Hi All,
Can the Magic Tool be used with multispectral imagery to develop spot-spraying or variable-rate application maps?
The use case is detecting thistles within clover—a green-on-green target. While high-resolution RGB imagery at approximately 0.5–1.5 cm GSD may identify individual plants, manually—or using the Magic Tool—scouring approximately 100 hectares at a 1 m × 1 m scale is not practical.
Can multispectral data distinguish thistles from the surrounding clover and automate target detection at this scale? If so, what workflow would you recommend?
Hi,
If multispectral data can help distinguish the thistles, depends if the multispectral data can create a unique spectral signature of the thistles. There is an “easy” way to find out.
On your dataset, you annotate the thistle patches as well as non-thistle patches with two sets of polygons. Then create single band indices and copy the annotations on each band layer. Use the statistics export to create a CSV per spectral band and feed those into any AI tool of your choice. Let it analyze whether there is a difference between those two sets.
If there is, it might even be able to suggest a vegetation index formula sensitive for that specifically.
If there is a spectral feature that can be exploited, the Magic Tool might be able to find it, but it will not change the workflow of how you use the Magic Tool currently. We will come up with an improvement in the future where you are able to save settings in the Magic Tool that would then solve what you are asking for.