For version 3.0b27; last updated 6/22/00
The guide camera image controller can control one or more guide cameras. For each camera it can take exposures, finds stars in images, determine accurate positions for stars and write images to FITS files. Image files can be retrieved via ftp and AppleShare. The Guider Image Controller is implemented in a Macintosh computer and commanded via a serial port. You cannot yet trigger a reset via the serial port, so if the Macintosh locks up you will probably have to go manually reboot it.
Commands are described in the Guider Image Controller Commands Manual and one of the commands is "help".
Each time the camera is read, the image is automatically saved to a fits file on a shared disk. The name of the most recently saved file is contained in the file "last.image". The number in the file name is automatically incremented and rolls over to one after it reaches a user-settable value (see the command setmaxfiles), so old images will be overwritten unless a copy is explicitly saved somewhere else.
Star finding is done using an algorithm developed by Jeff Morgan at the University of Washington. It involves dividing the image into "super pixels", each of which contain summary information about a square of regular pixels. The size of a super pixel is chosen to be the size of an ordinary star. Very bright stars are assumed to cover a region of 5x5 super pixels. If two very bright regions overlap then both are rejected. If three bright pixels line on a line along rows or columns, all three are rejected. (Two bright adjacent super pixels are fine, as they are assumed to be one star that lies on the boundary between the two super pixels). Other tests are also used to reject pixels. The remaining bright super pixels (or sets of super pixels in the case of very bright stars) are centroided.
The algorithm and code are the SDSS centroiding code, as originally written by Jim Gunn and reworked into SDSS form by Heidi Newberg. The main routine is atDACentroid, which calls atObjectCenterFind_Ptr to do the centroiding, atEllipseFind to measure the shape of the star and atSigmaFind for whatever reason.
Cameras that have bad pixels require "bad pixel maps" in order to find stars, centroid or even compute statistics for the image. These maps are used to determine which pixels to ignore when computing statistics, which stars to ignore when finding stars, and which stars to reject when centroiding.
Bad pixel maps are binary files containing pixel values of 0 (good) or any other value (bad). However, for the sake of coding simplicity, they have 16 bits per pixel, just like normal images. They are stored in the directory "bad pix maps" and are named as follows: "bad pixel map <camIDNum> <bin x> <bin y>". If a map doesn't exist for the particular binning factors that were used to take an image, then the operations mentioned above cannot be performed. (If this turns out to be too severe a restriction , I can change the code to calculate binned maps as needed. But the current system is faster and simpler).
Bad pixel maps may be created as follows: take an image that shows the defects in good contrast. A flat field would be a good candidate. View this image (for instance in NIH Image) and determine an upper and lower cutoff; pixels whose values are outside of this range are considered bad. You may wish to process the image to enhance the defective areas. If you have changed the image, save the changed version and manually load it. The issue the "makebadpixmap" command, using the threshold values you determined, save the resulting bad pixel map using the naming convention described above, and move it to the appropriate directory.
The controller is a Macintosh. The software can easily support any Photometric camera that can be driven by the PVCAM software library. As of this writing it supports three kinds of camera: two sizes of Photometrics SenSys cameras and one size of Photometrics PXL camera. Adding support for additional models of SenSys or PXL cameras consists of changing a data table defined in a header file, possibly modifying cam.init and recompiling and relinking the software. Adding support for other kinds of cameras requires modifying some interface subroutines, but should be fairly easy if the camera comes with a reasonable C or C++ interface library.
The system is written in Metrowerks C++. Most of the code is ANSI standard. The serial port code is Macintosh-specific, but can easily be replaced. The application contains a very tiny amount of additional Mac-specific code in "main" to handle opening the log window.
6/22/00 Added a troubleshooting section.
6/23/99 Previous version.
