IPP Software Navigation Tools IPP Links Communication Pan-STARRS Links

Changeset 6054


Ignore:
Timestamp:
Jan 18, 2006, 8:50:18 PM (21 years ago)
Author:
eugene
Message:

updates

Location:
trunk/doc/design
Files:
2 added
4 edited

Legend:

Unmodified
Added
Removed
  • trunk/doc/design/Makefile

    r6018 r6054  
    1 # $Id: Makefile,v 1.13 2006-01-16 18:58:45 eugene Exp $
     1# $Id: Makefile,v 1.14 2006-01-19 06:49:50 eugene Exp $
    22
    33PDFLATEX = env TEXINPUTS=../../latex/inputs:$(TEXINPUTS):.: pdflatex
     
    88        @echo "  targets: srs ssdd scd all"
    99
     10cdr: ippCDR.pdf
    1011srs: ippSRS.pdf
    1112ssdd: ippSSDD.pdf
  • trunk/doc/design/ippSSDD.tex

    r6049 r6054  
    1 %%% $Id: ippSSDD.tex,v 1.4 2006-01-19 03:51:41 eugene Exp $
     1%%% $Id: ippSSDD.tex,v 1.5 2006-01-19 06:49:50 eugene Exp $
    22\documentclass[panstarrs]{panstarrs}
    33
     
    137137PSDC-430-012  &   Pan-STARRS IPP Modules Supplementary Design Requirements Specification \\ \hline
    138138PSDC-430-014  &   Pan-STARRS IPP PS-1 Cluster Support \\ \hline
    139 \tbd{add the other subsystem SDDs}
    140139\DocumentsExternalSection
    141140Posix Standard & Open Group Based Specifications Issue 6, IEEE Std 1003.1, 2003 \\
    142141\DocumentsEnd
     142
     143\tbd{add the other subsystem SDDs}
    143144
    144145\section{Subsystem Overview}
     
    605606names.
    606607
    607 \subsubsubsection{House keeping}
    608 
    609 \paragraph {Lock sweeping} In the event that a Storage Object operation fails to complete successfully
     608\paragraph{House keeping}
     609
     610\subparagraph{Lock sweeping} In the event that a Storage Object operation fails to complete successfully
    610611stale locks will have to be identified and removed from the IPP Pixel
    611612Data Server Database.  This should be done periodically by comparing
     
    622623table.
    623624
    624 \paragraph{Consistency sweeping} Periodically the IPP Pixel Data Server meta-data and Storage Object will need
     625\subparagraph{Consistency sweeping} Periodically the IPP Pixel Data Server meta-data and Storage Object will need
    625626to be checked for sanity.  This would be similar to running fsck on a
    626627modern filesystem.  Consistency sweeping should include Lock sweeping
     
    20312032Phase 2.
    20322033
     2034\subsection{Summary of the Phase 1-3 Data Products}
     2035
     2036
    20332037%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
    20342038
     
    20422046sky, the sky cell, along with the associated pixels from a collection
    20432047of images which have been processed through phases 1--3.  The size and
    2044 exact representation of a static sky cell are yet to be determined.
    2045 The working concept is that the static sky cells contain roughly the
    2046 same number of pixels as an OTA (4k x 4k) and represent a portion of a
    2047 local tangent plane projection.  In order to meet the image
    2048 degradation requirements, the pixel scale of the static sky is planned
    2049 to be 0.2\arcsec, somewhat smaller than the 0.3\arcsec\ raw image
    2050 pixel scale.
    2051 
    2052 For each sky cell, the corresponding pixels are extracted from the
    2053 exposures being processed and mapped to the projection of the sky
     2048exact representation of a static sky cell are configurable parameters
     2049which may vary between surveys.  In some cases, it may be feasible to
     2050use a single large rectangular tangent plane image for the static sky
     2051and accept the distortion near the edges.  In other cases, the static
     2052sky cells may be more appropriate defined to represent a relatively
     2053small patch over which the distortion is known to be small.  The
     2054details of the static sky cell definition should not be a driver for
     2055the Phase 4 analysis.  In order to meet the image degradation
     2056requirements, the pixel scale of the static sky is planned to be
     20570.2\arcsec, somewhat smaller than the 0.25\arcsec\ raw image pixel
     2058scale;  the choice of this scale must be determined with some caution.
     2059If the pixels are too large, the degredation will be excessive; too
     2060small, and the total volume of the static sky image data will be
     2061excessively costly.
     2062
     2063\begin{figure}
     2064\begin{center}
     2065\resizebox{6in}{!}{\includegraphics{pics/phase4}}
     2066\caption{ \label{fig:phase2} Phase 2 dataflow - this diagram is old: update}
     2067\end{center}
     2068\end{figure}
     2069
     2070In the basic concept of the Phase 4 analysis, each sky cell is
     2071examined independently.  The corresponding pixels are extracted from
     2072the exposures being processed and mapped to the projection of the sky
    20542073cell. The pixels from the multiple input processed images are combined
    20552074into a single, cleaned image.  Outlier pixels may be optionally
     
    20612080The remaining pixels are added to the existing static sky image.
    20622081Object detection must be performed on the difference ($P4\Delta$) and
    2063 cleaned ($P4\Sigma$) images.
     2082cleaned ($P4\Sigma$) images.  This process is illustrated graphically
     2083in Figure~\ref{BasicP4}
    20642084
    20652085\subsubsection{Image Warping and Matching}
     
    21542174a time when the computing infrastructure is not under significant load.
    21552175
    2156 %\begin{figure}
    2157 %\begin{center}
    2158 %\resizebox{6in}{!}{\includegraphics{pics/phase4}}
    2159 %\caption{ \label{fig:phase4} Phase 4 dataflow}
    2160 %\end{center}
    2161 %\end{figure}
     2176\subsubsection{Magic and Phase 4 Modifications}
     2177
     2178\begin{figure}
     2179\begin{center}
     2180\resizebox{6in}{!}{\includegraphics{pics/phase4a}}
     2181\caption{ \label{fig:phase4a} Phase 4a}
     2182\end{center}
     2183\end{figure}
     2184
     2185The Pan-STARRS relationship with the U.S. Air Force has some
     2186implications for the data processing which place some interesting
     2187constraints on the IPP implementation.  The U.S. Air Force is funding
     2188the construction of PS-1, and we are thus subject to restrictions
     2189under which the Air Force must operate telescopes.  The Air Force has
     2190diplomatic and security concerns about publically releasing images in
     2191which artificial satelites are detected, particularly if those
     2192satelites are identified.  Historically, Air Force projects have been
     2193restricted from releasing images with identified satelites,
     2194considering the data to be of a sensitive nature. 
     2195
     2196The language which governs such Air Force projects is in the process
     2197of being modified so that images with identifable satelites can be
     2198treated under the classification of 'For Official Use Only'.  Even
     2199under such an arrangement, however, the Air Force requires that the
     2200satelites which appear in the PS-1 images be excised in such a way
     2201that their orbits cannot be reliably identified. 
     2202
     2203Satelite detected in the Pan-STARRS images will appear as long ($>
     2204100$ pixels) streaks.  The natural, and most effective, way to
     2205identify such streaks is to search for them in the images after they
     2206have been processed using the difference image processing in Phase 4.
     2207A team from Boeing has been contracted to develop a software module to
     2208idenfity streaks in these images.  They delivered the initial release
     2209of their software, which is currently called {\em Magic}, in the
     2210beginning of January 2006.  Further effort on that software will be
     2211required to confront it with the real PS-1 image parameters. 
     2212
     2213In order to mesh the operation of {\em Magic} with the IPP Phase 4
     2214analysis for PS-1, some modifications must be made to the operation
     2215sequence.  In the case of PS-1, unlike PS-4, the basic set of images
     2216which confronted with the Static Sky image are obtained in sequence,
     2217not essentially simultaneously.  This is important for Magic because
     2218the satelite streak in one image will not appear in the other three.
     2219Thus the Magic operation must be performed on intermediate difference
     2220images for individual exposures.  To facilitate this, the basic Phase
     22214 described above is divided into three stages: Phase 4a, in which the
     2222individual difference frames are constructed; Magic; and Phase 4b, in
     2223which the results of Magic are used to filter the images before the
     2224final stack and difference images are generated.
     2225
     2226Figure~\ref{phase4a} illustrates the operation of Phase 4a and Magic.
     2227The individual warped image (P4$_W$) is constructed and differenced
     2228against the corresponding Static Sky image.  In this analysis, poisub,
     2229which performs the kernel matching and difference image construction,
     2230is run with somewhat looser constraints on the image difference
     2231kernel.  The lower-accuracy difference kernel will make the difference
     2232image noisier in the vicinity of galaxies and bright stars.  Unlike
     2233the science transients, however, most of the pixels involved in a
     2234satelite streak are not in close proximity to another objects; for SNe
     2235and GRBs, the objects are preferentially located near host galaxies
     2236where it is most critical to have a high-quality subtraction.
     2237Allowing poisub a more relaxed image difference kernel results in a
     2238substantial improvement in the convolution speed.  The resulting
     2239difference image is called P4$\Delta$'. 
     2240
     2241{\em Magic} searches for streakes in the difference images by using a
     2242heirarchy of Huff transforms.  A single run of {\em Magic} expects to
     2243have access to all of the P4$\Delta$' images from a single FPA image.
     2244It starts with independent Huff transforms of the individual cells,
     2245which can be performed in parallel under the PanTasks paralleliztion
     2246scheme.  The results of these Huff transforms are then merged to yield
     2247the equivalent of a larger Huff-transformed image.  The heirarchy
     2248searches for significant streaks from the individual cells, the
     2249equivalent 2x2 cell sets, then 2x2 of those, etc, until the entire
     2250array has been searched.  Significant peaks detected from the Huff
     2251transforms are then used to restrict the pixels in the image space.
     2252These pixels are examined and streaks detected by requiring a certain
     2253fraction of the pixels along the purported streak to be lit.
     2254Filtering using the cross-streak PSF can also be applied to enhance
     2255the detection and minimize the false positives. Any streaks which are
     2256detected are then excisized.  In order to obfuscate the idenification
     2257of the satelite, this step masks the streak pixels, with a box
     2258somewhat wider than the streak (10-20 pixel wide), displaced by an
     2259unknown, random amount relative to the streak center-line.  This strip
     2260is extended to the ends of the FPA array.  Analysis of the expected
     2261density of streaks by the Boeing team, using known satelites as the
     2262input to the simulation, shows that we should expect to lose less than
     22631\% of the pixels due to these long excised regions.
     2264
     2265The result of the Magic process is a set of masks for each raw PS-1
     2266image.  Any images which are released beyond the IPP cluster must have
     2267these masks applied, or with manual inspection for visible streaks.
     2268\tbd{time-stamps? ok to release without the valid time? or only with }
     2269
     2270The final portion of the analysis, Phase 4b, is illustrated in
     2271Figure~\ref{phase4b}.  In this stage, the intermediate result images
     2272from Phase 4a can be used.  The P4$_W$ images are masked with the
     2273corresponding masks from Magic.  These images are then stacked as
     2274normal, and the resulting summed image is differenced with poisub,
     2275using a more stringent limit on the image difference kernel.
     2276
     2277\begin{figure}
     2278\begin{center}
     2279\resizebox{6in}{!}{\includegraphics{pics/phase4b}}
     2280\caption{ \label{fig:phase4b} Phase 4b}
     2281\end{center}
     2282\end{figure}
    21622283
    21632284%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
     
    27482869\item {\bf output stage} as a stand-alone program, psphot will produce
    27492870  output tables in various formats.
     2871\end{itemize}
    27502872
    27512873\subsection{psastro}
     
    27732895and 3.
    27742896
    2775 In the conceptualy most straightforeward mode, each readout contained
     2897In the conceptually most straightforeward mode, each readout contained
    27762898in the incoming FPA structure is treated independently.  The metadata
    27772899describing the approximate astrometry of the readout are used to guess
     
    28112933basic description of the astrometry is the collection of header
    28122934keywords which define the boresite center coordinates (RA, DEC), the
    2813 rotation and platescales
     2935location of the chip data arrays within the full mosaic (the IRAF
     2936DATASEC keywords).  More detailed astrometric information may be
     2937defined using the WCS keywords.  These unfortunately do not have a
     2938standard representation of higher-order terms.  Furthermore, the two
     2939competing systems which have been proposed define only a single
     2940transformation frame.  In the IPP, it is important to carry around
     2941more information which can be used to improve our astrometric
     2942solutions in the future.  Specifically, we would like to maintain at
     2943least the transformation for the telescope optics independent from the
     2944individual chip warps or tilts.  Even more, we would like to have a
     2945flexible astrometry definition format which can be extended in a
     2946flexible fashion.  We have defined a FITS table convention to carry
     2947all of the elements of the astrometric transformation of a full FPA.
     2948Within the table, transformations are generally defined to convert one
     2949layer (eg, focal plane) to another.  The form and the parameters of
     2950the conversion make up the columns of the table.  With this structure,
     2951it is possible to add arbitrary layers as needed.  The IPP, and
     2952portions of the project (particularly Otis), will share a common
     2953default astrometry model for the telescope.  This will be defined on
     2954the basis of measurements over the first weeks of observations.  The
     2955IPP will perform the astrometric analysis of individual images with
     2956psastro, and the results can be saved in this tabular format.  Over
     2957time, these result tables can be used to improve the astrometric model
     2958for the telescope, and to improve the astrometric reference catalog.
     2959
     2960\subsection{poisub}
     2961
     2962Poisub is the image difference analysis program.  \tbd{Paul: please
     2963  flesh this out!}.
     2964
     2965\subsection{stac}
     2966
     2967STAC is the program which warps and optimally combines images from the
     2968same region of the sky.  It consists of two major stages: the warping
     2969stage and the image combination stage with robust outlier rejection.
     2970\tbd{Paul: flesh this out!}
    28142971
    28152972\section{Interfaces}
     
    38173974------
    38183975
    3819 * poisub / stack
    3820 * re-org the Phase 4 stuff to discuss Magic
    3821 * astrometry calibration data formats
     3976* output data products
     3977* DVO
     3978* PanTasks
     3979* ipptools / ippMonitor?
    38223980* analysis stages, versions and iterations
    3823 * output data products
Note: See TracChangeset for help on using the changeset viewer.