Precision Agriculture最新文献

Abstract

This study presents a method based on retrofitted low-cost and easy to implement tracking devices, used to monitor the whole harvesting process in viticulture, to map yield and harvest quality parameters in viticulture. The method consists of recording the geolocation of all the machines (harvest trailers and grape harvester) during the harvest to spatially re-allocate production parameters measured at the winery. The method was tested on a vineyard of 30 ha during the whole 2022 harvest season. It has identified harvest sectors (HS) associated with measured production parameters (grape mass and harvest quality parameters: sugar content, total acidity, pH, yeast assimilable nitrogen, organic nitrogen) and calculated production parameters (potential alcohol of grapes, yield, yield per plant) over the entire vineyard. The grape mass was measured at the vineyard cellar or at the wine-growing cooperative by calibrated scales. The harvest quality parameters were measured on grape must samples in a commercial laboratory specialized in oenological analysis and using standardized protocols. Results validate the possibility of making production parameters maps automatically solely from the time and location records of the vehicles. They also highlight the limitations in terms of spatial resolution (the mean area of the HS is 0.3 ha) of the resulting maps which depends on the actual yield and size of harvest trailers. Yield per plant and yeast assimilable nitrogen maps have been used, in collaboration with the vineyard manager, to analyze and reconsider the fertilization process at the vineyard scale, showing the relevance of the information.

Abstract

Chickpea (Cicer arietinum L.) is a major grain legume grown worldwide as a staple protein source. Traditionally, it is a rain-fed crop, but supplemental irrigation can increase yields and counteract the challenges posed by the changing climate worldwide. A fast and non-destructive plant water status assessment method may streamline irrigation management. The main objective of this study was to remotely assess the leaf water potential (LWP) and leaf area index (LAI) of field-grown chickpea. Five irrigation treatments were applied in two farm experiments and two commercial fields. Ground hyperspectral canopy reflectance and Vegetation and Environment monitoring on a New Micro-Satellite (VENµS) images acquired throughout the study. In parallel, LWP and LAI measurements were captured in the field. Vegetation indices (VIs) and machine learning (ML) based on all spectral bands were used to calibrate and validate spectral estimation models. The normalized difference spectral index (NDSI) that used bands on 1600 and 1730 nm (NDSI_(1600,1730)) selected in the current study yielded the LWP lowest estimation error on independent validation (RMSE = 0.19 [MPa]) using linear regression. VENµS based VIs resulted in relatively lower LWP estimation accuracy (RMSE = 0.23–0.29 [MPa]) compared to VIs calculated from ground hyperspectral data (RMSE = 0.19–0.21 [MPa]). Artificial neural network (ANN) models for LWP from ground and space spectral data showed similar performances (RMSE = 0.15–0.17 [MPa]), and were both more accurate than VIs. LWP response to the irrigation treatments was faster than the LAI response and was captured by the NDSI_(1600,1730). The low correlation found between LWP and LAI (r = 0.08–0.44) supports the conclusion that spectral reflectance of chickpea canopy can be used to estimate LWP per se and is only partially affected by morphological changes induced by irrigation treatments and canopy development. The ability to rapidly estimate chickpea LWP may improve irrigation scheduling in the future.

Sensed and soil sample data are used in two main approaches for mapping soil properties in precision agriculture: management zones (MZs) and contour maps. This is the first of two papers that explores maps of MZs. Management zones based on variation in sensed data that are related to the more permanent soil properties assume that the zones are multi-purpose. Soil properties are then often sampled on a grid to provide the average values of each property per zone. This paper examines the plausibility of this approach by examining how the number of samples taken on a grid and the application of kriging affect mean soil property values for MZs. The suitability of MZs based on ancillary data for managing several agronomically important properties simultaneously is also considered. These concepts are examined with historic soil data from four field sites in southern UK with different scales of spatial variation. Results showed that when the grid sampling interval is large, there is less difference in the means of properties between MZs, but kriging the soil data increased the differences between zones when the sampling interval was large and sample small. Sensed data are used increasingly to aid the identification of MZs, but these could not be considered multi-purpose at all sites. The MZs produced were most useful for phosphorus (P), pH and volumetric water content (VWC) at the Wallingford site and useful for most properties at the Clays and Y215 sites. For the latter site this was true only when the most dense data were used to calculate MZ averages. The results show that sampling interval for MZ averages should relate to the scale of variation or the size of the MZs at a site. The sampling density could be based on the variogram range of ancillary data. This research suggests that there should be 6–8 samples per zone to obtain accurate averages of soil properties. Nutrient data for more than one year were examined at two sites and showed that patterns remained consistent in the short term unless variable-rate management was used, but also the range of values changed in the short term.

The estimation of pre-harvest fruit quality and maturity is essential for growers to determine the harvest timing, storage requirements and profitability of the crop yield. In-field fruit maturity indicators are highly variable and require high spatiotemporal resolution data, which can be obtained from contemporary precision agriculture systems. Such systems exploit various state-of-the-art sensors, increasingly relying on spectrometry and imaging techniques in association with advanced Artificial Intelligence (AI) and, in particular, Machine Learning (ML) algorithms. This article presents a critical review of precision agriculture techniques for fruit maturity estimation, with a focus on destructive and non-destructive measurement approaches, and the applications of ML in the domain. A critical analysis of the advantages and disadvantages of different techniques is conducted by surveying recent articles on non-destructive methods to discern trends in performance and applicability. Advanced data-fusion methods for combining information from multiple non-destructive sensors are increasingly being used to develop more accurate representations of fruit maturity for the entire field. This is achieved by incorporating AI algorithms, such as support vector machines, k-nearest neighbour, neural networks, and clustering. Based on an extensive survey of recently published research, the review also identifies the most effective fruit maturity indices, namely: sugar content, acidity and firmness. The review concludes by highlighting the outstanding technical challenges and identifies the most promising areas for future research. Hence, this research has the potential to provide a valuable resource for the growers, allowing them to familiarize themselves with contemporary Smart Agricultural methodologies currently in use. These practices can be gradually incorporated from their perspective, taking into account the availability of non-destructive techniques and the use of efficient fruit maturity indices.

Abstract

The fast and accurate provision of within-season data of green area index (GAI) and total N uptake (total N) is the basis for crop modeling and precision agriculture. However, due to rapid advancements in multispectral sensors and the high sampling effort, there is currently no existing reference work for the calibration of one UAV (unmanned aerial vehicle)-based multispectral sensor to GAI and total N for silage maize, winter barley, winter oilseed rape, and winter wheat.

In this paper, a practicable calibration framework is presented. On the basis of a multi-year dataset, crop-specific models are calibrated for the UAV-based estimation of GAI throughout the entire growing season and of total N until flowering. These models demonstrate high accuracies in an independent evaluation over multiple growing seasons and trial sites (mean absolute error of 0.19–0.48 m² m⁻² for GAI and of 0.80–1.21 g m⁻² for total N). The calibration of a uniform GAI model does not provide convincing results. Near infrared-based ratios are identified as the most important component for all calibrations. To account for the significant changes in the GAI/ total N ratio during the vegetative phase of winter barley and winter oilseed rape, their calibrations for total N must include a corresponding factor. The effectiveness of the calibrations is demonstrated using three years of data from an extensive field trial. High correlation of the derived total N uptake until flowering and the whole-season radiation uptake with yield data underline the applicability of UAV-based crop monitoring for agricultural applications.

Improving agricultural nitrogen management is one of the key objectives of the recent Green Deal in Europe. Current technological developments in agriculture offer new opportunities to improve nitrogen fertilization practices. The aim of this study was to adapt to Sentinel-2 data a proven delineation method initially developed for yield maps, in order to facilitate precise nitrogen management by farmers. The study was conducted in two steps. Firstly, an analysis at annual level was conducted to assess the relationship between vegetation indices and yield at the subfield scale, for different sensing period. The second step consisted in performing a pluri- annual analysis through the delineation of management zones and compare the results achieved from yield maps and from NDVI maps. Among different vegetation indices, NDVI proved to be an interesting candidate for subfield detection of yield variation, specifically when the index was sensed was sensed around the second half of May. In this area, this period usually corresponds to phenological development between the flag leaf stage and heading stage, just prior the initiation of winter wheat flowering. Using NDVI maps within Blackmore’s delineation approach instead of yield maps. Allowed to reach an accuracy of 69% on zone classification. However, as yields and NDVI distribution do not respond to similar statistical distributions, we considered that the delineation threshold used to differentiate high from low yielding zones had to be adapted. The adaptation of the “performance threshold” in favor of the median NDVI, made it possible to achieve a higher accuracy (71%) of the delineation. But above all, the improvement lies also in a more robust satellite-based delineation.

Abstract

Site-specific estimation of lime requirement requires high-resolution maps of soil organic carbon (SOC), clay and pH. These maps can be generated with digital soil mapping models fitted on covariates observed by proximal soil sensors. However, the quality of the derived maps depends on the applied methodology. We assessed the effects of (i) training sample size (5–100); (ii) sampling design (simple random sampling (SRS), conditioned Latin hypercube sampling (cLHS) and k-means sampling (KM)); and (iii) prediction model (multiple linear regression (MLR) and random forest (RF)) on the prediction performance for the above mentioned three soil properties. The case study is based on conditional geostatistical simulations using 250 soil samples from a 51 ha field in Eastern Germany. Lin’s concordance correlation coefficient (CCC) and root-mean-square error (RMSE) were used to evaluate model performances. Results show that with increasing training sample sizes, relative improvements of RMSE and CCC decreased exponentially. We found the lowest median RMSE values with 100 training observations i.e., 1.73%, 0.21% and 0.3 for clay, SOC and pH, respectively. However, already with a sample size of 10, models of moderate quality (CCC > 0.65) were obtained for all three soil properties. cLHS and KM performed significantly better than SRS. MLR showed lower median RMSE values than RF for SOC and pH for smaller sample sizes, but RF outperformed MLR if at least 25–30 or 75–100 soil samples were used for SOC or pH, respectively. For clay, the median RMSE was lower with RF, regardless of sample size.

Timely and accurate estimates of sugarcane yield provide valuable information for food management, bio-energy production, (inter)national trade, industry planning and government policy. Remote sensing and machine learning approaches can improve sugarcane yield estimation. Previous attempts have however often suffered from too few training samples due to the fact that field data collection is expensive and time-consuming. Our study demonstrates that unmanned aerial vehicle (UAV) data can be used to generate field-level yield data using only a limited number of field measurements. Plant height obtained from RGB UAV-images was used to train a model to derive intra-field yield maps based on 41 field sample plots spread over 20 sugarcane fields in the Udon Thani Province, Thailand. The yield maps were subsequently used as reference data to train another model to estimate yield from multi-spectral Sentinel-2 (S2) imagery. The integrated UAV yield and S2 data was found efficient with RMSE of 6.88 t/ha (per 10 m × 10 m pixel), for average yields of about 58 t/ha. The expansion of the sugarcane yield mapping across the entire region of 11,730 km² was in line with the official statistical yield data and highlighted the high spatial variability of yields, both between and within fields. The presented method is a cost-effective and high-quality yield mapping approach which provides useful information for sustainable sugarcane yield management and decision-making.

The leaf area index (LAI), a crucial biophysical indicator, is used to assess and monitor crop growth for effective agricultural management. This study assessed the LAI at the seedling stage after conducting a field experiment with rainfed groundnut. The study tests the performance of multiple machine learning regression algorithms (MLRAs) and empirical vegetation indices (VIs) in retrieving groundnut's LAI using freely available Sentinel-2 data. The bands at 665 nm, 705 nm, 842 nm, and 2190 nm are the most sensitive for retrieving groundnut's LAI, according to an analysis of its band spectrum. Results suggest that VIs computed with wavebands centered at red (665 nm), red edge (705 nm), and near-infrared (842 nm) exhibited optimal R² with Sentinel-2 data. Normalized difference vegetation index (NDVI), red edge normalized difference vegetation index (NDVIre), simple ratio (SR), red edge simple ratio (SRre), and green normalized difference vegetation index (gNDVI) were utilized as predictors for LAI. Regarding the results of the validation between estimated and measured LAI, SR demonstrated the highest accuracy for groundnut LAI prediction (r² = 0.67, RMSE = 0.89). Ten MLRAs were tested, and results indicate from the perspective of the accuracy of models, the Gaussian processes regression, GPR (r² = 0.73 and RMSE = 0.81), Kernel ridge regression, KRR (r² = 0.72 and RMSE = 0.82) and Support vector regression, SVR (r² = 0.70 and RMSE = 0.85) demonstrated to be the most suitable for LAI estimation for rainfed groundnut at the seedling stage. The systematic analysis based on the regression approaches tested here revealed that the GPR outperformed other models combined, therefore, most suitable for estimating rainfed groundnut LAI at the seedling stage. These findings serve as a benchmark for obtaining crop biophysical parameters in the framework of groundnut traits monitoring in a tropical West Africa.

Autonomous robot-based orchard operations will become an alternative solution in the field of precision agriculture. One of the keys to robotic work is to achieve autonomous navigation that is as accurate as possible to ensure the most accurate working effect. In this work, we propose an orchard path fitting and navigation method based on the fusion of improved A-Star algorithm and Support Vector Machine Regression (SVR) to meet the requirements of autonomous navigation crawler platform for autonomous navigation in orchard environment and ensure accuracy. In this study, the actual speed and turning radius of the left and right tracks of the crawler platform were collected under 5 different slopes and 400 sets of different theoretical speed combinations of left and right tracks through the design nesting test, and the motion model of the crawler platform was constructed based on SVR. Orchard point cloud data were obtained by 3D solid-state LiDAR, and the improved A-star algorithm was used to fit the navigation path and calculate the turning curvature radius. Taking this curvature radius as the optimal navigation target value, the motion model predicts the optimal theoretical speed of left and right tracks, which is used as a reference for autonomous navigation. The comparison experiment of autonomous navigation was carried out in two modes: traditional and improved A-Star algorithm. The results show that the average values of the maximum lateral and longitudinal deviation of the improved automatic navigation method between orchards row are 6.90 cm and 9.88 cm, respectively. Compared with the method combined with the traditional A-Star algorithm and SVR, the values were 8.94 cm and 10.88 cm and were optimized by 29.57% and 10.12%, respectively. The autonomous navigation method proposed in this paper can meet the needs of orchards rows autonomous navigation, and can be widely applied to different orchard site environments (slope, ground obstacles, bad surface conditions), which can provide reference for the production practices of unmanned orchards.