Adriaan van Niekerk
May 15, 20191 min
Updated: May 23, 2019
One of my MSc students, Barry Watkins, is working on a technique to automatically extract agricultural field boundaries from Sentinel-2 imagery. Trying to extract field boundaries from satellite imagery is not new, but most previous attempts made use of very high resolution (VHR) imagery (e.g. WorldView-2, GeoEye, SPOT-5). It is a bit more challenging using lower (10m) spatial resolution imagery. However, Barry’s approach is to use edge detection on a time series of Sentinel-2 images, which enhances the persistent boundaries (Figure 1). Watershed segmentation is then applied to the resulting multi-temporal edge layer to generate polygons representing individual fields. The technique works pretty well (Figure 2), with overall classification accuracies of more than 90% and mean spatial errors below 25m! For more details, check out the paper recently published in Computers and Electronics in Agriculture. Barry is currently applying this technique in various areas to test its transferability. I will keep you updated!
This article was originally published on www.remotesensing.blog, which is devoted to articles about the use of remote sensing and other geospatial technologies for agricultural and related applications.
See related posts:
What exactly is remote sensing?
These are the bands farmers should like
Using free satellite imagery for yield benchmarking and predictions