[Aavso-photometry] Automated star finding

Wed Aug 8 13:39:20 EDT 2007

Jim,

That is a good list to start from. One important omission form your list is 
Mira. Mira Pro with its MExtract Module will adderss the stated problem in a 
number of ways but it is more versatile even that this application (Mira AL 
offers a targeted subset of the MExtract capability). The MExtract module 
has been available for about 1-1/2 years, during which time it has been 
refined and expanded. At the algorithm level, automated object detection is 
far more complex than one might suspect. I hope no one minds me getting a 
little detailed here but I think that, by giving a lot of detail, people can 
appreciate some of the important issues behind automated object detection 
and property extraction. And I do appreciate up front that this will 
probably sound like a promotional ad for the MExtract Module.

To see how MExtract integrates into photometric measurements, let's start 
with Mira's aperture photometry package. The photometry package can work 
with any number of unknowns, any number of standards, and any number of 
images (that is, an image set of 1 to n images). Photometry can be done in 
the following ways:

    1) manually by marking objects (in at least one image, if measuring an 
image set),
    2) by loading a catalog file containing (RA,Dec) or (X,Y) coordinates, 
and, optionally containing some standard star ID's, mags, and/or weights, 
and
    3) internally, as part of the MExtract object extraction pipeline.

For any of these methods, the photometry routines produce a table of 23 
columns for each object, about half resulting from measurements and 
calculations Mira has made (other info, such as name filter, date, time, 
etc. are taken from the image header). I will focus on option 3, using 
MExtract. This tool involves processing a pipeline of 5 stages: Procedure, 
Background, Detect, Match or Difference, and Post-Process. All stages are 
"optional" so that a procedure can be configured to the application at hand. 
To use MExtract, you click on the tabs and controls to setup an MExtract 
"profile" (a collection of parameters) describing which stages to use in the 
procedure, which methods to perform in each stage, and whch parameters to 
use for each method. You can run that profile or save it with a profile 
name. Recalling a saved profile allows you to run the same pipeline on other 
images or image sets. For example, you might create a profile that produces 
only an object list for use in option 2 above (processing a catalog file 
through the photometry package). Or you might load the catalog as in option 
2, and manually add or delete objects (as in option 1) before processing the 
objects.

As I mentioned, Photometry is a method in the Post-Process stage of the 
MExtract procedure. To use it, you check it on the Post-process tab and 
configure its parameters (aperture  sizes, orientation, ellipticity, etc.). 
When using MExtract for photometry, the object coordinates and properties 
result from the Detect, Filter, and Match/Diff stages of the pipeline, as 
these feed the target list into the photometric measurements.

MExtract was designed to be extremely versatile so you can do things like 
find stars and tabulate photometry. But there are many possible variations 
on just that theme. For example, you might configure it to find all the 
objects meeting "criteria" (more below) and then measure them in one image. 
Or you can have it do that for an image set (also called a "stack"). You 
might have it toss out those objects that match, or do not match. at a 
"similar" coordinate in every image ("similar" means the detection 
coordinate is within some radius of a position in the previous image). This 
can feed a list of transient objects (objects that are moving or that 
suddenly appeared) into the photometry routines. You might also just have 
MExtract create a table of the star coordinates for future measurements on 
other images or image sets using photometry option 2. You can load that 
object list into the photometry package and add or delete objects from it, 
save it as your master catalog for monitoriing a field (see end of paragraph 
above).

I mentioned detecting transients: One application of MExtract is to search 
for variables and novae. One of the options for the post-processing stage is 
to compare the photometry of each object detected in an image set with a 
threshold value for the variation. For this, it uses the magnitude measured 
in each image. For example, you might have it find objects in an image set 
that vary by more than 0.2 mag.

The issue of detection "criteria" is an important one. The MExtract pipeline 
has a number of stages---they are Background, Detect, Filter, Match, Diff, 
and Post-Process. You select a Procedure consisting of which stages to use, 
and then specify parameters to use for each stage. All options are saved 
together as a "profile" (see above). In the Filter stage, you select which, 
if any, criteria you want to use to filter the object list coming out of the 
Detect stage. Here are some examples showing why Filtering is both important 
and useful:

    1) You might choose to keep objects that are within some range of 
roundness (i.e., stars vs galaxies, or to discard highly elongated cosmic 
ray hits),
    2) Keep objects within some range of orientation angle (e.g., for 
objects or PSF's that are elongated vertically to study your tracking or 
optics, or maybe you want your object list not to contain saturation bleeds 
and hot pixels),
    3) Keep objects within some range of size (filter out the large blobs 
and/or hot pixels), and/or
    4) Select only objects which peak within some brightness range (avoid 
objects that are too faint or which should be considered "saturated").
    5) In addition, if you just want a list of widely spaced objects to 
monitor, it can even reject objects that are too close to other objects.

You can choose any or all of these filtering criteria, and they become part 
of the profile you save. Setting up different profiles allows MExtract to do 
all sorts of projects.

Using Mira's the built-in analysis tools after an MExtract run you are ready 
to go research-wise with the photometry table produced by MExtract. For 
example, you might use the scatter plot tool to compare values from columns 
in the photometry table (or any other measurements table). For example, a 
very obvious thing to do is plot a light curve, with magnitude error bars, 
for object 5 which was reported as exceeding your specified magnitude 
variation threshold of 0.25 mag.

As a final note, I would add that our lowest price Mira AL includes a 
simplified version of the MExtract module that is tuned to detect and 
measure stars for photometry. Schools use this for someting as obvious as 
having students make an HR diagram for objects in a field or in a cluster. 
There are only a few adjustable parameters: 1) Detection sigmas above sky 
noise, 2) Maximum number of objects to find, and 3) whether to measure the 
entire image or just a rectangular region. It runs from one of the toolbar 
buttons on the Photometry toolbar. From my detailed description of the 
MExtract module for Mira Pro, you can imagine that, internally, Mira AL runs 
a pre-wired MExtract profile using parameters chosen for detecting "nice" 
stars on a single image.

I apologize for all the text here---I didn't intend it to be this long! But 
I hope you can get an appreciation for all the issues involved in doing this 
kind of work. I think our MExtract module has broken down the procedure into 
as painless a task as possible but one has to understand the details of how 
it works in order to use it optimally.

Michael Newberry

----- Original Message ----- 
From: "Jim Roe" <jroe at jamesroe.com>
To: "Aavso-Photometry" <aavso-photometry at mira.aavso.org>
Sent: Wednesday, August 08, 2007 7:11 AM
Subject: Re: [Aavso-photometry] Aavso-photometry Digest, Vol 45, Issue 6

>
>
> bailyhill at aol.com wrote:
>> Could you mention some of the windows programs that you have used
>> successfully?? How about Mira, Maxim, Tfocus, Allos, Temma2, Canopus,
>> AP4WIN?? I think these are some of the ones that AAVSOers would be
>> using.?
>>
>
> While you (or someone else?) is at it, would you tell me which, if any,
> program is capable of listing every detectable star in an image with its
> RA, Dec, instrumental magnitude, (plus all the other goodies in the
> header such as date, time, filter, exposure, etc).
>
> The closest I've come is AIP4Win that will list most of these but only
> against PixelX and PixelY coordinates - unless you use their primitive,
> human input required, plate solution.  (CF, Pinpoint or Charon which are
> automatic and nearly instantaneous.)  I suppose I could use the plate
> solutions by either Charon or Pinpoint LE (in Maxim) to transform the
> Pixel coordinates, but that seems clumsy.
>
> Charon does a nice job but requires one to pick off the stars in the
> image individually.  The good news is, I've asked Bill Gray to see if he
> would fix this and he said he would look at it - keep your fingers 
> crossed.
>
> It looks to me that Maxim is only good for time series (in the
> photometry front, that is - I like it a lot for other functions).  Am I
> missing something?
>
> BTW, I've been able to run WinXP in a virtual machine from a Linux host
> using VMware Player.  V.2 supports USB 2.0 and the only big problem I've
> encountered is using the host serial port - but that is easily fixed by
> using USB-Serial adapters ($10 each).
>
> Jim Roe
> Wentzville, Missouri
>
> _______________________________________________
>
> Aavso-photometry mailing list
> Aavso-photometry at mira.aavso.org
> http://www.aavso.org/mailman/listinfo/aavso-photometry
>
> 