Hi everyone,
I would like to ask for your opinion about the accuracy assessment tool on ArcGIS Pro. I am currently using the software to run drone image classification. I have done the accuracy assessment manually (from selecting random points to calculating confusion matrix on excel) and used the accuracy assessment tool on the software.
Pretty sure I am doing it right manually (I ran spatial autocorrelation test on my randomly selected points to make sure they were really randomly distributed as well), BUT still tried to do the accuracy assessment on the software with the thought of avoiding questioning on the project’s method since this tool is readily available and what’s the point of not using it. Now that both methods gave me a slightly different trend on results, I can’t decide which is the method I should go for.
I also realized I was only able to create accuracy assessment points on classified image and still have to identify those points visually on the software, in my point of view what it does is only create random point feature for me that has an attribute table I can edit and use it for confusion matrix calculation tool on the software. I wanted an even number of points for each classes I have so I used the “equalized stratified random” sampling strategy, but since those points are created from a classified image, there might be misclassification and ended up with an uneven number of points at last. It also takes a longer time to do the whole assessment on the software than on excel (I have my spreadsheet built to calculate confusion matrix for me once I key in classification agreement).
At this point, (1) I wonder if I am using the tool correctly. (2) What’s your thought on using exact even number of points for each classes. (3) Do you think what the software gives is reliable and worth the time? (4) If you have used that tool before, can I have your feedback?
Thank you.
- Badges
- Science topic