Files | |
| file | HoughMap.h |
|
| |
| file | Statistics.h |
| Functions prototypes and structures for calculating statistical properties of Hough maps. | |
| file | Velocity.h |
|
| |
, each residual spin-down parameter, and each refined sky location inside the patch. Notice, from the master equation, that the effect of the residual spin-down parameter is just a change in
, and, at any given time,
can be considered constant. Also, the Hough map is a histogram, thus additive. It can be seen as the sum of several partial Hough maps constructed using just one periodogram.
Therefore, we can construct the HM for any
and spin-down value by adding together, at different times, partial Hough maps (PHM) corresponding to different
values (or equivalently, adding their derivatives and then integrating the result).
In practice this means that in order to obtain the HM for a given frequency and all possible residual spin-down parameters, we have to construct a CYLINDER of around the frequency
. All of the phmd coming from data demodulated with the same parameters. The coordinates of the phmd locate the position of the source in the sky, and by summing along different directions inside the cylinder we refine the spin-down value. To analyze another frequency, for all possible spin-down parameters, we just need to add a new line to the cylinder (and remove another one, in a circular buffer) and then proceed making all the possible sums again.
For the case of only 1 spin-down parameter we have to sum following straight lines whose slope is related to the grid in the residual spin-down parameter. We can distinguish (at most) as many lines as the number of the different periodograms used.
History: Created by Sintes May 11, 2001 Modified by Badri Krishnan Feb 2003
In principle, the subroutines provided are valid for { any} Hough master equation of the form:
where
is the measured frequency of the signal at time
,
intrinsic frequency of the signal at that time,
location of the souce in the sky,
the center of the sky patch used in the demodulation procedure, and
any vector.
The form of this vector
depends on the demodulation procedure used in the previous step. In our case this corresponds to
and
where
is the velocity of the detector,
is the detector position,
is the time at the solar system barycenter (for a given sky location
),
,
the residual spin-down parameter,
the spin-down parameter used in the demodulation, and
,
the intrinsic frequency and spin-down parameters of the source at time
.\
Looking at the generic Hough master equation, one realizes that for a fixed time, a given value of
, and a measured frequency
(from a selected peak), the source could be located anywhere on a circle (whose center points in the same direction of
and is characterized by
, the angle between
and
). Since the Hough transform is performed on a set of spectra with discrete frequencies, a peak on the spectrum appearing at
could correspond to any source with a demodulated frequency in a certain interval. As a consequence, the location of the sources compatible with
and
is not a circle but an annulus with a certain width.\
Our purpose is to map these annuli on a discrete space. An estimation of the average thickness of the annuli tells us that the vast majority of annuli will be very thin, and therefore our algorithm should not be optimized for drawing thick annuli but for thin ones. Also, the mapping implementation should be one with a uniform probability distribution in order to avoid discretization errors. In order to remove border effects, we use a biunivocal mapping, which requires that a pixel in a partial Hough map can belong only to one annulus, just touched by one peak of the spectrum. The criteria for the biunivocal mapping is that if and only if the center of the pixel is inside the annulus, then the pixel will be enhanced.\
In order to simplify (reduce the computational cost of) this task we construct look up tables ({ lut}) where the borders of these annuli are marked for any possible
value. Since we work on a discrete space these { lut} are valid for many
values.\
1.5.2