KBO encounters Options

[john_s]

I can discern 8 parallel, nearly horizontal, rows of artifacts per image tile. Once you start looking for it you can see this on every image. Any idea what causes that? Maybe some side effect from the subtraction algorithm?

[Edit:] Well, I had about 30 images in a row with those parallel rows of artifacts; now I'm not seeing them any more.

If you see them again, maybe make a screen grab and send me a copy by PM...

Thanks,
John

[john_s]

Suggestion: Might not be a bad idea to explain in the tutorials why stars are black in the middle (because they are very distant point-sources of light & therefore wash out the exposure in a smaller area than would be expected for a KBO because the latter are much closer; helps people understand why blobs=good.)

Thanks for the suggestion- it would be nice to have a bit more technical explanation for those who are interested. However the presence of black blobs isn't due to stars being further away or being "smaller" sources (the KBOs themselves are point sources for all practical purposes, and are as small as the star images). Rather, it's because stars don't move, so each positive image of a star is combined with a negative image of the same star when the subtraction is done- imperfections in the subtraction leave some negative (black) pixels as a result. The KBOs have moved on by the time the negative image is taken, so there's no superimposed negative image- all you get is the unadulterated positive image. Hope that makes sense...

John

[nprev]

Swing and a miss...but knew there had to be a reason for the pronounced dichotomy! Thanks, John. At the very least, a 'why not' rationale for the most easily misidentified objects should improve your SNR a bit.

[hendric]

I figured the black star centers was because of a "perfect" subtraction, ie 255-255=0, and the outside stays white because the gaussian applied to the PSF isn't exact.

tfisher,
I also saw those same parallel lines. I think it's caused by a bad CCD in the array. It doesn't quite seem perfectly horizontal though, it seems like it slants down one direction.

I have seen some giant hot pixels, hence my previous "binary" questions. I was pretty excited about seeing two move in the same direction until I saw the same two objects in a couple of images. Is it sad that I can now recognize parts of the CCD based on the hot pixels? I'm not sure if I should mark the hot pixels as blobs, since they meet the criteria, but I think I can notice them now since the KBOs seem fuzzy-edged but the hot pixels are hard-edged.

Sample of the hot pixel images. (animated gif)


Also, seen something a few times that is either a rapidly changing dust/gas cloud, or shmutz on the CCD. Seen something similar in the same spot a few times, but only got one screen cap of it, so my money is schmutz.


Here's my best example of what looks like ringing.


And my favorite bunch of blobs so far.

[john_s]

I figured the black star centers was because of a "perfect" subtraction, ie 255-255=0, and the outside stays white because the gaussian applied to the PSF isn't exact.

Nope, because zero is actually gray- we stretch the differenced images to show both positive and negative residuals, and so we can properly see the noise in the sky (which is scattered around zero, for a good sky subtraction). We certainly don't do byte arithmetic.

I have seen some giant hot pixels, hence my previous "binary" questions. I was pretty excited about seeing two move in the same direction until I saw the same two objects in a couple of images. Is it sad that I can now recognize parts of the CCD based on the hot pixels? I'm not sure if I should mark the hot pixels as blobs, since they meet the criteria, but I think I can notice them now since the KBOs seem fuzzy-edged but the hot pixels are hard-edged.

Well spotted- yes, those are CCD defects, and the hard edges are diagnostic (sometimes the edges also show "ringing" due to the convolution). If you are sure you're seeing a hard-edged defect, don't mark it, but if in doubt, certainly err on the side over-clicking rather than under-clicking.

John

[hendric]

John,
Good to know, I wasn't implying you're doing byte math, just trying to give a simple example. I understand now about using -/+ math and stretching it to make gray the middle.

I think this is an example of the horizontal banding tfisher referred to. It's not perfectly horizontal, but displaces downwards a few pixels across the whole image.


Here's something interesting. Assuming these two asteroids are at the same distance (I'm using 2.5 AU, center of the belt ), they're only ~ 20,000 km apart.

[tfisher]

I think this is an example of the horizontal banding tfisher referred to. It's not perfectly horizontal, but displaces downwards a few pixels across the whole image.

Yes, that matches what I have been seeing.

Thinking a bit more, I bet it starts out completely horizontal as the image comes from the ccd. But to subtract two images, they first have to be reprojected to a common frame of reference, so it isn't quite horizontal anymore.

[tfisher]

Poking around more at the horizontal artifacts... They occur even in the images on the tutorial page. Here's a quick experiment. I took the "Image 1: Original from 2004-Jun-09 at 11:40 UT" and subtracted this image from itself translated vertically by one pixel. Then I remapped colors so nearly equal values are white and all others are black. The result is attached, showing approximately horizontal lines where there are equal pixel values just above one another.

Interestingly, the "Image 2" has almost to-the-pixel identical lines. If this was coming from a ccd readout problem I would have guessed they wouldn't match up so well. So maybe it is a bug from the image reprojection step?

Attached thumbnail(s)

 

[john_s]

Thanks for these examples, Richard and all!

Here's something interesting. Assuming these two asteroids are at the same distance (I'm using 2.5 AU, center of the belt ), they're only ~ 20,000 km apart.

Though of course if one's at 2.50 AU and the other's at 2.51 AU, they are 1,500,000 km apart... And they might be variable stars too- we're still getting a feel for all the stuff that's buried in these data.


John

[hendric]

Oh sure, if they're not the same distance. I figure the odds are better that two streaks in the same direction with very similar lengths are co-orbital vs happenstance alignment.

[stevesliva]

I thought this was a thing of beauty, as far as finding transients in dense starfields go.

[nprev]

Okay, I'll play: I see two, possibly binary.

This particular processing method seems to make them stand out, if I was correct.

[tfisher]

Are Kuiper belt objects distributed like a belt, or more like a shell?

When I find close asteroids (the ones with three closely spaced bright marks), they always seem to move approximately horizontally across the image. But when I find distant bright blobs with a corresponding dark blob, they seem just as likely to be separated in any direction.

Or is it just that there are so many other sources of variable brightness besides KBO's, that most of the time when there are a bright blob seeming paired with a dark one it is just a chance occurrence. Like two out-of-sync variable stars near each other from our viewpoint?

[john_s]

Most KBOs are in a flat disk, like the asteroids (though there are dramatic exceptions), and here we're looking in the plane of the disk so most KBOs we find will be trundling along in the same direction at similar rates. That's good, because it means most are heading in the general direction of the New Horizons trajectory (though we expect only a few percent of the KBOs we find to be accessible to the spacecraft).

And yes, most point-like variable or moving objects in the frames are probably *not* KBOs - they're variable stars or sometimes even artifacts such as CCD defects, as discussed previously. That's why we need all this help in cataloging everything, so we can sort out the few objects that really are of interest to us. So thanks again!

John

[nprev]

John, interim progress report. 238 objects identified in 801 images, which means I see a candidate on the average in 29% of the images...call it one of every three. How does this compare with the expected results? Might have been skewed in the first few dozen or so, but I think I got it down now.

Also, is anyone gonna follow up on asteroid finds? I assume so, since they're scored as well.

Want to say again how much I enjoy the user interface; it's well thought-out, VERY easy once you get used to the detection methodology, and downright addictive!

EDIT: Forgot to add a suggestion: a "back" button! A couple of times I spotted a candidate just after I clicked "done"; couldn't fix it!