Can anyone please help with the process where area can be considered during land use classification? For e.g. If you want to classify forest, you need to take into account 0.5 ha area and 10% canopy cover. how do we set those criteria during classification? Is it under object based classification? I am classifying green space of urban area and want to consider only more than 10 sq m. If you have any ideas, I would sincerely appreciate.
Thank you
Hi Anisa,
I'd start out with the following steps:
I hope I understood your problem correctly and that this helps.
Best,
Fabian
Dear Anisa,
I do not know the classification algorithm or method that you intend to use, but I suggest that you classify your image to different land-use classes first, then you can use any analysis software to select the classes based on the criteria you set. You cannot exclude areas with less than a specific threshold unless you do classification first.
Some suggested clustering, but you cannot apply any criteria (forests less than 0.5 ha as an example) to exclude some clusters unless you identify them or label them first.
Here you may consider segmentation algorithms for image processing, e.g., https://medium.com/@ODSC/processing-images-through-segmentation-algorithms-2cf1b1e7ad92
Thank you everyone for the suggestions. I truly appreciate. So, what I understand is I need to classify the image first to filter it by area afterwards. In that case, I would classify all green areas as green space using traditional image processing and then afterwards fit my criteria into it. Then will I be able to include the pixels which doesn't fit into my criteria as a separate class?
Regards
For non-overlapping classes, criteria for each class should be mutually exclusive w.r.t. other classes. For a color based classification, color needs to be quantized first into a specific number of bins, say 36, and then each of these bins can be defined as a set of 10 degree angle values in a HSV color model representation.
- Badges
- Science topic