How Do You Tell Real Asteroids From Imposters?

The telescopes and CCD cameras of the Astronomical Research Institute (ARI) try very hard to see as deeply as possible. The camera used for our images has been tweaked to go about a magnitude deeper, but the cost is that there is much more noise in the images. Combine this with the fact that nearly all new discoveries are going to be 20th magnitude or dimmer, and the limiting magnitude of the ARI systems is about the same... and you can see that noise is a big problem in trying to determine which objects are real asteroids. I will categorize below the different types of noise which might be assumed by the student or by Astrometrica, to be an asteroid candidate. But first...

The NOISE DIRECTION

The images are not all taken with the telescope pointed exactly in the same direction. Close, but not perfect. Yet, the silicon chip which is the detector has millions of pixels with their own little story. There are hot pixels, there are black pixels, there's noise of every kind, and often this noise is very much associated with the pixels and moves across the sky with the pixels. When you blink images, the software makes sure the stars always are perfectly on top of each other for the 3 images, but if the telescope is aimed slightly differently, then the noise will appear to jump around in its own pattern. Often there is a more or less consistent direction to this chip-caused noise - I call it the "noise direction". Nearly all of your false asteroids will be moving noise associated with this noise direction! You MUST first identify this noise direction and ignore any "asteroids" which move with it as just noise, or otherwise you'll waste time and excitement on... noise.

Let's recall what a genuine asteroid track will look like...

(1) It will move with constant velocity across the 3 images. This criterion can be hard to judge if it is very slow moving. A blurry, poorly centeroid'ed image moving slowly will not necessarily pass the constant velocity test, but yet still could be real. What the centroid routine does is to take a weighted average of where the light is in a 4 pixel diameter circle centered on where you click.You might have a faint star which interferes with one of the images and fools the centroid. So ultimately there is only one way to judge the constant velocity - your own trained eye. I've found that the ReportTester.xls software is too blunt an instrument and can throw out good objects which you could in fact retain if you are careful in guiding the software to centroid properly.

(2) It will be roughly the same brightness on the 3 frames. The 3 frames could have very different sky conditions, as clouds commonly come through. So, don't judge this one just by eye, judge by looking at the magnitude given for the 3 images. They shouldn't differ more than 0.3 or so. HOWEVER, the asteroid candidate could merge with a background star on one of the images, and this could make you (or the Excel Report tester) think it fails this criterion when it doesn't.

(3) It will show very little break between the 3 images, so that if is moving significantly from one frame to the next, it will also look like a streak instead of a round star on a single frame. Main Belt asteroids are more common, and move slowly. NEO's (Near Earth Objects) move much faster and will usually show a long streak on each image. TNO's (Trans Neptunian Objects) are very slow moving and may not show any movement at all on our photos. They are usually dimmer than the limiting magnitude of the 24" scope however.

(4) It will also almost certainly be moving in a direction other than the direction that background sky noise moves. So why would background sky noise move? Because the telescope doesn't track the stars perfectly, and so the chip is aimed slightly differently for each of the images. Yet, when the images are blinked, Astrometrica will stack each frame precisely on top of the other pictures such that the STARS all lie on top of each other! This means that the noise pattern will shift between images. Most of the noise is correlated with the chip itself. This background noise can have apparently star-like objects to it, and when you blink the images you'll see all this stuff moving in a constant direction. If you have a candidate asteroid which moves in the same speed and direction as all the rest of the noise on your images, then it is almost certainly noise too, even if it looks otherwise genuine. So, the first thing you should do when blinking is get very aware of how the background noise shifts and then ignore things which are doing exactly that same movement.

(5) It will show a point-spread function (PSF) which is hump-shaped (brightest pixel at the center, surrounded by dimmer pixels). This PSF is due atmospheric and optics-induced blurring, the wave nature of light, and small tracking errors. Stars and asteroids will never be single bright pixels ("hot pixels", perhaps due to cosmic ray hits, thermal noise, or just a bad pixel).

Unfortunately, I think it's possible you will find real objects but yet they will fail one or more of the tests. If the asteroid is moving too slow then the time baseline may not be enough to give a decent path length and so the linearity and speed tests might fail. It's also possible that a star may be very close to one of the images and fool the centroid'ing which happens when you click on it. When you click on an object, it looks out at neighboring pixels and measures the mean position of the light within that circle. If your object is close to another object, you can find the software centering on a close bright star or bit of noise. You may be able to prevent this by adjusting the size of the aperature circle. In astrometrica, click on the blue wrench on the toolbar, (same as file | settings), then in the pop-up, click on the program tab and in the object detection section, adjust the aperature radius down from 4 to 3 or maybe even 2 if the confounding neighbor is really close.

So here are Examples - Real and Imagined Asteroids...

1. A Real Asteroid - bright, and moving fast = big long streak!

2. Two Real Asteroids - not as fast moving. Potential new discovery is CAB0013. Previously known asteroid is red-boxed. When you click the "known object overlay" buttoon in Astrometrica, it will show these. The box might be off by about a box width but not likely to be more than that, from the true object. Note also that sometimes you'll look and see nothing inside these red boxes. That means they're too faint for this scope on this night.

3. A streak, but it doesn't move... probably an edge-on spiral galaxy

4. A streak which looks very much like a real asteroid.... but its only on one of the 3 images. It's noise!

5. A cosmic ray event or hot pixel. Note the saturated bright pixels which have no pixels of intermediate brightness surrounding them. Instead, the brightness falls immediately down to "sky" levels. If its a slow moving asteroid and therefore not a streak, it'll show a point-spread function looking like the stars.