1 | import _asap
|
---|
2 |
|
---|
3 | class linefinder:
|
---|
4 | """
|
---|
5 | The class for automated spectral line search in ASAP.
|
---|
6 |
|
---|
7 | Example:
|
---|
8 | fl=linefinder()
|
---|
9 | fl.set_scan(sc,edge=(50,))
|
---|
10 | fl.set_options(threshold=3)
|
---|
11 | nlines=fl.find_lines()
|
---|
12 | if nlines!=0:
|
---|
13 | print "Found ",nlines," spectral lines"
|
---|
14 | print fl.get_ranges(False)
|
---|
15 | else:
|
---|
16 | print "No lines found!"
|
---|
17 | sc2=poly_baseline(sc,fl.get_mask(),7)
|
---|
18 |
|
---|
19 | The algorithm involves a simple threshold criterion. The line is
|
---|
20 | considered to be detected if a specified number of consequtive
|
---|
21 | channels (default is 3) is brighter (with respect to the current baseline
|
---|
22 | estimate) than the threshold times the noise level. This criterion is
|
---|
23 | applied in the iterative procedure updating baseline estimate and trying
|
---|
24 | reduced spectral resolutions to detect broad lines as well. The off-line
|
---|
25 | noise level is determined at each iteration as an average of 80% of the
|
---|
26 | lowest variances across the spectrum (i.e. histogram equalization is
|
---|
27 | used to avoid missing weak lines if strong ones are present). For
|
---|
28 | bad baseline shapes it is reccommended to increase the threshold and
|
---|
29 | possibly switch the averaging option off (see set_options) to
|
---|
30 | detect strong lines only, fit a high order baseline and repeat the line
|
---|
31 | search.
|
---|
32 |
|
---|
33 | """
|
---|
34 |
|
---|
35 | def __init__(self):
|
---|
36 | """
|
---|
37 | Create a line finder object.
|
---|
38 | """
|
---|
39 | self.finder = _asap.linefinder()
|
---|
40 | return
|
---|
41 |
|
---|
42 | def set_options(self,threshold=1.7320508075688772,min_nchan=3,
|
---|
43 | avg_limit=8,box_size=0.2):
|
---|
44 | """
|
---|
45 | Set the parameters of the algorithm
|
---|
46 | Parameters:
|
---|
47 | threshold a single channel S/N ratio above which the
|
---|
48 | channel is considered to be a detection
|
---|
49 | Default is sqrt(3), which together with
|
---|
50 | min_nchan=3 gives a 3-sigma criterion
|
---|
51 | min_nchan a minimal number of consequtive channels,
|
---|
52 | which should satisfy a threshold criterion to
|
---|
53 | be a detection. Default is 3.
|
---|
54 | avg_limit A number of consequtive channels not greater than
|
---|
55 | this parameter can be averaged to search for
|
---|
56 | broad lines. Default is 8.
|
---|
57 | box_size A running mean box size specified as a fraction
|
---|
58 | of the total spectrum length. Default is 1/5
|
---|
59 | Note: For bad baselines threshold should be increased,
|
---|
60 | and avg_limit decreased (or even switched off completely by
|
---|
61 | setting this parameter to 1) to avoid detecting baseline
|
---|
62 | undulations instead of real lines.
|
---|
63 | """
|
---|
64 | self.finder.setoptions(threshold,min_nchan,avg_limit,box_size)
|
---|
65 | return
|
---|
66 |
|
---|
67 | def set_scan(self,scan,mask=None,edge=(0,0)):
|
---|
68 | """
|
---|
69 | Set the 'data' (scantable) to work with.
|
---|
70 | Parameters:
|
---|
71 | scan: a scantable
|
---|
72 | mask: an optional mask retreived from scantable
|
---|
73 | edge: an optional number of channel to drop at
|
---|
74 | the edge of spectrum. If only one value is
|
---|
75 | specified, the same number will be dropped from
|
---|
76 | both sides of the spectrum. Default is to keep
|
---|
77 | all channels
|
---|
78 | """
|
---|
79 | if not scan:
|
---|
80 | raise RuntimeError, 'Please give a correct scan'
|
---|
81 | if len(edge)>2:
|
---|
82 | raise RuntimeError, "The edge parameter should have two \
|
---|
83 | or less elements"
|
---|
84 | if mask is None:
|
---|
85 | from numarray import ones
|
---|
86 | self.finder.setscan(scan,ones(scan.nchan()),edge)
|
---|
87 | else:
|
---|
88 | self.finder.setscan(scan,mask,edge)
|
---|
89 | return
|
---|
90 | def find_lines(self,nRow=0):
|
---|
91 | """
|
---|
92 | Search for spectral lines in the scan assigned in set_scan.
|
---|
93 | Current Beam/IF/Pol is used, Row is specified by parameter
|
---|
94 | A number of lines found will be returned
|
---|
95 | """
|
---|
96 | return self.finder.findlines(nRow)
|
---|
97 | def get_mask(self,invert=False):
|
---|
98 | """
|
---|
99 | Get the mask to mask out all lines that have been found (default)
|
---|
100 |
|
---|
101 | Parameters:
|
---|
102 | invert if True, only channels belong to lines will be unmasked
|
---|
103 |
|
---|
104 | Note: all channels originally masked by the input mask or
|
---|
105 | dropped out by the edge parameter will still be excluded
|
---|
106 | regardless on the invert option
|
---|
107 | """
|
---|
108 | return self.finder.getmask(invert)
|
---|
109 | def get_ranges(self,defunits=True):
|
---|
110 | """
|
---|
111 | Get ranges (start and end channels or velocities) for all spectral
|
---|
112 | lines found.
|
---|
113 |
|
---|
114 | Parameters:
|
---|
115 | defunits if True (default), the range will use the same units
|
---|
116 | as set for the scan (e.g. LSR velocity)
|
---|
117 | if False, the range will be expressed in channels
|
---|
118 | """
|
---|
119 | if (defunits):
|
---|
120 | return self.finder.getlineranges()
|
---|
121 | else:
|
---|
122 | return self.finder.getlinerangesinchannels()
|
---|
123 |
|
---|
124 | def auto_poly_baseline(scan, mask=None, edge=(0,0), order=0,
|
---|
125 | threshold=3,insitu=None):
|
---|
126 | """
|
---|
127 | Return a scan which has been baselined (all rows) by a polynomial.
|
---|
128 | Spectral lines are detected first using linefinder and masked out
|
---|
129 | to avoid them affecting the baseline solution.
|
---|
130 |
|
---|
131 | Parameters:
|
---|
132 | scan: a scantable
|
---|
133 | mask: an optional mask retreived from scantable
|
---|
134 | edge: an optional number of channel to drop at
|
---|
135 | the edge of spectrum. If only one value is
|
---|
136 | specified, the same number will be dropped from
|
---|
137 | both sides of the spectrum. Default is to keep
|
---|
138 | all channels
|
---|
139 | order: the order of the polynomial (default is 0)
|
---|
140 | threshold: the threshold used by line finder. It is better to
|
---|
141 | keep it large as only strong lines affect the
|
---|
142 | baseline solution.
|
---|
143 | insitu: if False a new scantable is returned.
|
---|
144 | Otherwise, the scaling is done in-situ
|
---|
145 | The default is taken from .asaprc (False)
|
---|
146 |
|
---|
147 | Example:
|
---|
148 | sc2=auto_poly_baseline(sc,order=7)
|
---|
149 | """
|
---|
150 | from asap.asapfitter import fitter
|
---|
151 | from asap import scantable
|
---|
152 |
|
---|
153 | # setup fitter
|
---|
154 |
|
---|
155 | f = fitter()
|
---|
156 | f._verbose(True)
|
---|
157 | f.set_function(poly=order)
|
---|
158 |
|
---|
159 | # setup line finder
|
---|
160 |
|
---|
161 | fl=linefinder()
|
---|
162 | fl.set_options(threshold=threshold)
|
---|
163 |
|
---|
164 | if not insitu:
|
---|
165 | workscan=scan.copy()
|
---|
166 | else:
|
---|
167 | workscan=scan
|
---|
168 |
|
---|
169 | vb=workscan._vb
|
---|
170 | # remember the verbose parameter and selection
|
---|
171 | workscan._vb=False
|
---|
172 | sel=workscan.get_cursor()
|
---|
173 | rows=range(workscan.nrow())
|
---|
174 | for i in range(workscan.nbeam()):
|
---|
175 | workscan.setbeam(i)
|
---|
176 | for j in range(workscan.nif()):
|
---|
177 | workscan.setif(j)
|
---|
178 | for k in range(workscan.npol()):
|
---|
179 | workscan.setpol(k)
|
---|
180 | if f._vb:
|
---|
181 | print "Processing:"
|
---|
182 | print 'Beam[%d], IF[%d], Pol[%d]' % (i,j,k)
|
---|
183 | for iRow in rows:
|
---|
184 | fl.set_scan(workscan,mask,edge)
|
---|
185 | fl.find_lines(iRow)
|
---|
186 | f.set_scan(workscan, fl.get_mask())
|
---|
187 | f.x=workscan._getabcissa(iRow)
|
---|
188 | f.y=workscan._getspectrum(iRow)
|
---|
189 | f.data=None
|
---|
190 | f.fit()
|
---|
191 | x=f.get_parameters()
|
---|
192 | workscan._setspectrum(f.fitter.getresidual(),iRow)
|
---|
193 | workscan.set_cursor(sel[0],sel[1],sel[2])
|
---|
194 | workscan._vb = vb
|
---|
195 | if not insitu:
|
---|
196 | return workscan
|
---|