choose object based classification option(s)

[wcornwell]

I tried to do a quick review of the options

Image analysis techniques have been developed and applied for ITDCD since the 1990s [15]. In general, these techniques can be divided into two categories, individual tree detection and individual crown delineation. The former extracts tree locations or treetops as point features, while the latter delineates individual crowns as polygon areas.

from https://link.springer.com/article/10.1007/s40725-023-00184-3

[wcornwell]

there are a lot of people using eCognition which i guess is an expensive one that we don't have. Example: https://www.mdpi.com/2504-446X/7/7/421

[wcornwell]

wild deserts approach: https://zslpublications.onlinelibrary.wiley.com/doi/full/10.1002/rse2.375

Rox is not into plants though:

In Phase 2, we took the identified living vegetation in the classified image and classified it into ‘species level’ (mulga species complex, Acacia ligulata, Mariana spp., Senna artemisioides complex, “other vegetation” and “ground cover”). “Ground cover” was unclassifiable vegetation cover at ground level, consisting of small grasses and forbs although immature individuals of the classified species may have fallen in this grouping. ‘Other vegetation’ encompassed other vegetation species that were not species of interest, such as Hakea leucoptera and Dodonea viscosa.

[wcornwell]

this is an amazing menu: https://github.com/satellite-image-deep-learning/techniques

[adelegem]

potential useful ones from https://github.com/satellite-image-deep-learning/techniques

• https://github.com/weecology/DeepTreeAttention
• https://github.com/jonathanventura/canopy
  which were both developed for tree species classification using hyperspectral imagery