Home       Trees       Index       Help   

Project Homepage

Package asap :: Module asaplinefind :: Class linefinder
[show private | hide private]
[frames | no frames]

Class linefinder

The class for automated spectral line search in ASAP.

Example:
   fl=linefinder()
   fl.set_scan(sc)
   fl.set_options(threshold=3)
   nlines=fl.find_lines(edge=(50,0))
   if nlines!=0:
      print "Found ",nlines," spectral lines"
      print fl.get_ranges(False)
   else:
      print "No lines found!"
   sc2=sc.poly_baseline(fl.get_mask(),7)

The algorithm involves a simple threshold criterion. The line is
considered to be detected if a specified number of consequtive
channels (default is 3) is brighter (with respect to the current baseline
estimate) than the threshold times the noise level. This criterion is
applied in the iterative procedure updating baseline estimate and trying
reduced spectral resolutions to detect broad lines as well. The off-line
noise level is determined at each iteration as an average of 80% of the
lowest variances across the spectrum (i.e. histogram equalization is
used to avoid missing weak lines if strong ones are present). For
bad baseline shapes it is recommended to increase the threshold and
possibly switch the averaging option off (see set_options) to
detect strong lines only, fit a high order baseline and repeat the line
search.
Method Summary
  __init__(self)
Create a line finder object.
  find_lines(self, nRow, mask, edge)
Search for spectral lines in the scan assigned in set_scan.
  get_mask(self, invert)
Get the mask to mask out all lines that have been found (default)...
  get_ranges(self, defunits)
Get ranges (start and end channels or velocities) for all spectral lines found.
  set_options(self, threshold, min_nchan, avg_limit, box_size)
Set the parameters of the algorithm Parameters: threshold a single channel S/N ratio above which the channel is considered to be a detection Default is sqrt(3), which together with min_nchan=3 gives a 3-sigma criterion min_nchan a minimal number of consequtive channels, which should satisfy a threshold criterion to be a detection.
  set_scan(self, scan)
Set the 'data' (scantable) to work with.

Method Details

__init__(self)
(Constructor)

Create a line finder object.

find_lines(self, nRow=0, mask=[], edge=(0, 0)) 

Search for spectral lines in the scan assigned in set_scan.
Parameters:
     nRow:       a row in the scantable to work with
     mask:       an optional mask (e.g. retreived from scantable)
     edge:       an optional number of channels to drop at
                 the edge of the spectrum. If only one value is
                 specified, the same number will be dropped from
                 both sides of the spectrum. Default is to keep
                 all channels
A number of lines found will be returned

get_mask(self, invert=False) 

Get the mask to mask out all lines that have been found (default)

Parameters:
      invert  if True, only channels belong to lines will be unmasked

Note: all channels originally masked by the input mask or
      dropped out by the edge parameter will still be excluded
      regardless on the invert option

get_ranges(self, defunits=True) 

Get ranges (start and end channels or velocities) for all spectral
lines found.

Parameters:
      defunits  if True (default), the range will use the same units
                as set for the scan (e.g. LSR velocity)
                if False, the range will be expressed in channels

set_options(self, threshold=1.7320508075688772, min_nchan=3, avg_limit=8, box_size=0.20000000000000001) 

Set the parameters of the algorithm
Parameters:
     threshold    a single channel S/N ratio above which the
                  channel is considered to be a detection
                  Default is sqrt(3), which together with
                  min_nchan=3 gives a 3-sigma criterion
     min_nchan    a minimal number of consequtive channels,
                  which should satisfy a threshold criterion to
                  be a detection. Default is 3.
     avg_limit    A number of consequtive channels not greater than
                  this parameter can be averaged to search for
                  broad lines. Default is 8.
     box_size     A running mean box size specified as a fraction
                  of the total spectrum length. Default is 1/5
Note:  For bad baselines threshold should be increased,
       and avg_limit decreased (or even switched off completely by
       setting this parameter to 1) to avoid detecting baseline
       undulations instead of real lines.

set_scan(self, scan) 

Set the 'data' (scantable) to work with.
Parameters:
     scan:    a scantable

   Home       Trees       Index       Help   

Project Homepage

Generated by Epydoc 2.1 on Wed May 2 14:24:32 2007 http://epydoc.sf.net