Yesterday saw me reduce the object list from dolphot output from one of the image sets to 887 probable stars. Today I set out to visualize these probable stars on the reference image they were extracted from. I wrote a python program using knowledge from back when we wrote our own aperture photometry codes, and I successfully displayed circular apertures at each probable star’s coordinates overlaid on the reference image. Success! Until I realized that the source detection was horribly wrong.

Here is the whole reference image where small purple circles show probable star’s positions.

Here is a zoomed-in image near some sources

It is clear that the “probable stars” are really not the most obvious stars in the image if they are stars at all. I am fairly confident this is an issue with my culling process (and not dolphot). By excluding all but objects with a sharpness in a very small range I think I unintentionally excluded most of the real sources.

Now the question remains: How can we cut down the objects to a reasonable number but not exclude the real sources?

I am interested in roundness as a filtering parameter. Dolphin warns against using it to discriminate, but he acknolwedges it can help to exclude the sources in diffraction spikes, he doesn’t explain what a “good” roundness is though, so I made a histogram of the roundness values thinking there would be higher concentration around the “good” roundness point. Hmmm, I don’t know what this tells me, but it looks like roundness may also want to be close to zero for good source.

Ok, a bit more research and I found that roundness is generally a measure of ellipticity. This would mean that for a circular source, roundness should be zero. I will try only keeping roundnesses below .5 and see how that works.

I went back to the original dolphot output and filtered it down keeping SNR > 5, getting rid of any objects with quality flags, cutting sharpness to between -.3 and .3, and cutting roundnesses down to just those below .5. This yielded 69k objects. Too many, but I humored it and ran plotpos.py on it to see what it would look like.

This is what it looks like.

Zoomed in

Ok, well it is getting pretty late. I am really happy to have the objects on a reference image, but more than a bit flummoxed about what to do next. Clearly my second try at object culling was pretty unsuccessful as there are WAY too many objects identified. I will try to spend some time tomorrow as well. I will at least do some reflecting on here tomorrow.