Precision Agriculture最新文献

Visible-Near Infrared (vis-NIR) spectroscopy can provide a faster, cost-effective, and user-friendly solution to monitor leaf N status, potentially overcoming the limitations of current techniques. The objectives of the study were to develop and validate partial least square regression (PLSR) to estimate the total N contents of fresh and removed leaves of potatoes using the vis-NIR spectral range (350–2500 nm) generated from a handheld proximal sensor. The model was built using data collected from Hancock Agricultural Research Station, WI, USA in 2020 and was validated using samples collected in 2021 for four different conditions. The conditions included two sites (Coloma and Hancock), four potato varieties (Burbank, Norkotah, Goldrush, and Silverton), two N rates (unfertilized and 308 kg N ha⁻¹), and four growth stages (vegetative, tuber initiation, tuber bulking, and tuber maturation). The calibration and validation models had high predictive performance for leaf total N with R² > 0.8 and RPD > 2. The model accuracy was affected by the total N contents in the leaf samples where the model underpredicted the samples with total leaf N contents greater than 6%.

Drought is a major abiotic stress that affects the productivity of sugarcane worldwide. Water deficiency during sugarcane growth will lead to a reduction in leaf pigment content, such as chlorophyll, known as chlorosis. Although changes in spectral reflectance signature were identified a conspicuous sign of chlorophyll content changes caused by drought stress, the quantitative relationships between leaf chlorophyll content and spectral reflection signatures are still poorly explored. In this study, we present our contribution in systematically establishing a model for estimating leaf chlorophyll content in drought-affected sugarcane using VIS/NIR reflectance spectroscopy and characteristic band extraction techniques. Leaves of sugarcane plants at early elongation stage under different controlled irrigation conditions were used for spectra data collection, and the chlorophyll contents were collected with standard analytical methods. Different characteristic band extraction techniques and regression models were compared and discussed to obtain a chlorophyll content estimation model with the best performance. As the quantitative results, the combination of characteristic bands extracted by the successive projection algorithm (SPA) with a Stacking regression model achieved a high chlorophyll content estimation performance (R² = 0.9834, RMSE  = 0.0544 mg/cm²) with only 4.3% of original spectral variables as inputs. This study provides a theoretical basis for accurate and non-invasive drought stress level estimation in large-scale cultivation.

Site-specific crop management is based on the postulate of varying soil and crop requirements in a field. Therefore, a field is separated into homogenous management zones, using available data to adapt management practices environment to maximize productivity and profitability while reducing environmental impacts. Due to advancing sensor technologies, crop growth and yield data on more minor scales are common, but soil data often needs to be more appropriate. Crop growth models have shown promise as a decision support tool for site-specific farming. The Decision Support System for Agrotechnology Transfer (DSSAT) is a widely used point-based model. To overcome the problem of inappropriate soil input data problem, this study introduces an external plug-in program called Soil Profile Optimizer (SPO), which uses the current DSSAT v4.8 to calibrate soil profile parameters on a site-specific level. Developed as an inverse modelling approach, the SPO can calibrate selected soil profile parameters by targeting available in-season plant data. Root Mean Square Error (RMSE) and normalized RMSE as error minimization criteria are used. The SPO was tested and evaluated by comparing different simulation scenarios in a case study of a 3-yr field trial with maize. The scenario with optimized soil profiles, conducted with the SPO, resulted in an R² of 0.76 between simulated and observed yield and led to significant improvements compared to the scenario conducted with field scale soil profile information (R² 0.03). The SPO showed promise in using spatial plant measurements to estimate management zone scale soil parameters required for the DSSAT model.

Hailstorms pose a direct threat to agriculture, often causing yield losses and worsening farmers’ agricultural activity. Traditional methods of hail damage estimation, conducted by insurance field inspectors, have been questioned due to their complexity, partial subjectivity, and lack of accounting for spatial variability. Therefore, remote sensing integration in the estimation process could provide a valuable aid. The focus of this study was on winter wheat (Triticum aestivum L.) and its response to damage in the near-infrared (NIR) spectral region, with a particular emphasis on the study of brown pigments as a proxy for yield damage estimation and mapping. An experiment was conducted during two cropping seasons (2020–2021 and 2021–2022) at two sites, simulating hail damage at critical flowering and milky stages using a specifically designed prototype machinery with low, medium, and high damage gradients compared to undamaged conditions in plots with a minimum of 400 m² area. After the damage simulation, hyperspectral visible-NIR reflectance was measured with Unmanned Aerial Vehicle (UAV) flights, and measurements of chlorophyll and of leaf area index (LAI) were contextually taken. Final yield per treatment was recorded using a combine. An increase in absorbance in the NIR region (780–950 nm) was observed and evaluated using a spectral mixture analysis (SMA) after selecting representative damaged and undamaged vegetation spectra to map the damage. The abundance of damaged endmember pixels per treatment resulted in a good relationship with the final yield (R² = 0.73), identifying the most damaged areas. The absorbance feature was further analysed with a newly designed multispectral index (TAI), which was tested against a selection of indices and resulted in the highest relationship with the final yield (R² = 0.64). Both approaches were effective in highlighting the absorbance feature over different dates and development stages, defining an effective mean for hailstorm damage mapping in winter wheat.

Plant nitrogen (N) is one of the key factors for its growth and yield. Timely assessment of plant N at a spatio-temporal scale enables its precision management in the field scale with better N use efficiency. Airborne imaging spectroscopy is a potential technique for non-invasive near real-time rapid assessment of plant N on a field scale. The present study attempted to assess plant N in a wheat field with three different irrigation levels (I₁–I₃) along with five nitrogen treatments (N₀–N₄) using a UAV hyperspectral imager with a spectral range of 400 to 1000 nm. A total of 61 vegetative indices were evaluated to find suitable indices for estimating plant N. A hybrid method of R-Square (R²) and Variable Importance Projection (VIP) followed by Variance Inflation Factor was used to limit the best suitable N-sensitive 13 spectral indices. The selected indices were used as feature vectors in the Artificial Neural Network algorithm to model and generate a spatial map of plant N in the experimental wheat field. The model resulted in R² values of 0.97, 0.84, and 0.86 for training, validation, and testing respectively for plant N assessment.

Soil compaction is considered one of the main threats to structural soil degradation, and it promotes increased densification of soil particles, impairs ecosystem services, the plant development, and therefore affects agricultural profitability. In this sense, this study aimed to analyze the feasibility of using a Remotely Piloted Aircraft System (RPAS) by relating parameters derived from aerial images based on Vegetation Indices (VIs) and the Canopy Height Model (CHM) with soil compaction in a coffee plantation area. The study was conducted in a commercial coffee plantation with the cultivar Mundo Novo with 14 years of implantation. Two aerial surveys were carried out, the first to determine the CHM and define the sampling points and the second for radiometric calculations of VIs. In the sampling point were collected data plant height, soil characterization, soil penetration resistance and productivity. Images were processed by Pix4D software, and the data analysis at QGIS and RStudio. As at results, no statistically significant differences were detected between the different plant height zones in the soil chemical analysis; significant statistical differences between plant height zones were detected for penetration resistance, which is correlated to productivity data; and the radiometric data presented a correlation with the penetration resistance data, making it possible to determine VIs (NDRE and MTCI) with correlation to the compaction data allowing the estimation of such variable. In this way, the possibility of monitoring the height variations of the coffee crop using RPAS to demarcate compacted zones was evidenced.

Fruit cracking is a preharvest physiological rind disorder in citrus, sometimes causing considerable yield loss. In recent years, reports from Israel and other countries suggest that cracking incidence has increased, which might indicate that climate change intensifies the phenomena. The study aims to develop a machine learning (ML) model for predicting the effect of climate measures (i.e., temperature, radiation, and humidity with daily resolution) along with management and environmental characteristics in two citrus mandarins, ‘Nova’ and ‘Ori’, one is prone to cracking and the other is less sensitive. ML model was developed based on data from approximately 250 citrus orchards across Israel collected over three seasons from 2019 to 2021. Our approach uses TSFRESH to extract and select features and SHAP (SHapley Additive exPlanations) to explain the factor’s intensity using trained classification and regression models based on the H2O-AutoML package. Gathered data skewed toward a low cracking percentage better predicted low and medium cracking levels, with a classification accuracy of 76% and regression mean absolute error (MAE) of 4.78%. Our study reaffirms the genetic background’s primary role in cracking. Notably, our analysis unveils fresh insights into cracking causes needing further exploration. The 40% quantile temperature (23.5 °C) is a novel finding as a learned threshold. ‘Nova’ may elevate cracking by 10%, ‘Ori’ could reduce it by 4%. Additionally, tree age exhibits a linear correlation when trees over 20 years correlate with up to 4% less cracking. These insights are crucial for comprehending, addressing, and managing the phenomenon at a significant spatial scale. The model, with further data support, may provide farmers with an effective tool for treating the severity of cracking incidence by developing a spatial–temporal decision-support system as a protocol to reduce the phenomenon on a regional scale and selecting regions that are relevant for citrus plantations.

Accurately monitoring Canopy Nitrogen Concentration (CNC) is a prerequisite for precision nitrogen (N) fertiliser management at the farm scale with carbon and N budgeting across the landscape and ecosystems. While many spectral indices have been proposed for CNC monitoring, their applicability and accuracy are often adversely affected by confounding factors such as aboveground biomass (AGB), crop type, growth stages, and environmental conditions, limiting their broader application and adoption; with AGB being one of the most dominant signals and confounding factors at canopy scale. The confounding effect can become more challenging as AGB is also physiologically linked with CNC across the growth stages. Additionally, the interplay between index form, selection of optimal wavebands and their bandwidths remains poorly understood for CNC index design. This study proposes robust and cost-effective 2- and 4-waveband multispectral (MS) CNC indices applicable across a wide range of crop conditions. We collected 449 canopy reflectance spectra (400–980 nm) together with corresponding CNC and AGB measurements across four growth stages of ryegrass (winter and summer), and five growth stages of barley (winter-spring) in Victoria, Australia, in 2018 and 2019. All possible waveband (400–980 nm) combinations revealed that the best combination varied between seasons and crop types. However, the visible spectrum, particularly the blue region, presented high and consistent performance. Bandwidths of 10–40 nm outperformed either very narrow (2 nm) or very broad bandwidths (80 nm). The newly developed 2-waveband index (416 and 442 nm with 10-nm bandwidth; R² = 0.75 and NRMSE = 0.2) and 4-waveband index (512, 440, 414 and 588 nm with 40-nm bandwidth; R² = 0.81 and NRMSE = 0.17) exhibited the best performance, while validation with an independent dataset (from a different growing period to those used in the model development) obtained NRMSE values of 0.25 and 0.24, respectively. The 4-waveband index provides enhanced performance and permits use of broader bandwidths than its 2-waveband counterpart.

Accurate, non-destructive forecasting of carrot yield is difficult due to its subterranean growing habit. Furthermore, the timing of forecasting usually occurs when the crop is mature, limiting the opportunity to implement alternative management decisions to improve yield (during the growing season). This study aims to improve the accuracy of carrot yield forecasting by exploring time series and multivariate approaches. Using Sentinel-2 satellite imagery in three Australian vegetable regions, we established a time series of carrot phenological stages (PhS) from ‘days after sowing’ (DAS) to enhance prediction timing. Numerous vegetation indices (VIs) were analyzed to derive temporal growth patterns. Correlations with yield at different PhS were established. Although the average root yield (t ha⁻¹) did not significantly differ across the regions, the temporal VI signatures, indicating different regional crop growth trends, did vary as well as the PhS at when the maximum correlation with yield occurred ((PhS_{{R2_{max} }} )) with two of the regions producing a delayed (PhS_{{R2_{max} }}) (i.e. 90–130 DAS). The best multivariate model was identified at 70 DAS, extending the forecasting window before harvest between 20 to 60 days. The performance of this model was validated with new crops producing an average error of 16.9 t ha⁻¹ (27% of total yield). These results demonstrate the potential of the model at such early stage under varying growing conditions offering growers and stakeholders the chance to optimize farming practices, make informed decisions on selling, harvesting, and labor planning, and adopt precision agriculture methods.

Medjool date thinning automation is essential for reducing Medjool production labor and improving fruit quality. Thinning automation requires motion planning based on feature extraction from a segmented fruit bunch and its components. Previous research with focused bunch images attained high success in bunch component segmentation but less success in establishing correct association between the two components (a rachis and spikelets) that form one bunch. The current study presents an algorithm for improved component segmentation and association in the presence of occlusions based on integrating deep neural networks, traditional methods building on bunch geometry, and active vision. Following segmentation with Mask-R-CNN, segmented component images are converted to binary images with a Savitzky–Golay filter and an adapted Otsu threshold. Bunch orientation is calculated based on lines found in the binary image with the Hough transform. The orientation is used for associating a rachis with spikelets. If a suitable rachis is not found, bunch orientation is used for selecting a better viewpoint. The method was tested with two databases of bunches in an orchard, one with focused and one with non-focused images. In all images, the spikelets were correctly identified [intersection over union (IoU) 0.5: F1 0.9]. The average orientation errors were 18.15° (SD 12.77°) and 16.44° (SD 11.07°), respectively, for the focused and non-focused databases. For correct rachis selection, precision was very high when incorporating orientation, and when additionally incorporating active vision recall (and therefore F1) was high (IoU 0.5: orientation: precision 0.94, recall 0.44, F1 0.60; addition of active vision: precision 0.96, recall 0.61, F1 0.74). The developed method leads to highly accurate identification of fruit bunches and their spikelets and rachis, making it suitable for integration with a thinning automation system.