CYBERSKA

 A Cyberinfrastructure platform to meet the needs of data intensive radio astronomy on route to the SKA

  • Public

Output of SAD source finding challenge

Last updated 1818 days ago by Russ Taylor

AIPS SAD was run on the three challenge data images.  The run was carred out by Larry.  Due to the limit on the number of sources that can be fit, Larry split the the challenge 1 and 2 images into east and west portions and ran SAD independently on each.  These SAD output files were run throught three pythons scripts

  1. sadfile.py - to parse the SAD output file into a clean source list with the quantities required for submission to the challenge.   The script also creates a ds9 region file with the locations of all detected sources.
  2. filter.py  - filters the sources list against two criteria. First by signal to noise ratio, S/N.  The S/N is taken the be the ratio of the fitted peak value to the local rms.  The second filter is to check the map value at the pixel coordinates of the source.  If the map value is less than 30% of the fitted peak then the source is taken to be a spurious detection and removed.  Creates a ds9 region file with sources that pass the filters. 
  3. mergefiles.py - merges the east and west version of challenges 1 and 2 and the high and low files from challenge 3 while removing any duplicates. 
One of the issues encountered was deteremining the rms to use for the S/N calculation in 2.  We use the sensitivity files provided with the challenge data set.  However, we found that the values in these files were incorrect.  Larry created images of the rms by measuring the variance in the SAD residual files as a function of position.  The rms images and the sensivity images were compared.  Plots of the values in the rms image versus values in the corresponding sensitivity images are shown below for challenge 1 east, challenge 2 east and challenge 3 high resolution.  All show a linear relationship but the slope differs signficantly from 1.   Note also that it is not clear that the intercept is (0,0),  but I am ignoring that issue for now.

image 

FIgure 1.  Plot of values in the rms image versus the sensitivity image for Challenge 1 east.

 

image 

Figure 2.  Plot of values in the rms image versus sensivity image for challenge 2 east.

 

image

Figure 3.  Plot of values in rms image versus sensitivity image for challenge 3 high resolution. 

 

Values of the slopes measured by eye and assuming (0,0) intercept are in the table 1.  For the low resolution version of challenge 3 it was to difficult to measure the ratio due to confusion in the rms image.  I will take the challenge 3 high value.    Assuming that the differences between east and west are measurement error, I average the results to derive ratios of 292.5, 37.7 for challenges 1, 2 respectively.

Table 1.  Ratio of rms image values to sensitivity image values
              Challenge                              rms/sensitivity             
challenge 1 East 288.20
challenge 1 West 296.91
challenge 2 East 37.08
challenge 2 West 38.40
challenge 3 High 39.4
challenge 3 Low  90.2

 

Applying a threshold signal-to-noise of 4.5 seems to do a good job of removing the souces around the edge of the mosaic but leaving the sources in the middle.   I also filtered out a small number of sources for which the flux in the map at the source location did not support the presence of a source at that location. 

Table 2. Numbers of input sources and sources after filter S/N > 3.0.
        challenge            raw souce number            after filtering      
challenge 1 East 17564 8313
challenge 1 West 17690 8418
challenge 2 East 10247 2375
challenge 2 West 9909 2347
challenge 3 high 10276 3804
challenge 3 low 168  141

 

East/west and low/high files were merged into single ch1, ch2 and ch3 source files.  Repeat sources detected in both east/west and high/low  were rationalized.

challenge  Number of sources
1 16707
2 4717
3 3943