EVALUATION OF SPECTRAL DATA FROM UAVS TO MEASURE SOIL PARAMETERS AND CROP YIELDS

• Sponsoring Institution: National Institute of Food and Agriculture
• Project Status: COMPLETE
• Funding Source: HATCH
• Reporting Frequency: Annual
• Accession No.: 1007388
• Project Start Date: Oct 1, 2015
• Project End Date: Sep 30, 2019
• Program Code: [(N/A)]- (N/A)

Recipient Organization
CORNELL UNIVERSITY
(N/A)
ITHACA,NY 14853

Performing Department
Crop & Soil Sciences

Non Technical Summary
Corn is grown for grain and silage by cash-grain & dairy farmers in New York on over 1 million acres of land. In debates on how to optimize estimates of corn yields, opinions range among continuing conventional measurements of soil properties, sampling at greater depth in the root zone, soil health testing, focusing on plant status as an integrator of the conditions that affect growth, and using precision-agriculture methods. Precision agriculture (PA), justified by its ability to address within-field spatial heterogeneity, is used increasingly to improve management toward maximizing profits.However, a number of authors have argued that the basic assumptions of current PA methodologies do not adequately account for the relationship between the nutrient status of soils, spectral properties of imagery, and crop growth pattern (e.g., Baveye and Laba, 2015). New approaches need to be explored, possibly using some of the new technologies that have become available in the last 10 years.Among these new technologies, Unmanned Aerial Systems (UASs) still have been used relatively little in agricultural applications (DeGloria et al., 2014). Nevertheless, the ease with which they allow repeated measurements at a generally low elevation with low ground speed and high spatial resolution can be taken advantage of to gain a better understanding of the relationship between plant yields, soil properties and spectral response patterns. The use of UASs for agricultural purposes is projected to become a multi-billion dollar industry in the next ten years. With one of six national UAS testing sites located in New York, farmers and Cornell Cooperative Extension educators have the unique opportunity to be on the cutting edge of evaluating this new technology.In this general context, the proposed research will involve farmers, researchers and extension agents in an initial evaluation of how UASs can be used to sharpen some of the technical assumptions on the basis of which PA is generally predicated. This information can improve and inform future corn management strategies and increase educators' ability to convey the value of UASs to farmers. We shall use a UAS to monitor bare soil before crop emergence and soil-crop cover of corn before harvest. We will also measure soil nutrient status and plant biomass before harvest. This data will be used to determine how UAS spectral measurements are correlated to soil characteristics and crop yield. These observations will determine when to use UASs, how to make use of the information they provide to improve corn management, and how to use this new technology to maximize crop yields.

Animal Health Component

100%

Research Effort Categories

Basic

(N/A)

Applied

100%

Developmental

(N/A)

Classification

Knowledge Area (KA)	Subject of Investigation (SOI)	Field of Science (FOS)	Percent
205	0110	2020	100%

Knowledge Area
205 - Plant Management Systems;

Subject Of Investigation
0110 - Soil;

Field Of Science
2020 - Engineering;

Keywords

spectral data

imagery

uav

precision agriculture

sensor

characterization

soil management

Goals / Objectives
This project seeks to evaluate the use of an innovative data collection device, UAS, to obtain process and distribute high spatial- and temporal- resolution imagery to relate spectral data with crop growth patterns and soil data to inform management strategies and predict crop yields. Potential benefits of this methodology may include reduction of the cost and time associated with soil and vegetation sample analyses; reduced labor costs of monitoring on large farms; and reduced inputs of fertilizer and pesticides, providing decreased environmental impacts.Specific objectives are to:1. Determine how successfully UAV-collected spectral data on bare soil spectral data predicts soil moisture and organic matter.2. Determine whether UAV-collected spectral data is more highly correlated with organic matter in conventional depth (5-cm depth) soil samples versus organic matter in deeper root zone samples.3. Determine how successfully UAV-collected spectral data on corn vegetative cover predicts crop yield.4. Evaluate conventional soil line and vegetative indices and soil-modified vegetative indices developed from UAV-collected spectral data.

Project Methods
Our primary source of spectral data will be a Lancaster Hawkeye III UAS airborne sensor that collects data in the visible (8mm/pixel) and multispectral (2.7cm/pixel; blue, green and near-infrared) ranges. Its sensing platform is comprised of off-the-shelf cameras in which standard filters have been replaced with other filters of the same optical thickness. To better understand the spectral data being collected, we will use an OceanOptics (owned by the Cornell Dept of Civil & Environment Engineering) to determine sensor specifications (e.g., camera and filter characteristics), including focal length, filter types and behavior, radiometric resolution, field of view, and angle of viewing.We will scan 4 corn fields in western NY that have FAA approval for UAV flights during the 2016/17/18 growing seasons. To correlate the spectral response to soil and plant biomass, imagery will be collected prior to planting (bare soil) and prior to harvest (maximum canopy). All images will be geo-referenced to the Universal Transverse Mercator (UTM) coordinate system, Zone 17, North American Datum of 1983 (NAD83). Four stationary markers will be placed around each cornfield. Images containing these markers will be geo-referenced to the GPS coordinates of the markers. All other images will be mosaicked to those base images using the geo-processing capabilities available within ArcGIS10.3.1.When working with remotely sensed data, the ideal sampling unit is the individual pixel. However, the accuracy of a GPS unit combined with the positional accuracy of the UAS itself could add a potentially high level of uncertainty in locating individual pixels on the ground. Therefore, we will try to use field plots that encompass the rooting zone of the plants being sampled and are, at least, four times the minimum cell size of the UAS imagery (6 cm2 multiplied by 4 ~ 29 cm2). As an added precaution, onscreen digitizing of an ARCVIEW shape file of field plots will be created using Digital Orthophoto Quarter Quads (DOQQ), GPS points, and visual inspection of the plots. The digitizing will be done on site when possible. The shape and size of the sample plot will be estimated and recorded. Minimally, two points will be recorded. Field plots will be overlaid on the image to verify their spatial and compositional accuracy. GPS point data will be collected using a Trimble Geo Explorer XH with sub-foot post correction accuracy.To complement the UAS-acquired data, we will obtain hyperspectral data using a proximal sensor, the ASD FieldSpec Pro (owned by the Cornell Section of Soil & Crop Sciences) at each soil/vegetation sampling point, with the sensor suspended above the center of each sampling location. For each sampling date, ASD data will be collected in a consistent manner. The ASD is a spectroradiometer able to acquire data from 350 nm to 2500 nm, via three sensors (Visible/Near Infrared sensor for 350 - 1050 nm; and two Short-Wave Infrared sensors for 900 - 1850 nm and 1700 - 2500 nm; the controlling software automatically accounts for the overlap in wavelength intervals) (Analytical Spectrum Devices, Inc., 2002). Proximal hyperspectral data has been used for determining soil characteristics in agricultural applications for some years (Mulder et al., 2011; DeGloria et al., 2014). The hyperspectral data responds in a measurable manner to soil moisture and to organic matter. In the visible range, soil moisture and organic matter both darken the image, but it is not possible to differentiate the cause of the darkening without further data. In this application, we will compare the results from the UAS-acquired data with the results of the ASD FieldSpec Pro to investigate the ability to sense soil moisture and organic matter content with the UAS-acquired data.The plant composition, soil properties and landscape features at each sampling location will be characterized and recorded. Twenty stratified random bare soil samples at a 5cm depth will be obtained in each of the four fields. Two sets of soil samples, at the same locations that were sampled for bare soil, will be gathered prior to harvest. After a 5cm depth sample has been acquired, soil samples at different depths in the root zone of individual plants, which are closest to the original sampling locations, as defined by the permanent marker, will be gathered in each field. Root zone samples will be composited to obtain a statistical average of the soil composition in the root zone, which may be very different than the same parameter right at the soil surface as viewed by remote sensing equipment.Standard soil analyses for N,P,K content, pH, organic matter and humidity level will be performed on all soil samples. Above-ground biomass measurements will be made on the plants associated with the 20 plant-samples (Caldwell et al., 2014; Liu et al., 2010) and belowground biomass, corn roots will be sampled (Mazilli et al., 2015). For the same plants, and in parallel of the sampling of the soil in depth, we shall also determine the spatial pattern of propagation of roots.Imagery will be compared to soil sample test results, field yield and individual plant/root biomass measurements to determine if there is a correlation between spectral signatures and plant biomass, soil nutrient status or yield. A regression analysis between spectral values within different regions of the UAS-acquired data (predictor variable) and soil properties and plant/root biomass measurements (response variables) will be performed using the linear model function in R (R Core Team, 2014). Combinations of predictor variables (spectral responses in different areas of the spectrum) that produce the best models will be selected from the larger pool of spectral information through a stepwise regression. Once the best models are identified, a final "best" model that has an adequate fit while using the fewest parameters will be selected by implementing the Akaike Information Criterion.Soil line analysis and vegetation indices will be calculated to determine the validity of these tools when applied to UAV-collected data (Yoshioka et al., 2010). Vegetation indices (generally a ratio of red and near-infrared (NIR) spectral regions) have been used with satellite remotely-sensed data, and have undergone efforts to minimize factors such as atmospheric variations and soil brightness (Ustin et. al., 1999; Gitelson et al, 2002; Prabhakara et al., 2015). Other ratios will be tested for use with this platform, which does not collect in the red. UAV-acquired data will be examined for advantages due to low altitude (potentially avoiding atmospheric variations) and the ability to observe vegetative cover and soil exposure (Geipel and Claupein, 2014). We will examine the ability of these techniques to provide information about the vegetation and soil in the low altitude, high spatial-resolution imagery.The Lancaster Hawkeye system does not include spectral data processing software. We will explore the use of open-source data (such as VisualSM or MS Photosynth) to process the raw imagery, allowing creation of custom mosaics, stereo images, and ability to make spectral measurements with the UAV-acquired data. Open-source data and processing procedures will be readily available to users who wish to have more control of the processing or make more innovative use of the data.Evaluation: Indicators of success will be completion of a study that allows us to state clearly the capacity and limitations of the use of UAV-acquired data for crop and soil management; to provide guidance on both acquisition and use of these data; and to provide methods for interpreting data acquired from the Lancaster Hawkeye III UAS sensor.Milestones:Characterization of the sensor.Successful georeferencing methods.Acquisition of imagery and biomass/ soil samples each season.Interpretation of data each year.Final interpretation of data.Results summarized and submitted for publication.

Progress 10/01/18 to 09/30/19

Outputs
Target Audience: Nothing Reported Changes/Problems:The original proposal for image acquisition was predicated on the participation of a Cornell Cooperative Extension (CCE) educator qualified to fly an unmanned aerial vehicle owned by CCE. The educator had agricultural producers committed to allowing flights over their corn and broccoli fields, as well. The educator resigned from CCE prior to the project start date, placing us in the position of finding a qualified pilot. We were able to piggy-back on another 2-year project that is looking at the use of aerial and proximal imagery for detecting differences in vegetation due to varying N application. We assisted with the project in exchange for access to the aerial imagery obtained by unmanned aerial vehicle. The imagery is of research plots rather than fields. The change in scale will affect the investigation potential, including the process of stitching images together and ground control. In the summer of 2017, the end of the second year, we were able to collaborate with other faculty on a suitable field site, a privately-owned 38-acre field planted in corn. We sub-contracted with an external pilot to fly imagery, although that entailed gaining the collaboration of the Rochester Institute of Technology as well to have access to an appropriate camera/sensor. We completed 2 flights in 2017.In 2018, the group of faculty working at the privately-owned site moved to the Cornell Musgrave Farm, Aurora, NY. This field site is approximately 40 acres, planted in corn, and will allow us to remain at the site for several years. For flights over this site, we contracted directly with the Rochester Institute of Technology and completed three flights. Some data processing and analysis has been undertaken, but the late acquisition of imagery due to the issues of pilots and cameras/sensors has severely delayed data analysis. Due to these delays, we have received a no-cost extension on the project. Finally, downloading data from RIT servers to Cornell servers has failed numerous times due to the volume of data. Transferring data on external hard drives is more reliable but extremely time consuming and has taken a great deal of manual handling of data. What opportunities for training and professional development has the project provided?We have shared data with Cornell colleagues for their research purposes. Colleagues at RIT have access to data, as well. How have the results been disseminated to communities of interest?We have shared data with colleagues for their research purposes via Cornell Box. What do you plan to do during the next reporting period to accomplish the goals?We will complete analysis of data and consideration of software packages for processing data.

Impacts
What was accomplished under these goals? During Year 4, we re-loaded data from several flights in an attempt to capture data lost during downloads. We are considering recommendations or guidance on handling the volume of data, a major hurdle identified during the field work. This will be a drawbackfor most smaller growers. We have continued analysis of the imagery compared to proximal data, real-time yield data, and soil analyses.

Publications

Progress 10/01/17 to 09/30/18

Outputs
Target Audience:CCE educators and faculty colleagues learned about the work through working meetings. Changes/Problems:The original proposal for image acquisition was predicated on the participation of a Cornell Cooperative Extension (CCE) educator qualified to fly an unmanned aerial vehicle owned by CCE. The educator had agricultural producers committed to allowing flights over their corn and broccoli fields, as well. The educator resigned from CCE prior to the project start date, placing us in the position of finding a qualified pilot. We were able to piggy-back on another 2-year project that is looking at the use of aerial and proximal imagery for detecting differences in vegetation due to varying N application. We assisted with the project in exchange for access to the aerial imagery obtained by unmanned aerial vehicle. The imagery is of research plots rather than fields. The change in scale will affect the investigation potential, including the process of stitching images together and ground control. In the summer of 2017, the end of the second year, we were able to collaborate with other faculty on a suitable field site, a privately-owned 38-acre field planted in corn. We sub-contracted with an external pilot to fly imagery, although that entailed gaining the collaboration of the Rochester Institute of Technology as well to have access to an appropriate camera/sensor. We completed 2 flights in 2017. In 2018, the group of faculty working at the privately-owned site moved to the Cornell Musgrave Farm, Aurora, NY. This field site is approximately 40 acres, planted in corn, and will allow us to remain at the site for several years. For flights over this site, we contracted directly with the Rochester Institute of Technology and completed three flights. Some data processing and analysis has been undertaken, but the late acquisition of imagery due to the issues of pilots and cameras/sensors has severely delayed data analysis. Due to these delays, we have received a no-cost extension on the project. What opportunities for training and professional development has the project provided?PI Grantham collaborated with colleague Senior Extension Associate Susan Hoskins to establish a Geospatial Sciences Program Work Team and to begin professional development for CCE educators through that structure. We held three events in 2018 to familiarize CCE educators with UAS procedures, data analysis techniques, and geospatial data available from CUGIR. How have the results been disseminated to communities of interest? Nothing Reported What do you plan to do during the next reporting period to accomplish the goals?During Year 4, we will be considering recommendations or guidance on a major hurdle identified during the field work. The volume of data is enormous and simply downloading the data from the UAS memory to external hard drives or to a server is very time consuming. In some cases, data was lost and had to be refreshed from the original memory. This will be a drawback for most smaller growers. In Year 4, a no-cost extension year, we will continue analysis of the imagery compared to proximal data and soil analysis.

Impacts
What was accomplished under these goals? Research Objectives 1 through 4: In Year 3, we re-located to a field site on Cornell property, greater than 30 acres, that was planted in corn. We collected the same data with the same instruments and platforms as in Year 2. Ground control markers were located with a handheld GPS unit. We obtained data at 200- and 400-feet elevations on bare soil before planting, at approximately V7 and 50 cm high, and on mature corn, just prior to harvest. At each flight event, we collected proximal hyperspectral data at ground control points. We collected soil samples at ground control points at the time of the bare soil flight. We have made progress in image analysis and evaluating the functions of some software packages for this purpose. Images collected for Bowman Farm, Aurora, New York in August 2017 were used to develop a mosaicking protocol in Pix4D. The twenty-three GPS points collected in the field were also used in Pix4D to geo-reference the images to UTM zone 18, WGS84. Mosaics have successfully been produced for each of 5 bands collected by the UAS sensor. We are currently working on streamlining the process and moving the output into ArcGIS. We have continued to search the literature, and have identified additional open source software that we will use to compare results with software accompanying the unmanned aerial system. Research Objectives 1 through 4: In Year 3, we re-located to a field site on Cornell property, greater than 30 acres, that was planted in corn. We collected the same data with the same instruments and platforms and in Year 4, a no-cost extension year, we will continue analysis of the imagery compared to proximal data and soil analysis.

Publications

Progress 10/01/16 to 09/30/17

Outputs
Target Audience:Year 2 research was conducted on a private farmer's property and he has requested summaries of our results. We have worked with him on ground control and to receive yield data. Once we have finished analysis, we will provide him with our results. A group of CCE educators has received briefings and previewed some educational materials as part of the Smith-Lever funded project on which PI Grantham and colleague Hoskins are collaborating. Changes/Problems:As stated earlier, the size of the research plots used in Year 1 restricted our efforts. The Year 2 field of 38 acres was an excellent field site. That site will not be planted in corn this coming season so we will work in a Cornell field of over 30 acres for Year 3. The fields are geographically close, so soil types are similar and the Cornell field will be planted in corn. What opportunities for training and professional development has the project provided?A group of CCE educators has received briefings and previewed some educational materials as part of the Smith-Lever funded project on which PI Grantham and colleague Hoskins are collaborating. PI Grantham attended the 2017 ASPRS conference that focused on UAS development, including one workshop that over-viewed a number of software packages. How have the results been disseminated to communities of interest?A group of CCE educators has received briefings and previewed some educational materials as part of the Smith-Lever funded project on which PI Grantham and colleague Hoskins are collaborating. Several colleagues at Cornell and in the private sector have been invited to the Cornell Box to access data acquired in this project. Colleague Hoskins has used some of the imagery in training for CCE educators working with youth and for exercises being developed for 4-H Youth Development programming. Another colleague, Professor Philpot, has downloaded images to compare with other imagery he is working with. What do you plan to do during the next reporting period to accomplish the goals?Research Objectives 1 through 4: The Year 2 field site will be in a soybean rotation in the third year, so we have identified a field on Cornell property, greater than 30 acres, that will be planted in corn on which to conduct research in Year 3. We will collect the same data with the same instruments and platforms and continue analysis of the imagery compared to proximal data and soil analysis. Based on the Year 2 experience, we will modify the physical aspects of the ground control to reduce the labor of installation and removal.

Impacts
What was accomplished under these goals? Research Objectives 1, 2, and 3: Although images obtained in Year 1 can be compared to proximal hyperspectral data (ASD FieldSpecPro) and soil data, the research plots used in year 1 were small enough to be captured in one image, even at low altitude. The size of the plot also prevented meaningful ground control tests. Therefore, in the second year, we located a privately-owned 38-acre field planted in corn. The size of the field provided soil moisture and drainage variability and a good test of ground control. We designed and established of ground control prior to flights; developed a flight plan and management that included safety measures such as visual observers, job hazard assessment, safety briefings, and personal protection equipment; support for radiometric calibration using ground control panels and calibration panels; delivery of data; and data processing following flights. This plan can be a model for others at Cornell and in the agricultural industry. Review of recent literature, which compared of the number and placement of ground control for UAS photogrammetry, indicated that five ground control points in a rectangular study area provided good results, with diminishing additional accuracy as more ground control was added [1,2]. Placement near the four corners with a ground control point in the center was shown to provide good results for horizontal and vertical accuracy. Based on this, GPS data was taken at twelve points and semi-permanent monumentation was installed in the field site. 18" pieces of reinforcing rod (rebar) were driven into the ground for the semi-permanent monumentation and orange rebar caps were put on top so that the control points could be visually located after the corn was cut. Twelve points were placed as shown in Figure 2 so that the number of test points available for quality assurance verification and accuracy calculation was comparable to the number of ground control points. Position data were collected with a mapping-grade GPS receiver, and accuracy was calculated as 15 cm at a 94% confidence level. These points were signalized with white, 5-gallon bucket lids nailed to the end of 6' wooden stakes that were driven into the ground next to the rebar. Raising the lids off the ground was required to make them visible from the air in the fully grown corn. GeoPDF maps were generated in ArcGIS so that the study team could return to the ground control points to capture radiometric data on flying days. Radiometric calibration panels were also deployed on flying days to support that activity. Three complete flight collections were conducted over four flying days. We obtained data at 200- and 400-feet elevations on mature corn, just prior to harvest, and on bare soil following harvest. The MicaSense RedEdge camera collected data at spectral bands blue, green, red, red edge (region of rapid change in reflectance of vegetation in the near infrared (IR) range of electromagnetic spectrum), and near-IR. At each flight event, we collected proximal hyperspectral data (400 nm to 2400 nm) and soil samples at ground control points. In addition, the landowner provided yield data after harvest. Imagery is being analyzed using Pix4D software and we will compare those efforts and results to other software packages. The images for one flight of the Year 2 field are georeferenced and merged together into one complete image by band, 5 separate georeferenced images for each flight. 1 Abdullah, Q, 2017. "Mapping Matters, Your Questions Answered," Photogrammetric Engineering and Remote Sensing, 83:4, April 2017, American Society for Photogrammetry and Remote Sensing. pp 255-260. 2 Day, D, W. Weaver, 2017. "Ground Control Configuration Analysis for Small Area UAV Imagery Based Mapping, Presented at the American Society for Photogrammetry and Remote Sensing Imaging and Geospatial Technology Forum (IGTF), 15 March, 2017. Research Objective 4: A filing system has been developed for the data on the Cornell Box (not a dropbox function) and several colleagues have been invited to share that data. Imagery is sorted by type (ground control, hyperspectral, imagery, GPS, and soil analysis), date of acquisition, and altitude of imagery. We have made progress in image analysis and evaluating the functions of some software packages for this purpose. We have continued to search the literature, and have identified additional open source software that we will use to compare results with software accompanying the unmanned aerial system.

Publications

Progress 10/01/15 to 09/30/16

Outputs
Target Audience:With the resignation of Educator Bill Verbeeten from Cornell Cooperative Extension, we do not have the farmer participation that we originally anticipated. However, we have worked with Cornell Cooperative Extension of Oneida County to develop a NE SARE proposal for training on the use of unmanned aerial systems in agriculture, for farmers and their advisors. That proposal is in review. PI Grantham with Susan Hoskins, Cornell Institute for Information Sciences, were awarded a Smith-Lever proposal to develop curriculum on geospatial sciences for Cornell Cooperative Extension educators. It will complement the training that Cornell Cooperative Extension of Oneida County hopes to offer and we will continue to work with them to reach both educators and farmers. In addition, PI Grantham was invited by ASD, Inc., to present by webinar on the project and the issues that are being investigated. The presentation was held on September 27, 2016, and had 168 participants from across the US and abroad. The majority of the participants were researchers, agricultural consultants and producers, and educators who were interested in learning how to use geospatial sciences in their work. Changes/Problems:The original proposal for image acquisition was predicated on the participation of a Cornell Cooperative Extension (CCE) educator qualified to fly an unmanned aerial vehicle owned by CCE. The educator had agricultural producers committed to allowing flights over their corn and broccoli fields, as well. The educator resigned from CCE prior to the project start date, placing us in the position of finding a qualified pilot. We were able to piggy-back on another 2-year project that is looking at the use of aerial and proximal imagery for detecting differences in vegetation due to varying N application. We assisted with the project in exchange for access to the aerial imagery obtained by unmanned aerial vehicle. The imagery is of research plots rather than fields. The change in scale will affect the investigation potential, particularly ground control. What opportunities for training and professional development has the project provided?Senior Research Associate Laba worked with an intern in Professor Ketterings group in the summer, 2016, to help her understand the science behind the imagery and interpreting imagery with soil and yield data. PI Grantham collaborated with Cornell Cooperative Extension of Oneida County on a proposal to develop unmanned aerial systems curriculum. PI Grantham presented on September 27, 2016 on an ASD, Inc., hosted webinar for 168 participants. How have the results been disseminated to communities of interest? Nothing Reported What do you plan to do during the next reporting period to accomplish the goals?Research Objectives 1 and 2: We will continue the original soil sampling regimen through the next growing season: bare soil prior to N application and soil samples immediately after harvest. We will compare soil indices obtained with UAV-collected spectral data with soil analysis. Research Objective 3. We will compare measured yield to yield predicted by vegetation indices obtained from the unmanned aerial systems. Research Objective 4. We will compare using open-source software, commercial software independent of the unmanned aerial system, and proprietorial software accompanying the unmanned aerial system for calculating soil and vegetation indices.

Impacts
What was accomplished under these goals? Research Objectives 1, 2, and 3: We have identified a research group with which to collaborate, led by Professor Quirine Ketterings and Professor Elson Shields. These experiments are being conducted on corn alone. We developed a ground control system for the research plots, obtained airborne imagery of the plots at bare soil, and obtained proximal imagery of bare soil using the ASD FieldSpecPro, a hyperspectral radiometer. We sampled and analyzed the soil prior to N application and at the time of obtaining airborne and proximal imagery. The research group with whose project we are collaborating obtained GreenSeeker data, also proximal. Imagery of the research plots was obtained throughout the growing season. Professor Ketterings group obtained yield data at the end of the season. We designed and conducted some experiments on unmanned aerial vehicle-borne camera calibration in which we qualitatively compared camera outputs to outputs of a hyperspectral radiometer, Ocean Optics. Those experiments confirmed the published camera wavelength range. Research Objective 4: We have conducted literature searching, and identified and obtained some open source and proprietorial software that we will use to compare results with software accompanying the unmanned aerial system.

Publications