respectively; for mixed peas and oats, they were 0.76 and 0.18 in the booting to heading stage, while for mixed legumes and spring barley, they were 0.88 and 0.16 in the reproductive growth stages. In terms of straw mass estimation, R2 was 0.96, 0.83, and 0.86, and NRMSE was 0.12, 0.24, and 0.33 respectively. This research contributes to, and confirms, the use of the AutoML framework in hyperspectral image analysis to increase implementation flexibility and reduce learning costs under a variety of agricultural resource conditions. It delivers expert yield and straw mass valuation two months in advance before harvest time for decision-makers. This study also highlights that the hyperspectral system provides economic and environmental benefits and will play a critical role in the construction of sustainable and intelligent agriculture techniques in the upcoming years.

The application of unmanned aerial systems (UAS) in forest research includes a wide range of equipment, systems, and flight settings, creating a need for enhancing data acquisition efficiency and quality. Thus, we assessed the effects of flying altitude and lateral and longitudinal overlaps on digital aerial photogrammetry (DAP) processing and the ability of its products to provide point clouds for forestry inventory. For this, we used 18 combinations of flight settings for data acquisition, and a nationwide airborne laser scanning (ALS) dataset as reference data. Linear regression was applied for modeling DAP quality indicators and model fitting quality as the function of flight settings; equivalence tests compared DAP- and ALS-products. Most of DAP-Digital Terrain Models (DTM) showed a moderate to high agreement (R2 > 0.70) when fitted to ALS-based models; nine models had a regression slope within the 1% region of equivalence. The best DAP-Canopy Height Model (CHM) was generated using ALS-DTM with an R2 = 0.42 when compared with ALS-CHM, indicating reduced similarity. Altogether, our results suggest that the optimal combination of flight settings should include a 90% lateral overlap, a 70% longitudinal overlap, and a minimum altitude of 120 m above ground level, independent of the availability of an ALS-derived DTM for height normalization. We also provided insights into the effects of flight settings on DAP outputs for future applications in similar forest stands, emphasizing the benefits of overlaps for comprehensive scene reconstruction and altitude for canopy surface detection.

A significant trend has developed with the recent growing interest in the estimation of aboveground biomass of vegetation in legume-supported systems in perennial or semi-natural grasslands to meet the demands of sustainable and precise agriculture. Unmanned aerial systems (UAS) are a powerful tool when it comes to supporting farm-scale phenotyping trials. In this study, we explored the variation of the red clover-grass mixture dry matter (DM) yields between temporal periods (one- and two-year cultivated), farming operations [soil tillage methods (STM), cultivation methods (CM), manure application (MA)] using three machine learning (ML) techniques [random forest regression (RFR), support vector regression (SVR), and artificial neural network (ANN)] and six multispectral vegetation indices (VIs) to predict DM yields. The ML evaluation results showed the best performance for ANN in the 11-day before harvest category (R² = 0.90, NRMSE = 0.12), followed by RFR (R² = 0.90 NRMSE = 0.15), and SVR (R² = 0.86, NRMSE = 0.16), which was furthermore supported by the leave-one-out cross-validation pre-analysis. In terms of VI performance, green normalized difference vegetation index (GNDVI), green difference vegetation index (GDVI), as well as modified simple ratio (MSR) performed better as predictors in ANN and RFR. However, the prediction ability of models was being influenced by farming operations. The stratified sampling, based on STM, had a better model performance than CM and MA. It is proposed that drone data collection was suggested to be optimum in this study, closer to the harvest date, but not later than the ageing stage.

No abstract

The recent trend of automated machine learning (AutoML) has been driving further significant technological innovation in the application of artificial intelligence from its automated algorithm selection and hyperparameter optimization of the deployable pipeline model for unraveling substance problems. However, a current knowledge gap lies in the integration of AutoML technology and unmanned aircraft systems (UAS) within image-based data classification tasks. Therefore, we employed a state-of-the-art (SOTA) and completely open-source AutoML framework, Auto-sklearn, which was constructed based on one of the most widely used ML systems: Scikit-learn. It was combined with two novel AutoML visualization tools to focus particularly on the recognition and adoption of UAS-derived multispectral vegetation indices (VI) data across a diverse range of agricultural management practices (AMP). These include soil tillage methods (STM), cultivation methods (CM), and manure application (MA), and are under the four-crop combination fields (i.e., red clover-grass mixture, spring wheat, pea-oat mixture, and spring barley). Furthermore, they have currently not been efficiently examined and accessible parameters in UAS applications are absent for them. We conducted the comparison of AutoML performance using three other common machine learning classifiers, namely Random Forest (RF), support vector machine (SVM), and artificial neural network (ANN). The results showed AutoML achieved the highest overall classification accuracy numbers after 1200 s of calculation. RF yielded the second-best classification accuracy, and SVM and ANN were revealed to be less capable among some of the given datasets. Regarding the classification of AMPs, the best recognized period for data capture occurred in the crop vegetative growth stage (in May). The results demonstrated that CM yielded the best performance in terms of classification, followed by MA and STM. Our framework presents new insights into plant–environment interactions with capable classification capabilities. It further illustrated the automatic system would become an important tool in furthering the understanding for future sustainable smart farming and field-based crop phenotyping research across a diverse range of agricultural environmental assessment and management applications.