Extracting compact sources

Say you have a FITS image (e.g. input.fits) with your radio observations and you want to extract compact and point-sources. You can use the executable FindSource or FindSourceMPI (which supports parallel computation) installed in the CAESAR bin directory to this aim:

$ ./FindSource --config=[config file]

$ mpirun -np [NPROC] ./FindSourceMPI --config=[config file] [--no-mpi]

A sample configuration file to search sources in the test image ScorpioSNRField.fits is reported below:

#===================
#==    MAIN       ==
#===================
inputFile = ScorpioSNRField.fits                    | Input image filename (ROOT or FITS)
##
//===========================
//==   BKG OPTIONS         ==
//===========================
useLocalBkg = true                                                                                                                                  | Use local background calculation instead of global bkg (T/F)
bkgEstimator = 2                                                                                                                                            | Background estimator (1=Mean,2=Median,3=BiWeight,4=ClippedMedian)
useBeamInfoInBkg = true                             | Use beam information in bkg box definition (if available) (T/F)
boxSizeX = 20                                                                                                                                                         | X Size of local background box in #pixels
boxSizeY = 20                                                                                                                                                         | Y Size of local background box in #pixels
gridSizeX = 0.2                                                                                                                                                     | X Size of local background grid used for bkg interpolation
gridSizeY = 0.2                                                                                                                                                     | Y Size of local background grid used for bkg interpolation
###
###
//====================================
//==  SOURCE FINDING OPTIONS        ==
//====================================
searchCompactSources = true                                                                                                 | Search compact sources (T/F)
minNPix = 5                                                                                                                                                           | Minimum number of pixel to consider a source
seedThr = 5                                                                                                                                                                 | Seed threshold in flood filling algo for faint sources
mergeThr = 2.6                                                                                                                                                      | Merge/aggregation threshold in flood filling algo
compactSourceSearchNIters = 2                       | Number of iterations to be performed in compact source search (default=10)
seedThrStep = 0.5                                   | Seed threshold decrease step size between iteration (default=1)
###
###
//===========================================
//==  NESTED SOURCE FINDING OPTIONS        ==
//===========================================
searchNestedSources = false                                                                                                 | Search for nested sources inside candidate sources (T/F)
//==================================
//==  SOURCE FITTING OPTIONS      ==
//==================================
fitSources = false                                  | Fit sources with multi-component gaus fit (T/F)

With the above configuration, after computing the significance map, the image is scanned for compact sources with a flood-fill algorithm, assuming a seed significance threshold of 5 sigmas and an aggregation threshold of 2.6 sigmas. A minimum number of pixels equal to 5 is considered in blob finding. Two search iterations are performed, decreasing detection threshold by 0.5 sigmas at each iterations. No negative excess wrt background are searched. No nested sources are searched. No fitting is performed over the detected sources.

A region file is created with the detected source as well as a ROOT file with full source information stored in the SourceInfo ROOT TTree. Each source in the tree is a Source object. You can view the original image with detected sources in DS9 as:

$ ds9 ScorpioSNRField.fits -regions ds9.reg

The image below shows the compact sources detected in the sample field

Input image with detected compact sources (blue lines) and point-sources (red lines)

You can also get full access to detailed source info (parameters, contours, etc) from the ROOT output file using macros, CLI, etc. Source info is stored in the SourceInfo ROOT tree. For example, the following code gets access to all detected sources and dumps some basic source parameters:

{
  //Open output file with source finding results
  TFile* f= new TFile("output.root","READ");

  //Get access to source data tree
  TTree* data= (TTree*)f->Get("SourceInfo");

  Source* source= 0;
  data->SetBranchAddress("Source",&source);

  //Loop over all detected sources
  for(int i=0;i<data->GetEntries();i++)
  {
     //Get i-th source
     data->GetEntry(i);

     //Dump source info
     source->Print();

     //Do other stuff with i-th source object (see Source API)
     //...
  }
}