Automatic detection and quantification of invasive shrubs

Hi everybody,

I am trying to use Pix4DFields to automatically detect invasive shrubs in a prairie using RGB and multispectral orthomaps and quantify their numbers and coverage. It is not as easy as to distinguish between grasses/forbs vs. shrubs, as there are also native shrubs. Some types of shrubs can be identified by their color or by their extended leaf phenology. I have created RGB and multispectral maps. One type of invasive shrub has silvery (blue-green) leaves and can be distinguished on RGB, especially late in fall when most other shrubs have shed their leaves. It is even easier to do so in the multispectral maps based on the NDVI index and maybe the MCARI, but I think this can be possibly improved with a custom index tailored to a specific wavelength range of the invasive shrub?

Does anybody have any suggestions about improving the indexing, and how I can then quantify the number of shrubs, distribution in different size classes, and percentage of cover relative to the entire prairie parcel?

Thanks a lot in advance!

Thomas Wassmer, Professor of Biology, Siena Heights University

Hi,

could you share screenshots of some examples? The MagicTool is actually built to handle those tasks well and we have users who specifically use it for invasive species mapping and targeted spraying.

If you have enough spectral information, you can try building custom vegetation indexes.For this, it would make sense to annotate a few of those shrubs by hand and then extract from the individual spectral bands the raw values as statistics for those manual annotations. Then, for example, you can use any state-of-the-art LLM to let it crunch the numbers and find a correlation between the band values and the shrub you want to detect and let it come up with a vegetation index you can then calculate in Pix4DFields.

That would be my very hands-on suggestion to start with this topic.

  1. See if the MagicTool does the trick

  2. On a second approach, investigate if there is a correlation with the spectral data you already have

Thanks Julius. I have attached a RGB and MS map of some of the area from 11/6 this year. Most of the remaining foliage is from invasive shrubs. However, there are also a few evergreen coniferous trees and maybe still a few native shrubs. In addition, I would like to distinguish the two main invasive shrub genera, one of which has silvery-bluish leaves, the other more grass-green foliage.

I have tried the magic tool, and it does get the larger shrubs but misses smaller ones and cannot distinguish between the shrub types.

In addition, I would like to also analyze the size distribution into a few categories, such as juvenile (smallest), adult (intermediate), and mature (largest) shrubs.

Thanks in advance!

Tom

RGB map

MS Map with Annotation (Magic Tool)