Spatial accuracy of aerial imagery: Why it matters06/25/2019 | Crop Production Research, Weed Issues
By Hayden Walker, ISA Analytics intern, and Peter Kyveryga, ISA Analytics Director
Farmers today are using unmanned aerial vehicle (UAV) systems, or drones, to locate trouble spots in fields. Using imagery from a drone eliminates walking a field and the imagery encompasses the entire field so areas aren’t missed.
But sometimes spatial data sent from drone imagery can be inaccurate. Spatial accuracy is necessary for:
- developing detailed stand count maps,
- identifying weeds,
- pest and disease-affected areas for targeting spraying,
- marking drainage tile locations, and
- mapping compacted or high-traffic areas.
The Iowa Soybean Association’s (ISA) Analytics team is working on a project to find out how spatially accurate or inaccurate images from different systems can be and identify ways to correct them.
Aerial images collected by airplane and UAV systems are geo-referenced: the objects on the imagery can be projected to the exact locations on the ground. However, several factors can affect the accuracy of spatial imagery. Errors may arise during the stitching of individual images to form the final product, or the accuracy of the GPS used during the imagery collection process may be off. Other factors affecting accuracy include field topography and imagery resolution (pixel size).
This spring, the Analytics team installed ground reference points or markers at the ISA imagery calibration site near Collins in Story County. The markers are located inside the 200-acre field and on the perimeter of the field. The ground-truth locations of the markers were determined using a Topcon GR-5 GPS unit with sub-inch accuracy.
Markers were placed on the corners of six large and six small calibration tarps of precise reflectance values in shades of dark gray to white. The tarps have been in the exact location for several years and can be used as location reference points when comparing images for location accuracy.
When comparing the ground reference points from 2019 to the corners of the calibration tarps in 2018, the team saw a shift in the imagery of approximately 24 feet, a relatively large error (Figure A). When examining the imagery collected this year, the shift was only about 4.5 feet (Figure B). These are rough estimations as the imagery resolution should also be considered. The current industry standard for imagery accuracy is about 10 feet.
The Analytics team also learned that the deviations from the controlled geo-referenced points on the ground were not in the same directions in the 2019 imagery as in the 2018 imagery (Figure B), indicating that the errors are not systematic but likely a result of adjustment made during the imagery stitching process or random errors.
Controlled markers aren’t readily available for farmers so other methods will be needed to check spatial accuracy of aerial imagery. When comparing field images for location accuracy, farmers can use other spatial sources such as road or topographic maps, satellite or historical imagery and overlay them with the current images. As farmers collect their own imagery, they should consider these important elements of spatial accuracy for best results.
Contact Peter Kyveryga at email@example.com.
Additional Recommended Articles for this topic:
For media inquiries, permission to republish articles or to request high-res photos, please contact Katie James, ISA Public Relations Manager at firstname.lastname@example.org. © 2020 Iowa Soybean Association. All rights reserved.