Surface geometry "resolution" is sometimes wrong

[rfrenchseti]

From pds-webserver created by rfrenchseti: SETI/pds-webserver#13

From pds-opus created by rfrenchseti : SETI/pds-opus#455

On 8/30/2018 5:04 PM, Rob French wrote:

I've been playing around with surface geometry searches, and I'm finding things that I don't understand.

For example, take COISS_2076-1727054847_1727147261-N1727119421_1, which is a distant shot of Mimas approximately 80x80 pixels in a 1024x1024 image:

Intended Target Name: Mimas Body Center Resolution: 5.12952 Body Center Distance: 859864.88 Finest Observed Resolution (Min): 5.1293 Finest Observed Resolution (Max): 5.12937 Coarsest Observed Resolution (Min): 29.95767 Coarsest Observed Resolution (Max): 40.56602

This resolution is believable - at 400 km diameter, Mimas should be about 80 pixels.

Now take COISS_2060-1644743986_1644781734-N1644777828_1, which is an extreme closeup of Mimas where Herschel Crater fills more than the 1024x1024 image size.

Intended Target Name: Mimas Body Center Resolution: 1.55525 Body Center Distance: 16294.177 Finest Observed Resolution (Min): 1.53693 Finest Observed Resolution (Max): 1.5439 Coarsest Observed Resolution (Min): 1.56995 Coarsest Observed Resolution (Max): 2.45734

How can image #1 be 859864/16294 = 53X farther from Cassini, and yet the body center resolution is only 5.129/1.555 = 3.3X different? In fact I don't believe the body center resolution at all. I think in image #2 Mimas is at least 4000 pixels wide, so the resolution should be at least 4000/80 = 50X different (which agrees with the distance ratio). Yet this resolution would have me believe Mimas would fit in less than 400 pixels, much smaller than the FOV.

Now then, here are the metadata links for the two examples I sent:

https://pds-rings.seti.org/viewmaster/metadata/COISS_2xxx/COISS_2076/COISS_2076_moon_summary.tab/N1727119421_1

CENTER_RESOLUTION = 5.12952 CENTER_DISTANCE = 859864.88

https://pds-rings.seti.org/viewmaster/metadata/COISS_2xxx/COISS_2060/COISS_2060_moon_summary.tab/N1644777828_1

CENTER_RESOLUTION = 1.55525 CENTER_DISTANCE = 16294.177

The data here agree with what I saw in OPUS, showing that the import pipeline is correct.

I went back in time using the last CDAPS navigation I did on 1/12/18, where I keep all the metadata used to generate the navigation. The metadata for N1727119421 roughly agrees with what we're seeing now, to within a few percent. But the metadata for N1644777828_1 is completely different! In fact the old version says the resolution is 0.0976, not 1.555. This is roughly 1/53 the resolution of the first image, which is what I would expect based on the distance.

Conclusion: The current surface geometry files are massively wrong in some cases, and whatever is going wrong happened since Jan 12.

[rfrenchseti]

From Matt:

We can test this keyword against the PIXEL_SCALE keyword, which was calculated by the Imaging Team and delivered to the Imaging Node.

I first tried to click over to the image’s delivered label from the metadata page to which Rob linked. This does not have a direct link that I could find (perhaps it should), but I clicked down the tree a few levels and then up again to “data” and found it: https://pds-rings.seti.org/holdings/calibrated/COISS_2xxx/COISS_2060/data/1644743986_1644781734/N1644777828_1_CALIB.LBL Unfortunately, the label appears to include only the keywords that were created at the project level, and not those created by the Imaging Team.

I then decided to find the image in the Imaging Node’s atlas. Even though I already knew by name the image I wanted, this took me at least 20min and is strongly tempting me to sympathize with the critics of PDS. Part of the problem is that Imaging Node does not appear to offer the Imaging Team keywords on their own, but only shows them as part of the search tool. So I had to construct a search that would include the image we want, which I finally succeeded in doing: https://pds-imaging.jpl.nasa.gov/search/?fq=ATLAS_INSTRUMENT_NAME%3A%22iss+nac%22&fq=START_TIME%3A%5B2010-02-13T00%3A00%3A00Z+TO+2010-02-13T23%3A59%3A59Z%5D&q=*%3A*&start=72 Apparently, added columns do not come through in this URL, so from this link you must follow these steps: 1) click on “Sort View” 2) click on the pull-down menu for “Add field to sort by” and select PIXEL_SCALE. 3) click on the down-arrow under “Start Time” so that it will go back to sorting by descending start time. We are still on page 4, since that is included in the URL (and we have to go by pages because the time constraints do not allow resolution finer than one day!). 4) Finally, text search this page for “1644777828”, and you’ll find the image we’re discussing. You can now see that the PIXEL_SCALE keyword is calculated at 0.0964 km/px.

Conclusion: Rob is correct to say that he has found an error in our calculated metadata.

Best, -Matt

[rfrenchseti]

This has been fixed for COISS but still needs to be done for other instruments.

[jnspitale]

See Sec. 3.1.3.3 in https://docs.google.com/document/d/1JVEhksHlru2jk8ZW6WPyZeGxReG0ccB-3GEQ2vyX4vg/edit