Realtime Automated Analysis

There are two programs that act in tandem to produce calibrated object lists, and they are referred to as ``pacman'' programs, because they ``eat'' one file at a time as they are produced during the analysis process, in analogy to the arcade game figure that eats a row of little pellets.

The process begins when camserverd records a new image to disc and creates a link to that image file in the specified link directory (Section 1.2.4). A Perl script called sexpacman.pl monitors that directory for new links, and when one appears, it corrects the image with the available dark- and flat-fields, and processes the result through SExtractor to produce a list of objects in the field. This list is called an sobj file. sexpacman.pl then adds the name of the sobj file to a list of other sobj files, sorted according to priority (most images are processed chronologically, but GRB follow-up images outrank other images, and prompt burst images outrank late-time follow-up images. See the explanation for the scheduler in Section 4.3.4 for more information on these triggers.).

The list of sobj files is monitored by an idl process called idlpacman.pro, hereafter referred to simply as idlpacman. The first file in the list is read by idlpacman and calibrated against the USNO A2.0 catalog to produce a list of estimated R-band magnitudes and celestial locations for these sources. These lists are called cobj files. The name of the processed sobj file is removed from the list, and idlpacman moves to the next file name on the list, or waits for a name to be added, if no more are present. The time elapsed to the creation of a calibrated object list is typically 45 s from the first appearence of the raw image.

Figure 8.1: Flowchart depicting the ROTSE-III data analysis pipeline. Most of these steps have been implemented in an automated realtime process. Based on the file names, the software decides how far to pursue the analysis; prompt burst response data will be compiled into match structures and scanned for highly variable, uncataloged objects. Object lists from other kinds of triggers are created and stored for future processing. The relative photometry process is only intended to be run manually on these archived data.