Micasense Altum LWIR processing

PIX4Dmapper
1

We’re deciding between processing LWIR separate from the 5-band reflective multispectral or doing it together. Sometimes we’ve been able to process them together – seemingly when there’s enough thermal detail. But we’re not clear about how GCPs might be applied when LWIR is combined with multispectral vs separate. In either case, the GCPs are not visible in the 80-cm LWIR band, but we aren’t clear whether Pix4D performs some magic to make the matching more accurate when combined.

In terms of working with the data after the fact, it doesn’t matter; we can either combine the bands with the LWIR or not. But we are just trying to find out whether there’s any difference in accuracy.

Our study site is a large meadow flown at 120m, where we get 1 km^2 per flight plan (3 flights with changed batteries at two locations).

2

Hi, If you process the thermal images separately, they might fail to calibrate as the resolution is very low. The LWIR is included in the altum rig in our database. Pix4D uses the green band images to process all the other band images (including thermal) as the green band has lots of features/variability in case of vegetation. For marking GCPs, we recommend marking teh GCP in all the bands or marking it only in the reference band (which is green by default).

3

Thanks for this information, very useful to know! But I’m not quite sure how best to proceed, since we usually run into errors that cause the processing to fail when we include LWIR. Also, it’s impossible to see our GCP targets (most of which are 2’x2’ tarp sections with duck tape X, either white tape on blue tarp or black tape on white tarp). So my process has been to run it separately, which has been working reliably, though only the 5-band is processed with GCPs.

In terms of calibrating, surprisingly the LWIR is calibrating ok, but our concern is the spatial accuracy in the end.

I’m very curious about the notion of just marking the green bands with GCPs. Why would that help to connect the LWIR? And how would that make the other non-green bands to register the same?

Thanks again, and we’d really like to figure out our best process.

Jerry

4

I just checked the most recent lwir-only project I ran, for a montane meadow in the northern Sierra from last August with mostly herbaceous cover, probably no standing water, and a bit of road surface (going to be hot), and 588 out of 611 images calibrated. The 5-band also calibrated well and produced a nice composite image. My inclination is to use ArcGIS Pro to georeference the lwir to the 5band using the road surface as a reference, and maybe only use one link to translate the position if it’s a bit off; shouldn’t need to rescale or rotate I’m assuming but could use a couple or three points to accomplish that if needed. I’m just not convinced that we can use GCPs and don’t see how Pix4D can magically connect it to the other 5 bands. But maybe I’m missing something.

5

Hi, all the bands in a rig are connected with translation values, same for altum. Our database has these values (x, y, z). The green band is used as default by the reference band, you can have a look at the image properties editor. Thus everything will depend on the green band. If you mark GCPs only on the green band, it should theoretically work. You can do a test by marking GCPs on all bands (except LWIR) and then another project where you mark GCP only on the green band. then check the accuracy of both. Let us know how it goes :slight_smile: Using ArcGIS is also a good idea.