MTH 448/563 Data-Oriented Computing

Fall 2019

Day 17, Wednesday, October 23

Reminder: SQL test Monday, Oct 28. Please prepare thoroughly for this.

Quiz: 'partner'. What is the first name of your partner from Monday's class?

Sloan Digital Sky Survey (SDSS)

An extensive moderate (0.5 arcsecond) resolution survey of the sky: an example of big data with a SQL API.


Querying the SDSS

Sloan Digital Sky Survey, SkyServer DR14 (faster than DR15 server): SQL Search

SQL Search Tutorial

My own observations of Halley's comet using binoculars

Quiz "Halley": In which year did I make those observations?

Exercise: your patch of sky

Stake out a small patch of sky for yourself: choose a squarish (ra,dec)-rectangle with the same area as your thumbnail at arm's length. somewhere in the DR14 footprint,

sky_patch_geometry.png thumbnail_angle.png

Better sketch with different variable names:


Best if you avoid the plane of our galaxy (Milky Way) and do not include any bright foreground star.

Your answer will be in the format: a < ra < b, c < dec < d. You could look at a sky map like this one to get your bearings.

Exercise: Please enter the coordinates of your own patch of sky, all in degrees, on this spreadsheet (change the Z to a G).

SDSS objects in your patch

Exercise: Get the coordinates and some properties of all of the objects in your own patch of sky that were captured by the SDSS.

The Schema Browser is useful.

Measures of brightness: Apparent magnitudes is the negative logarithm of the brightness as seen from here: larger number therfore means dimmer. Magnitude 9.5 is about as faint as you can see with binoculars. Measures of flux and magnitude in SDSS: explanation here. Here is a chart of the filters used by SDSS.

Exercise: make a plot of all your objects?

Galaxy count

Exercise: How many galaxies did the SDSS find in your own patch of sky? Enter it in the spreadsheet.

Algorithms for recognizing things in data



Recognizing stars in photographs with no metadata

Group exercise: identify the stars in a photo with no metadata. Assuming you have a list of all the bright stars in the sky with their coordinates, how would you identify the stars in any photo of a portion of the sky, given no info except the photo itself?



Exercise to tax our own visual recognition abilities:

Here is a photo of Halley's comet taken on March 8.

Figure out the correspondences between objects in the photo and objects in a star chart. Then, using the photo and the chart we saw earlier, determine the location in celestial coordinates (Epoch 1950.0) of (the nucleus of) Halley's comet on March 8. Give the ra and dec.


Let's use Inkscape to try to line up the photo and the chart. Be sure to hold Ctrl to maintain aspect ratio while rescaling: we must apply only rigid motions. NOTE for Mac users: You will need to install XQuartz in order for Inkscape to run. Save my_halley_observations_detail.jpg and wikipedia_commons_2_2a_Lspn_comet_halley_march_08.jpg, and import them into Inkscape. Then shift, rescale, and rotate as necessary to bring them into alignment.

Back to the goal of automatically recognizing the stars in a photo.

Can we recognize a single star? I took this photo on the NY State Thruway at twilight one evening.


How about a pair of stars?


Exercise: Let's see if we can do it (recognize triples of stars), starting with synthesized data. We will first process all the stars, then present ourselves with triples that have been subjected to arbitrary translations, rotations and dilations.

This will be the subject of Report 4, due 8am Saturday, Nov 2.

How about a triangle of stars?

How to make a canonical description of a triangle that is invariant under translations, rotations, and dilations? A moduli space.