Location: Hydrology and Remote Sensing LaboratoryTitle: Remote sensing with simulated unmanned aircraft systems for precision agriculture applications Author
Submitted to: IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 4/8/2014
Publication Date: 6/16/2014
Publication URL: http://handle.nal.usda.gov/10113/62567
Citation: Hunt Jr, E.R., Daughtry, C.S., Mirsky, S.B. 2014. Remote sensing with simulated unmanned aircraft systems for precision agriculture applications. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing. 7(11):4566-4571.
Interpretive Summary: Precision agriculture is a group of technologies based on global positioning systems for managing areas within a field differently, depending on crop requirements. By applying a lower rate of fertilizer on areas with more fertile soils, less total fertilizer is applied, reducing costs to the farmers and reducing impacts on water quality. Using unmanned aircraft systems (UAS, also called drones or UAVs) for agricultural remote sensing has potential advantages by acquiring images with smaller pixel sizes at much lower cost. For safety, UAS flights are currently restricted in the USA by the Federal Aviation Administration, but the restrictions may be lifted for agriculture within a few years. At the Beltsville Agricultural Research Center, we conducted a fertilization experiment using cereal rye planted as a winter cover crop. We obtained true-color and color-infrared images using a camera mounted looking down from a tall pole to simulate data obtained from UAS. We found that estimates of plant cover and leaf chlorophyll content were more accurate using very small pixel sizes (0.5 mm) compared to using larger pixel sizes (0.7 by 1 m). Usually, soil properties vary within a field at spatial scales from 5 to 20 m, so crop fertilizer requirements will also vary at this same scale. Furthermore, it is relatively easy to acquire and process remotely sensed imagery with pixel sizes at this spatial scale. Our results suggest that better information for crop fertilizer requirements may be obtained with pixel sizes ten thousand times smaller. Processing image data with very-small pixel sizes is difficult; however, it is not necessary to produce an image covering the whole field by stitching together hundreds to thousands of low-altitude photographs. Instead, photographs could be analyzed as if they were small sample plots obtained along transects across the field. Computer vision algorithms will be necessary to extract automatically information from so many photographs.
Technical Abstract: An important application of unmanned aircraft systems (UAS) may be remote-sensing for precision agriculture, because of its ability to acquire images with very small pixel sizes from low altitude flights. The objective of this study was to compare pixel sampling with plot-scale metrics for the remote sensing of biomass and chlorophyll content. Cereal rye (Secale cereal) was planted at the Beltsville Agricultural Research Center for a winter cover crop. In a randomized block design with four replicates, plots of cereal rye were fertilized in the fall with different nitrogen rates, which produced differences of biomass. In the spring, plots were fertilized again with different nitrogen rates, which produced differences in leaf chlorophyll content. UAS imagery over the fertilization experiment was simulated by placing a Fuji IS-Pro UVIR digital camera at 3-m height looking nadir over each plot. An external UV-IR cut filter was used to acquire true-color images; an external red cut filter was used to obtain color-infrared-like images with bands at near-infrared, green and blue wavelengths. We averaged all of the pixels in the center portion (0.7 by 1.0 m, with pixel sizes of about 0.5 by 0.5 mm) of each photograph and calculated plot-scale spectral indices. The Green Normalized Difference Vegetation Index was correlated with dry aboveground biomass (r = 0.58); however, the Triangular Greenness Index (TGI) was not correlated with leaf chlorophyll content. We used the SamplePoint program to select 100 pixels in each true-color and color-infrared image, and for each pixel, we visually identified the cover type and acquired the digital numbers. The number of rye pixels in each image was better correlated with biomass (r = 0.73), and the average TGI from only well-illuminated leaf pixels was negatively correlated with leaf chlorophyll content (r = -0.72). These results suggest that better information on crop nitrogen requirements may be obtained using very small pixel sizes, which are possible from imagery acquired at very low altitudes. However, the spatial scale for crop nitrogen application is about 5 to 20 m. It may not be necessary to geospatially register the large numbers of photographs with very small pixel sizes, in order to provide a synoptic view of a field. Instead, each image could be analyzed as a single plot in transects across a field.