Martian Anomalies

I finally spent some time locating, downloading, and correctly reading images from the HiRise instrument on the MRO (Mars Reconnaissance Orbiter) from the Planetary Data System.  Usually always a pain, and this was no exception.  I think there are things out there to 'help' me get the data I want, but I'm old school and it's usually much easier for me to write some simple C code to extract the data from the calibrated datasets and then do what I want with it.

I can process these images the same as I do the LRO (Lunar Reconnaissance Orbiter), although they're slightly different as far as dimensions go.  For one, they're all 546 pixels wide.  512 pixels are legit, with 18 at the beginning and 16 at the end used for other purposes.  The images can be pretty long.  The first one I downloaded was 25000 pixels long.  You can pack about 2700 64x64 subarrays into that size.  I've seen other images that are 35000 pixels long.  There's some variation in size, so I'm going to have to deal with that if I want to automate the data production.

Looking at MRO dataset ESP_065269_1695_BG12_0.  Data collected on 29 June 2020 at about 06:25 UTC.  I'm not sure what the pixel scale is.  I'd imagine pretty big compared to LRO resolution.

Wow check out those Martian cliffs!

Almost looks like a series of roads:

these are some very small portions of a 546 by 25000 pixel image.  Not quite I'm still playing with the software a little, but I'm getting what I'd call acceptable results at this point.  Here's the top 256 64x64 image sections.  The autoencoder seems to be working as advertised: The higher the ranking more different the images are from each other (i.e. 'Not Mars'); towards the end of the 256 they look almost all the same.  Looks like many of the 'not mars' images are shadows and sharp edges.  The dunes in the bottom ones are stunning.

Anyhow, now I have software and scripts that can deal with this data as well.  So I'll start to work with it a little and see what happens.  My primary focus is on lunar data, but I'd planned on adding Mars at some point.  Now seems to be the time.

Rolling Lunar Rocks

Some cool looking paths apparently carved by rolling rocks.

This is from LRO dataset M1415751069RC:

Lunar Anomalies #18

Weird crater bottom.  LRO dataset M1424544254RC:

same, zoomed in a little more:

Here's a nice looking outcropping of some sort.  LRO dataset M1415751069LC:

Lunar Anomalies #17

LRO dataset M1369989563LC.  A pretty strange crater bottom shape:

Craters nearby have similar:

From LRO dataset M1351197461LC, large boulders and boulders rolling downhill!

some closeup details:

LRO dataset M1311260861LC.  An Octagon???

Another hexagon with some interesting looking dirt LRO dataset M1278345738LC:

Looking For Lunar Anomalies, continued more

I haven't posted a 'Top 256' image in a while.  These are the 256 highest scoring anomalies for a particular section of an LRO dataset.  An LRO dataset is 5064 pixels wide and 52224 pixels long.  That a pretty big bite, so I divide this dataset up into eight sections, each of which are 5064 by 6528 pixels.  At 1m per pixel resolution, that's an area of about 33 square kilometers.

LRO dataset M1379380302LC section 'E' looks like this:

The Top 256 are 64 by 64 pixel image sections.  Here's a mosaic of the Top 256 (left to right, top to bottom):

Here are the Top 256 showing their location marked with red boxes (kinda hard to see, I know):

Here's a zoom-in where there's a concentration of red squares:

Lunar Anomalies #16

From LRO M1524704530RC.  A complicated crater scored #1 with the autoencoder for the section that it's in.

Same dataset, here's a weird looking groove and crater (scored #4 for it's section).  This is right on the edge of the image, but since I see no other distortions I have to assume this is the actual shape.

LRO Dataset M1517686238LC.  I don't get this one at all:

Nice little hexagon in LRO dataset M1426310663LC:

Looking for Lunar Anomalies, continued again

This is an agreement map of 20 different models.  Any shade of orange indicates that all 20 models agreed that this was one of the top 256 anomalies in this region.  Grayscales indicate number of models that agreed on that particular location, brighter meaning more models in agreement.

And here's the average agreement map.  Bright areas indicate strong model agreement AND a high anomaly score:

LRO dataset M1493691723RC