I've been trying to run the 'random trees' classifier using training polygons (approx 600 training polygons across 8 classes) on a large (9 Gb+) hyperspectral image (426 bands) and keep getting the 'ERROR: 999999' code and I'm wondering if the classifier just can't handle such a large image? Since according to the documentation it sounds like the classifier is designed for an RGB image, I use the RGB version of the image as the 'in_raster' and use the large 426 band hyperspectral image as the 'in_additional_raster'. I was able to run the classifier on the RGB image alone, but as soon as I try using the large image the dreaded 'ERROR:999999' shows up.
Has anyone used this classifier on a large image like this and had any success? I've tried switching the order of the input rasters, changed the format of the input rasters (from .tif to .img), tried in desktop and Pro, etc. I also tried the same image and training polygons using the SVM classifier and get the same error. Please help!
Hi Andrew, thanks for reporting this issue. We are more than happy to work with you and get your hyperspectral image (426 bands) working in our image classification tool.
First of all, may I know please which hyperspectral data did you use and saw this issue? If you could share with us a small area of your data (e.g., 1000 by 1000, but has to keep same number of bands) for testing the fix of this issue, that will be great. If you could not, it is also fine, we will try to find similar hyperspectral data for testing.
You could contact me directly through my Esri email: Ling_tang@esri.com. Thanks and I am looking forward to your reply.