Model of the Citadelle site and Rochette workspace made from high-resoution images taken on Sol 178-190 with the Mastcam-Zs, Navcams, and Hazcams on the Perseverance Rover. This model is 23m meters wide (east-west) and 24m long (north-south). The scalebar is 10 meter long, 1m wide, and points North. The colors are enhanced and gamma corrected to show contrast. Credits: NASA/JPL/MSSS/ASU/Cornell
The Rochette colection
5 comments
Great Work! Added this 3D model to my GlobeViewer Mars App. Thank you!
rab-software.de/en/actual-projects/globe...
@cdt59 Sorry for the delay in my reply and thank you for taking the time to explain. Metashape seems to be the direction to go for large scale meshes like yours. I will reach out to Jim Bell about the spectral filters and see if they have specifics, but I read a paper some time back on their wavelength and can come up with my own table. Right now I use my own experience to work off of my own FFT frequency (low & high) analysis based image conversions and heightmap generation.
Also, do you stitch "all" the images together to create your meshes, or do you sections and then stitch the meshes together? I have found that there is too much lens distortion (which is why you mentioned the rad-calibrated images) to bring in to many images at once. I have also do a lot stitching by hand now because my results are better in Reality Capture, but it is a time suck. Finally I have been pushing the Camera team to start using Watson like a remote photogrammetry arm, especially around the geological interests. WatsonCam has good detail and the arm is articulated and precise enough to create excellent overlapped mosaics. Maybe you already saw this application and suggested it as well. At least it would be great to get a few images of objects not seen from the Mastcams, which gets me back to your mesh. Wow, the fact you have such a large swath, is really impressive.
PS if you have access to lens parameters, could you share them? Also, if you also have access to LUTs created by NASA JPL for the different cameras, that would be great as well Currently I create my own based on the color targets for the MastCam-Z take each SOL, but I notice the NavCam has only recently been taking an image or two of these targets.
I misspoke when I said the full-frame Navcam and other engineering camera images are on the PDS. They are still tiled, unfortunately, and that's a bummer for the 3D modeling community. I'll ask what is up with that.
@mebalzer thanks a lot. I make these unofficial models to help the Mars 2020 scientists interpret the various rock geomorphologies. I'm glad you can enjoy these models too.
The raw images are immediately available here mars.nasa.gov/mars2020/multimedia/raw-im... Unfortunately, the Navcam raw images are tiled, and as you know, it's important to have those reconstructed into their original full-frame images before alignment. After a period of time (three months, I believe) the full-frame, rad-calibrated, good-stuff is released to the PDS: pds-geosciences.wustl.edu/missions/mars2...
I did not know about PBR textures, but that sounds very relevant for the multispectral Mastcam-Z images. Thanks for the tip. Many rocks that the Mastcam-Zs image have specular glints, so it would be fantastic to characterize the material's metalness independent from its color and roughness. If you know of a blog or video that explains more about a workflow for that, I'd know to investigate.
I'm using Agisoft Metashape Pro for these models. Thankfully, the academic license is affordable. I hear really good things about Reality Capture, though. I think both programs are equally powerful, and I just prefer Metashape because I am used to its workflow. :)
This one is really impressive.
I was hoping to have as many images to create a similar version using the Navcam from 184 but one of the images came back corrupted. Sadly it was the one with the rock that the core sample was taken. If you have access to a better image than the public has, that would be great if you could create a newer model. I also appreciate that you make your models available. Thank you.
I do have a couple of questions though. 1st. Have you tried to create PBR textures based on the other filters captured from MastCam-Z. I have been working out the wavelengths but sadly Reality Capture does not seem to UV the same way, when converting using the RGB image files versus the filtered monochromatic image files. I have had to unwrap manually and do image projection to come up with a series of different texture sets to evaluate roughness, metal, specular and delight for an alebedo texture. 2nd, what photogrammertry application are you using?