Are unmanned aerial vehicle-based hyperspectral imaging and machine learning advancing crop science?

The past few years have seen increased interest in unmanned aerial vehicle (UAV)-based hyperspectral imaging (HSI) and machine learning (ML) in agricultural research, concomitant with an increase in published research on these topics. We provide an updated review, written for agriculturalists, highlighting the benefits in the retrieval of biophysical parameters of crops via UAVs relative to less sophisticated options. We reviewed >70 recent papers and found few consistent results between similar studies. Owing to their high complexity and cost, especially when applied to crops of low value, the benefits of most of the research reviewed are difficult to explain. Future effort will be necessary to distill research findings into lower-cost options for end-users.

UAV-Based Hyperspectral Monitoring Using Push-Broom and Snapshot Sensors: A Multisite Assessment for Precision Viticulture Applications.

Hyperspectral aerial imagery is becoming increasingly available due to both technology evolution and a somewhat affordable price tag. However, selecting a proper UAV + hyperspectral sensor combo to use in specific contexts is still challenging and lacks proper documental support. While selecting an UAV is more straightforward as it mostly relates with sensor compatibility, autonomy, reliability and cost, a hyperspectral sensor has much more to be considered. This note provides an assessment of two hyperspectral sensors (push-broom and snapshot) regarding practicality and suitability, within a precision viticulture context. The aim is to provide researchers, agronomists, winegrowers and UAV pilots with dependable data collection protocols and methods, enabling them to achieve faster processing techniques and helping to integrate multiple data sources. Furthermore, both the benefits and drawbacks of using each technology within a precision viticulture context are also highlighted. Hyperspectral sensors, UAVs, flight operations, and the processing methodology for each imaging type' datasets are presented through a qualitative and quantitative analysis. For this purpose, four vineyards in two countries were selected as case studies. This supports the extrapolation of both advantages and issues related with the two types of hyperspectral sensors used, in different contexts. Sensors' performance was compared through the evaluation of field operations complexity, processing time and qualitative accuracy of the results, namely the quality of the generated hyperspectral mosaics. The results shown an overall excellent geometrical quality, with no distortions or overlapping faults for both technologies, using the proposed mosaicking process and reconstruction. By resorting to the multi-site assessment, the qualitative and quantitative exchange of information throughout the UAV hyperspectral community is facilitated. In addition, all the major benefits and drawbacks of each hyperspectral sensor regarding its operation and data features are identified. Lastly, the operational complexity in the context of precision agriculture is also presented.

Assessing Grapevine Biophysical Parameters From Unmanned Aerial Vehicles Hyperspectral Imagery.

Over the last 50 years, many approaches for extracting plant key parameters from remotely sensed data have been developed, especially in the last decade with the spread of unmanned aerial vehicles (UAVs) in agriculture. Multispectral sensors are very useful for the elaboration of common vegetation indices (VIs), however, the spectral accuracy and range may not be enough. In this scenario, hyperspectral (HS) technologies are gaining particular attention thanks to the highest spectral resolution, which allows deep characterization of vegetative/soil response. Literature presents few papers encompassing UAV-based HS applications in vineyard, a challenging conditions respect to other crops due to high presence of bare soil, grass cover, shadows and high heterogeneity canopy structure with different leaf inclination. The purpose of this paper is to present the first contribution combining traditional and multivariate HS data elaboration techniques, supported by strong ground truthing of vine ecophysiological, vegetative and productive variables. Firstly the research describes the UAV image acquisition and processing workflow to generate a 50 bands HS orthomosaic of a study vineyard. Subsequently, the spectral data extracted from 60 sample vines were elaborated both investigating the relationship between traditional narrowband VIs and grapevine traits. Then, multivariate calibration models were built using a double approach based on Partial Least Square (PLS) regression and interval-PLS (iPLS), to evaluate the correlation performance between the biophysical parameters and HS imagery using the whole spectral range and a selection of more relevant bands applying a variable selection algorithm, respectively. All techniques (VIs, PLS and iPLS) provided satisfactory correlation performances for the ecophysiological (R 2 = 0.65), productive (R 2 = 0.48), and qualitative (R 2 = 0.63) grape parameters. The novelty of this work is represented by the first assessment of a UAV HS dataset with the expression of the entire vine ecosystem, from the physiological and vegetative state to grapes production and quality, using narrowband VIs and multivariate PLS regressions. A correct non-destructive estimation of key parameters in vineyard, above all physiological parameters which must be measured in a short time as they are extremely influenced by the variability of environmental conditions during the day, represents a powerful tool to support the winegrower in vineyard management.

Beyond the traditional NDVI index as a key factor to mainstream the use of UAV in precision viticulture.

In the last decade there has been an exponential growth of research activity on the identification of correlations between vegetational indices elaborated by UAV imagery and productive and vegetative parameters of the vine. However, the acquisition and analysis of spectral data require costs and skills that are often not sufficiently available. In this context, the identification of geometric indices that allow the monitoring of spatial variability with low-cost instruments, without spectral analysis know-how but based on photogrammetry techniques with high-resolution RGB cameras, becomes extremely interesting. The aim of this work was to evaluate the potential of new canopy geometry-based indices for the characterization of vegetative and productive agronomic parameters compared to traditional NDVI based on spectral response of the canopy top. Furthermore, considering grape production as a key parameter directly linked to the economic profit of farmers, this study provides a deeper analysis focused on the development of a rapid yield forecast methodology based on UAV data, evaluating both traditional linear and machine learning regressions. Among the yield assessment models, one of the best results was obtained with the canopy thickness which showed high performance with the Gaussian process regression models (R2 = 0.80), while the yield prediction average accuracy of the best ML models reached 85.95%. The final results obtained confirm the feasibility of this research as a global yield model, which provided good performance through an accurate validation step realized in different years and different vineyards.

Evaluation of novel precision viticulture tool for canopy biomass estimation and missing plant detection based on 2.5D and 3D approaches using RGB images acquired by UAV platform.

BACKGROUND: The knowledge of vine vegetative status within a vineyard plays a key role in canopy management in order to achieve a correct vine balance and reach the final desired yield/quality. Detailed information about canopy architecture and missing plants distribution provides useful support for farmers/winegrowers to optimize canopy management practices and the replanting process, respectively. In the last decade, there has been a progressive diffusion of UAV (Unmanned Aerial Vehicles) technologies for Precision Viticulture purposes, as fast and accurate methodologies for spatial variability of geometric plant parameters. The aim of this study was to implement an unsupervised and integrated procedure of biomass estimation and missing plants detection, using both the 2.5D-surface and 3D-alphashape methods. RESULTS: Both methods showed good overall accuracy respect to ground truth biomass measurements with high values of R2 (0.71 and 0.80 for 2.5D and 3D, respectively). The 2.5D method led to an overestimation since it is derived by considering the vine as rectangular cuboid form. On the contrary, the 3D method provided more accurate results as a consequence of the alphashape algorithm, which is capable to detect each single shoot and holes within the canopy. Regarding the missing plants detection, the 3D approach confirmed better performance in cases of hidden conditions by shoots of adjacent plants or sparse canopy with some empty spaces along the row, where the 2.5D method based on the length of section of the row with lower thickness than the threshold used (0.10 m), tended to return false negatives and false positives, respectively. CONCLUSIONS: This paper describes a rapid and objective tool for the farmer to promptly identify canopy management strategies and drive replanting decisions. The 3D approach provided results closer to real canopy volume and higher performance in missing plant detection. However, the dense cloud based analysis required more processing time. In a future perspective, given the continuous technological evolution in terms of computing performance, the overcoming of the current limit represented by the pre- and post-processing phases of the large image dataset should mainstream this methodology.

A Low-Cost and Unsupervised Image Recognition Methodology for Yield Estimation in a Vineyard.

Yield prediction is a key factor to optimize vineyard management and achieve the desired grape quality. Classical yield estimation methods, which consist of manual sampling within the field on a limited number of plants before harvest, are time-consuming and frequently insufficient to obtain representative yield data. Non-invasive machine vision methods are therefore being investigated to assess and implement a rapid grape yield estimate tool. This study aimed at an automated estimation of yield in terms of cluster number and size from high resolution RGB images (20 MP) taken with a low-cost UAV platform in representative zones of the vigor variability within an experimental vineyard. The flight campaigns were conducted in different light conditions and canopy cover levels for 2017 and 2018 crop seasons. An unsupervised recognition algorithm was applied to derive cluster number and size, which was used for estimating yield per vine. The results related to the number of clusters detected in different conditions, and the weight estimation for each vigor zone are presented. The segmentation results in cluster detection showed a performance of over 85% in partially leaf removal and full ripe condition, and allowed grapevine yield to be estimated with more than 84% of accuracy several weeks before harvest. The application of innovative technologies in field-phenotyping such as UAV, high-resolution cameras and visual computing algorithms enabled a new methodology to assess yield, which can save time and provide an accurate estimate compared to the manual method.

Development of Low-Cost Air Quality Stations for Next Generation Monitoring Networks: Calibration and Validation of PM2.5 and PM10 Sensors.

A low-cost air quality station has been developed for real-time monitoring of main atmospheric pollutants. Sensors for CO, CO2, NO2, O3, VOC, PM2.5 and PM10 were integrated on an Arduino Shield compatible board. As concerns PM2.5 and PM10 sensors, the station underwent a laboratory calibration and later a field validation. Laboratory calibration has been carried out at the headquarters of CNR-IBIMET in Florence (Italy) against a TSI DustTrak reference instrument. A MATLAB procedure, implementing advanced mathematical techniques to detect possible complex non-linear relationships between sensor signals and reference data, has been developed and implemented to accomplish the laboratory calibration. Field validation has been performed across a full "heating season" (1 November 2016 to 15 April 2017) by co-locating the station at a road site in Florence where an official fixed air quality station was in operation. Both calibration and validation processes returned fine scores, in most cases better than those achieved for similar systems in the literature. During field validation, in particular, for PM2.5 and PM10 mean biases of 0.036 and 0.598 µg/m³, RMSE of 4.056 and 6.084 µg/m³, and R² of 0.909 and 0.957 were achieved, respectively. Robustness of the developed station, seamless deployed through a five and a half month outdoor campaign without registering sensor failures or drifts, is a further key point.

An open-source and low-cost monitoring system for precision enology.

Winemaking is a dynamic process, where microbiological and chemical effects may strongly differentiate products from the same vineyard and even between wine vats. This high variability means an increase in work in terms of control and process management. The winemaking process therefore requires a site-specific approach in order to optimize cellar practices and quality management, suggesting a new concept of winemaking, identified as Precision Enology. The Institute of Biometeorology of the Italian National Research Council has developed a wireless monitoring system, consisting of a series of nodes integrated in barrel bungs with sensors for the measurement of wine physical and chemical parameters in the barrel. This paper describes an open-source evolution of the preliminary prototype, using Arduino-based technology. Results have shown good performance in terms of data transmission and accuracy, minimal size and power consumption. The system has been designed to create a low-cost product, which allows a remote and real-time control of wine evolution in each barrel, minimizing costs and time for sampling and laboratory analysis. The possibility of integrating any kind of sensors makes the system a flexible tool that can satisfy various monitoring needs.

CrossVit: enhancing canopy monitoring management practices in viticulture.

A new wireless sensor network (WSN), called CrossVit, and based on MEMSIC products, has been tested for two growing seasons in two vineyards in Italy. The aims are to evaluate the monitoring performances of the new WSN directly in the vineyard and collect air temperature, air humidity and solar radiation data to support vineyard management practices. The WSN consists of various levels: the Master/Gateway level coordinates the WSN and performs data aggregation; the Farm/Server level takes care of storing data on a server, data processing and graphic rendering; Nodes level is based on a network of peripheral nodes consisting of a MDA300 sensor board and Iris module and equipped with thermistors for air temperature, photodiodes for global and diffuse solar radiation, and an HTM2500LF sensor for relative humidity. The communication levels are: WSN links between gateways and sensor nodes by ZigBee, and long-range GSM/GPRS links between gateways and the server farm level. The system was able to monitor the agrometeorological parameters in the vineyard: solar radiation, air temperature and air humidity, detecting the differences between the canopy treatments applied. The performance of CrossVit, in terms of monitoring and reliability of the system, have been evaluated considering: its handiness, cost-effective, non-invasive dimensions and low power consumption.

Carbon dioxide and acetone air-sea fluxes over the southern Atlantic.

Measurements of CO2 and acetone fluxes have been made over a large-scale, naturally occurring high latitude phytoplankton bloom in the remote South Atlantic. Shipborne micrometeorological methods for direct atmospheric flux measurement have been applied to determine the direction and size of the CO2 and acetone fluxes. Previous results suggest that high latitude oligotrophic ocean regions are sinks of acetone, whereas high productivity regions are sources. The observed CO2 fluxes are into the ocean and on the order of 1 micromol m(-2) s(-1) at most. The acetone fluxes measured show a significant relationship with chlorophyll in the region of the phytoplankton bloom. Although the uncertainty is very high due to the very low signal-to-noise ratio, significant positive acetone mean fluxes of the order of 0.01 nmol m(-2) s(-1) have been observed in bloom areas, whereas near zero, negative, or highly variable low acetone fluxes have been measured elsewhere. Based on these results we estimate that the global acetone source from bloom affected areas is small in comparison to the uptake from the much larger oligotrophic regions, and that the ocean is globally a net sink for acetone.