Precision Agriculture最新文献

Purpose

The hull scrape and vegetation indices are widely used for predicting peanut maturation, but they are time-consuming, subjective, labor-intensive, and fail to account for climate variables, reducing their accuracy.Thus, the objective was to verify the potential of using artificial intelligence associating IV and climate variables to predict the variability of peanut pod maturity in the field

Methods

For this purpose, peanut maturity data collected on different dates in commercial fields in Brazil and the United States. In addition, high-resolution satellite images were used to calculate nine IV and four climatic variables for each area were acquired using the NASA-POWER platform. Four machine learning models were tested and the input for the training were selected using the Random Forest feature selection. Thus, the models were trained using 70% of the data for training and 30% for testing and applied the cross validation with K-fold.

Results

The best results were obtained for the XGBoosting model with R² test values varying 0.90, 0.89, 0.93 and 0.87 and a minimum MAE and RMSE of 0.05. Except for the Georgia dataset where the MLP model presents the highest performance R² value of 0.93, MAE 0.05 and RMSE 0.06 for the test. The RBF models present the worst results with a low index of agreement (d) 0.4 for all the datasets demonstrating a low agreement between the predicted and observed values.

Conclusion

Combining the climatic variables was able to improve the model’s performance, however detailed information about the field such as topographic conditions and soil type seem to be a different approach to enhance the model performance. Using the calibrated model for overall dataset peanut farmers from any localities can use to monitor and map the PMI variability in the fields, improve the decision-making, decrease the loss and increase the kernels quality.

Purpose

Precision agriculture technologies play an important role in optimising practices to increase yields and reduce costs, contributing to socio-economic progress and environmental well-being, and playing a key role in addressing climate change. Viticulture is a strategic, input-intensive agricultural sector where precision technologies can make the use of resources more efficient without compromising profitability. The aim of this study is to evaluate the profitability of implementing precision farming systems, such as unmanned aerial vehicle surveying for the production of vigour maps, compared to the conventional cultivation system in two Italian wineries.

Methods

The profitability of using precision farming tools in viticulture compared to conventional management techniques has been investigated in two Italian wineries over a four-year period, before and after the introduction of UAV technology.

Results

The results demonstrate the usefulness and economic viability of precision agriculture technologies in viticulture. The vigour maps produced by the data collected with UAV technology allow both the identification of problems such as diseases, and consequently the planning of phytosanitary treatments, and selective grape harvesting, which allows a significant improvement in the quality of the harvested grapes.

Conclusion

The results demonstrate the usefulness of precision technologies for cost-effective and sustainable vineyard management, satisfying a market segment made up of stakeholders who are increasingly sensitive to environmental issues.

Although unmanned aerial vehicle (UAV) remote sensing is widely used for high-throughput crop monitoring, few attempts have been made to assess nitrogen content (NC) at the organ level and its association with nitrogen use efficiency (NUE). Also, little is known about the performance of UAV-based image texture features of different spectral bands in monitoring crop nitrogen and NUE. In this study, multi-spectral images were collected throughout different stages of winter wheat in two independent field trials - a single-variety field trial and a multi-variety trial in 2021 and 2022, respectively in China and Germany. Forty-three multispectral vegetation indices (VIs) and forty texture features (TFs) were calculated from images and fed into the partial least squares regression (PLSR) and random forest (RF) regression models for predicting nitrogen-related indicators. Our main objectives were to (1) assess the potential of UAV-based multispectral imagery for predicting NC in different organs of winter wheat, (2) explore the transferability of different image features (VI and TF) and trained machine learning models in predicting NC, and (3) propose a technical workflow for mapping NUE using UAV imagery. The results showed that the correlation between different features (VIs and TFs) and NC in different organs varied between the pre-anthesis and post-anthesis stages. PLSR latent variables extracted from those VIs and TFs could be a great predictor for nitrogen agronomic efficiency (NAE). While adding TFs to VI-based models enhanced the model performance in predicting NC, inconsistency arose when applying the TF-based models trained based on one dataset to the other independent dataset that involved different varieties, UAVs, and cameras. Unsurprisingly, models trained with the multi-variety dataset show better transferability than the models trained with the single-variety dataset. This study not only demonstrates the promise of applying UAV-based imaging to estimate NC in different organs and map NUE in winter wheat but also highlights the importance of conducting model evaluations based on independent datasets.

Through targeted spray applications, precision agriculture can provide not only environmental benefits but also lower production costs, improving farm competitiveness. Nevertheless, few studies have focused on the cost-effectiveness of precision agriculture sprayers in vineyards, which are among the most widespread specialty crops. Therefore, this is the first study that aims to evaluate the cost-effectiveness of variable rate technology (VRT) and unmanned aerial vehicle (UAV) sprayers compared to a conventional sprayer in a hypothetical and representative vineyard area of southern Italy. The economic analysis, based on technological parameters in the literature, enabled the identification of the minimum farm size (break-even point) for introducing precision agriculture sprayers (PAS), considering the annual cost of the pesticide treatments (equipment and pesticide costs). Our findings revealed that the UAV sprayer—if permitted by law—could be the most convenient option for farms larger than 2.27 ha, whereas the VRT sprayer should be chosen by farms over 17.02 ha. However, public subsidies, such as those provided by the Italian Recovery Plan, make adopting VRT sprayers also economically viable for areas as small as 3.03 ha. Finally, the sensitivity analysis confirmed that the purchase price and pesticide cost are the most sensitive parameters affecting the break-even points. Our findings shed light on the economic sustainability of these innovative sprayers, a key driver for their adoption by farmers and for setting future strategies for facing the current agricultural crisis.

Purpose

Creating spatial weed distribution maps as the basis for site-specific weed management (SSWM) requires determining the occurrence and densities of weeds at georeferenced grid points. To achieve a field-wide distribution map, the weed distribution between the sampling points needs to be predicted. The aim of this study was to determine the best combination of grid sampling design and spatial interpolation technique to improve prediction accuracy. Gaussian copula as alternative method was tested to overcome challenges associated with interpolating weed densities such as smoothing effects.

Methods

The quality of weed distribution maps created using combinations of different sampling grids and interpolation methods was assessed: Inverse Distance Weighting, different geostatistical approaches, and Nearest Neighbor Interpolation. For this comparison, the weed distribution and densities in four fields were assessed using three sampling grids with different resolutions and arrangements: Random vs. regular arrangement of 40 grid points, and a combination of both grid types (fine grid).

Results

The best prediction of weed distribution was achieved with the Kriging interpolation models based on weed data sampled on the fine grid. In contrast, the lowest performance was observed using the regular grid and the Nearest Neighbor Interpolation. A patchy distribution of weeds did not affect the prediction quality.

Conclusion

Using the Gaussian copula kriging did not result in a reduction of the smoothing effect, which still represents a challenge when employing spatial interpolation methods for SSWM. However, using a randomly distributed raster with a fine resolution could further optimize the precision of weed distribution maps.

Spring dead spot is a disease of bermudagrass (Cynodon dactylon L. Pers) caused by Ophiosphaerella spp., of fungi which infect the below ground structures of plants, causing damage to the turf canopy. Previous research suggests that precision management strategies based on manually identified disease within unmanned aerial vehicle (UAV) imagery using GIS software and global navigation satellite systems (GNSS)-equipped sprayers can reduce the fungicide required for spring dead spot management. However, this methodology is time consuming and impractical for golf course superintendents. This paper introduces a novel approach to spring dead spot identification utilizing a custom Python script, the Simple Ophiosphaerella Damage Detector (SODD), to identify and record locations of spring dead spot from UAV imagery using basic feature extraction techniques. Initial tests comparing the outputs from SODD to spring dead spot manually identified by researchers on four fairways, comparisons of K-means cluster maps showed similarities ranging between 71 and 88% although incidence counts were inconsistent. Precision treatment methods based on SODD were evaluated across 16 golf course fairways at three locations in Virginia organized as a randomized complete-block design with four replications and four treatment methods; spot and zonal treatments based on SODD identified incidence and density, respectively, compared against full-coverage and non-treated controls. Applications were made with a Toro Multipro5800 with GeoLink GNSS-equipped sprayer in Fall of 2021. Spot and zonal treatment strategies showed similar control to full-coverage applications (p≤0.001) while reducing the percentage of the fairways treated by 48% and 52%, respectively (p≤0.001). These results highlight the capabilities for SODD as a tool for disease map generation.

Counting soybean plants is a crucial strategy for assessing sowing quality and supporting high production. Despite its importance, the laborious nature of traditional assessment methods makes them unreliable and not scalable. Additionally, innovative image-based solutions have demonstrated limitations in detecting dense crops such as soybeans. Therefore, in this study, we developed neural network models to analyze a set of RGB and multispectral images and perform plant classification in a comprehensive dataset, which included data collected at three vegetative stages of soybean (VC, V1, and V2). Our results demonstrated high accuracy in classifying plants using either RGB (98%) or multispectral images (92%). A significant strength of this study is the ability to classify highly dense plants, without a trend for misclassification. Clearly, our findings provide stakeholders with a timely and effective approach to counting soybean plants, reducing labor and time, while increasing reliability.

Precision Agriculture relies significantly on yield data obtained from combine harvesters, which constitutes a pivotal tool for optimizing crop productivity. Despite its potential, challenges in data accuracy persist, necessitating the development of novel automated postprocessing protocols for yield data refinement. In this paper, different automatic postprocessing protocols were evaluated using remote sensing data, specifically Sentinel-2 satellite imagery. Various automatic postprocessing protocols were applied to a dataset spanning 946 hectares over a four-year period. Commercial sensors on combine harvesters acquired the yield data. The analysis included global (field-level) adjustments and local adjustments at a finer scale (40 × 40 m² level), employing interval mean ± n·(standard deviation) calculations. Three n values (1, 1.5, and 2.5) were tested, resulting in thirteen distinct postprocessing variations. Finally, a mean filter was also applied. The results demonstrated that the yield correlation with satellite data increased with the reduction of yield variability at the pixel level (10 m). The best results were obtained using n = 1 with a 3 × 3 mean filter, where Sentinel-2 pixels remained unaffected, and the average Root Mean Square Error (RMSE) during validation was 0.572 t·ha⁻¹. In addition, the geostatistical parameters (coefficient of variation, semivariance, and range within a 10 m pixel) reached optimal values. Finally, the postprocessing uncertainty was determined to be 0.200 t·ha⁻¹. These results validate the efficacy of a novel postprocessing protocol for refining yield data in rice crops. The integration of pixel-level data from combine harvesters with Sentinel-2 imagery emerges as a promising approach for optimizing crop management, offering valuable insights for the advancement of Precision Agriculture.

Accurate estimation and spatial allocation of economic optimum nitrogen (N) rates (EONR) can support sustainable crop production systems by reducing chemical compounds to be applied to the ground while preserving the optimum yield and profitability Smart Farming (SF) techniques such as historical precision agriculture (PA) machinery data, satellite multispectral imagery, and on-machine nitrogen adjustment sensors can bring together state-of-the-art precision in determining EONR. The novelty of this study is in introducing an efficient optimization framework using SF technology to enable real-time and prescription based EONR application execution. An optimization strategy called response surface modelling (RSM) was implemented to support decision making by fusing multiple sources of information while keeping the underlying computation simple and interpretable. Here, a field of winter wheat with an area of 7 ha was used to prove the proposed concept of determining EONR for each location in the field using auxiliary variables called multispectral indices (MSIs) derived from Sentinel 2. Three different image acquisition dates before the actual N application were considered to find the best time combination of MSIs along with the best MSIs to model yield. The best MSIs were filtered out through three phases of feature selection using analysis of variance (ANOVA), Lasso regression, and model reduction of RSM. For the date 2020.03.25, 14 out of 21 MSIs exhibited a significant interaction with the N applied as determined through an on-machine N sensor. For dates 2020.03.30 and 2020.04.04, the numbers of significant indices were identified as 6 and 10, respectively. Some of the MSIs were no longer significant after five days of the growth period (5-day interval between Sentinel 2 revisits). The best model demonstrated an average prediction error of 14.5%. Utilizing the model’s coefficients, the EONR was computed to be between 43 kg/ha and 75 kg/ha for the target field. By incorporating MSIs into the fitted model for a given N range, it was demonstrated that the shape of the yield-N relation (RSM) varied due to field heterogeneity. The proposed analytical approach integrates farmer engagement by participatory annual post-mortem analysis. Using the determined RSM approach, retrospective assessment compares economically optimal N input, based on observed MSIs values to each location, with the actual applied rates.