Perseverance Imagery, technical discussion of processing, cameras, etc. Options

[scalbers]

It's not great - but this is my agisoft metashape results with the Navcam images so far
https://sketchfab.com/3d-models/m2020-landi...7215aa7db2fb0c8

Looks pretty nice given the available vantage points (one rover location). Is there a possibility the "1st Person" navigation mode would work with this model?

[MarT]

Stereo imagery viewed in a VR headset is a bit underwhelming - reconstructing geometry using photogrammetry to create a fully 3D representation of a landscape is much more interesting.

I did this with some Curiosity imagery a few years ago, with fascinating results. If you have a SteamVR capable VR headset, you can have a look here: https://steamcommunity.com/sharedfiles/file...s/?id=928142301 - I typed up some fairly detailed notes in the description which will broadly apply to Perseverance imagery.

(Full disclosure: I work for Valve, creators of SteamVR. The Mars stuff was a fun personal project which turned into something a bit larger...)

Thats awsome! Great to see you here. I just ran into that two days ago. It is something I always wanted to do, but never fully found the time to work for a longer period of time. I have tried Photogrammetry on InSight, this is the result two years ago: https://www.youtube.com/watch?v=cBYAwTm_ArE...eature=youtu.be

Big thanks for the info here an on steam! It will be surely very helpfull to others too.

To stay on topic: I had a run with a short sequence of the true raw data from the down-look cams: https://www.youtube.com/watch?v=l4WKIoTjE4c...p;pbjreload=101
I will just wait for all the EDL data to download and will have a run with the data in Agisoft. I tried with the MARDI cam. While it was capable of clearly registering the images, I must have done something wrong, as the last image ended up "under the ground". Anyway the Perseverance data surely look promising in that matter.

I am working on a stabilized 360 video as with the MARDI cam. This is a quick version of the underlaying layer simulating approximately the view from 10km above the surface. I made it from the debayered data. Propably impacted by the FFMPEG compression.

Attached thumbnail(s)

 

[phase4]

About image pointing, the algorithm I use for MSL, based on this post, seems to work fine for Percy, with the "rover_attitude" quaternion field replaced with "attitude".

Thank you for this hint, fredk. I didn't realize the Spice toolkit works just as fine without kernels.
For future use, does anyone know how to obtain the zoom value from the JSON information? Should it be derived from the CAHVOR data?

[mcaplinger]

For future use, does anyone know how to obtain the zoom value from the JSON information? Should it be derived from the CAHVOR data?

Should be if the CAHVOR model is set correctly, which they may or may not be at this point. See https://github.com/bvnayak/CAHVOR_camera_model and https://agupubs.onlinelibrary.wiley.com/doi...29/2003JE002199

[John Whitehead]

Do color calibration targets ever fade (change color) from solar (and worse) radiation on Mars?
How are they tested on Earth and proven to not fade?

[mcaplinger]

Do color calibration targets ever fade (change color) from solar (and worse) radiation on Mars?
How are they tested on Earth and proven to not fade?

https://mastcamz.asu.edu/mars-in-full-color/

At the University of Winnipeg, the effect of intense Mars-like ultraviolet (UV) light on the colors of the eight materials was studied, confirming that the materials will only change very little with UV-exposure through a long mission at the Martian surface.

Typically, they get dusty before fading would be an issue. Hopefully these magnets work better than the last time.

[JohnVV]

just came across a paper on the cameras and mic

"The Mars 2020 Engineering Cameras and Microphone on the Perseverance Rover: A Next-Generation Imaging System for Mars Exploration"

https://link.springer.com/article/10.1007/s11214-020-00765-9

[phase4]

Should be if the CAHVOR model is set correctly, which they may or may not be at this point. See https://github.com/bvnayak/CAHVOR_camera_model and https://agupubs.onlinelibrary.wiley.com/doi...29/2003JE002199

Thanks for the links. I tried the code but it gave very low values and fluctuating results for the Mastcam-Z focal length.
Will try again when new images arrive.

[mcaplinger]

Thanks for the links. I tried the code but it gave very low values and fluctuating results for the Mastcam-Z focal length.

Seems like, for the one case I looked at, that the values in the JSON are in the order VORCAH instead of CAHVOR as one would expect. I don't know if this is intentional or a bug. And I'm not sure if these models are actually correct anyway. But you could take a look.

[djellison]

Jim has put out a Mastcam-Z filename decode guide

https://mastcamz.asu.edu/decoding-the-raw-p...mage-filenames/

It includes digits in there that describe the focal length in mm as three digits.

All the stereo pan images report 34mm for that - which matches the widest angle of Mastcam-Z MSL Mastcam Left.

[mcaplinger]

It includes digits in there that describe the focal length in mm as three digits.

Oh yeah, duh. Well, it was much more interesting to extract it from the camera model

FYI, widest angle of MCZ is 26mm.

[djellison]

Ahh ok - they miss spoke at yesterdays thing then. This matches MSL Mastcam Left then

[mcaplinger]

This matches MSL Mastcam Left then

Correct. Original MCZ spec was for 34mm to 100mm but we ended up with some extra credit range and can do 26mm to 110mm. Typically values 26, 34, 48, 63, 79, 100, and 110 will be used, but AFAIK nobody is sure yet how they'll be chosen. Zoom has to be stowed at 26mm for driving, so that might motivate use of that setting sometimes.

[phase4]

Seems like, for the one case I looked at, that the values in the JSON are in the order VORCAH instead of CAHVOR as one would expect.

So close. Processing the CAHVOR as VORCAH gave more plausible results, although one decimal off.
It was a fun exercise but no longer nescessary... thanks for posting the decode guide Doug!

[mcaplinger]

So close. Processing the CAHVOR as VORCAH gave more plausible results, although one decimal off.

Pixel size is 7.4e-6 m = 7.4e-3 mm, is that what you were using?