Micasense Altum Thermal: conversion, resolution questions

[aggieair]

We are flying the Altum camera at 400ft AGL.

An Agisoft project was created using all 6 bands as in the instructions link:

https://agisoft.freshdesk.com/support/solutions/articles/31000148381-micasense-altum-processing-workflow-including-reflectance-calibration-in-agisoft-metashape-professi

I am experimenting with different ways to convert the thermal band into Cel, ie, Agisoft's Raster Calculator, ArcGIS's Raster Calculator.

In the 6-band Agisoft project (after the 5-band relfectance was exported),  removed the lines with B1/32768; B2/32768; B3/32768; B4/32768; B5/32768 lines, and replaced it with a single line: (B6/100) - 275.13. I've exported the ortho (Index Value selected). Then also another thermal raster where it only said B6 in one line, and did the calculation to Cel in ArcGIS Raster Calculator. Just to compare. Values were same. Good.

Both resulting thermal images from the original 6-band Agisoft project have super fine resolution of 5cm (same for the BRGNIRVRE bands). I am not sure what it's doing to the thermal, upscaling? Can anyone explain?

Then I created a new Agisoft project, but only using the Altum thermal images. This final raster had a 69cm pixel resolution, closer to the 80cm that Micasense reports for a 400ft flight. I also tried to find the difference between my 69cm thermal raster and 5cm thermal raster. While not apples to apples, or a perfect comparison, there were some bigger numbers than we were expecting.

Can anyone tell me how Agisoft is processing the thermal band when it's part of the 6-band multicamera project and why it makes the resolution much finer?

Also those with the Altum, which software do you prefer to convert your centiKelvins to Cel?

[Alexey Pasumansky]

Hello aggieair,

When you are generating a multi-band orthomosaic, it would have the same resolution for all the layers, therefore the thermal band orthorectified images will have large scale compared to thermal band only processing. Visually the individually orthorectified images for both approaches should be similar, but due to the larger resolution for the multi-band approach the blending of those orthorectified images would likely give more detailed result.

[aggieair]

Can I actually report to my client that the TIR has a 5cm pixel resolution?

It is documented that the TIR resolution/GSD is going to be much more coarse (80cm on the Altum specs sheet). Hence the confusion on what to report.

[Paulo]

AggieAir,

here is the GSD report for last 2 bands NIR and Thermal (LWIR) of an Altum multi sensor camera flown at 115 m AGL...

The average GSD of NIR Band as well as Blue, Green, Red and RedEdge is around 5.2 cm while Thermal avg GSD is 72.5 cm so upscale is quite significant:

2020-05-29 14:31:58 -------------------------------------------------------
2020-05-29 14:31:58                  Reporte GSD
2020-05-29 14:31:58 -------------------------------------------------------
2020-05-29 14:31:58 Camara: Altum, NIR (8mm)
2020-05-29 14:31:58 Dimensiones Sensor (mm): 7.1208 x 5.3268, 8.8927
2020-05-29 14:31:58 Distancia Focal (mm): 8.0
2020-05-29 14:31:58 Pixel Size (mm): 0.00345000000526429 Pixel size
2020-05-29 14:31:58 Distancia Focal - ajustada (pix): 2244.75
2020-05-29 14:31:58 Altura Camara Promedia (m): 476.18
2020-05-29 14:31:58 Altura Terreno Promedia (m): 361.16
2020-05-29 14:31:58 Altura Vuelo Promedia (m): 115.01 AGL
2020-05-29 14:31:58 Dimensiones Imagen (pix): 2064 x 1544
2020-05-29 14:31:58 GSD - promedio (cm/pix): 5.041
2020-05-29 14:31:58 GSD - ajustado (cm/pix): 5.207
2020-05-29 14:31:58 FootPrint (m): 107.48 x 80.40
2020-05-29 14:31:58 Horizontal FOV es 49.38 grados(interno) y 50.29 grados(externo)
2020-05-29 14:31:58 Diagonal FOV es 59.72 grados(interno) y 61.03 grados(externo)
2020-05-29 14:31:58 -------------------------------------------------------
2020-05-29 14:32:00 -------------------------------------------------------
2020-05-29 14:32:00                  Reporte GSD
2020-05-29 14:32:00 -------------------------------------------------------
2020-05-29 14:32:00 Camara: Altum, LWIR (1.77mm)
2020-05-29 14:32:00 Dimensiones Sensor (mm): 1.9200 x 1.4400, 2.4000
2020-05-29 14:32:00 Distancia Focal (mm): 1.77
2020-05-29 14:32:00 Pixel Size (mm): 0.0119999996337891
2020-05-29 14:32:00 Distancia Focal - ajustada (pix): 161.48
2020-05-29 14:32:00 Altura Camara Promedia (m): 476.32
2020-05-29 14:32:00 Altura Terreno Promedia (m): 361.12
2020-05-29 14:32:00 Altura Vuelo Promedia (m): 115.21 AGL
2020-05-29 14:32:00 Dimensiones Imagen (pix): 160 x 120
2020-05-29 14:32:00 GSD - promedio (cm/pix): 79.369
2020-05-29 14:32:00 GSD - ajustado (cm/pix): 72.498
2020-05-29 14:32:00 FootPrint (m): 116.00 x 87.00
2020-05-29 14:32:00 Horizontal FOV es 52.71 grados(interno) y 56.13 grados(externo)
2020-05-29 14:32:00 Diagonal FOV es 63.53 grados(interno) y 69.82 grados(externo)
2020-05-29 14:32:00 -------------------------------------------------------
2020-05-29 14:32:00 Script finished in 7 seconds.

Note: these GSD values correspond to adjusted focal lengths while respective GSDs for published focal lengths (8 mm for MS bands and 1.77 mm for Thermal) are 5 cm and 79.4 cm which are consistent with Altum specs...

[aggieair]

Thank you Paulo! I will say to the client that the TIR has been upscaled to match the other bands.

Micasense shared a bunch of GitHub links today that have the explanation somewhere in there. I didn't know these existed until I made a help ticket with them.

https://github.com/micasense/imageprocessing

https://github.com/micasense/imageprocessing/blob/master/Alignment.ipynb