SKICAT: Sky Image Cataloging and Analysis Tool

1 Introduction

In astronomy and space sciences, we currently face a data glut crisis. The problem of dealing with the huge volume of data accumulated from a variety of sources, of correlating the data and extracting and visualizing the important trends, is now fully recognized. This problem will become more acute very rapidly, with the advent of new telescopes, detectors, and space missions, with the data flux measured in terabytes. We face a critical need for information processing technology and method­ ology with which to manage this data avalanche in order to produce interesting scientific results quickly and effi­ ciently. Developments in the fields of machine learning and AI can provide at least some solutions. Much of the future of scientific information processing lies in the implementation of these methods.

We present an application of supervised classification to the automation of the tasks of cataloging and ana­ lyzing objects in digitized sky images. The Sky Image Cataloging and Analysis Tool (SKICAT) was developed for use on the images resulting from the 2nd Palomar Observatory Sky Survey (POSS-11) conducted by the California Institute of Technology ( Caltech). The photo­ graphic plates collected from the survey are digitized at the Space Telescope Science Institute. This process will result in about 3,000 digital images of 23, 040 x 23, 040 16-bit pixels each, totalling over 3 terabytes of data. When complete, the survey will cover the entire northern sky in three colors, detecting virtually every sky object down to a B magnitude of 22. This is at least one magni­ tude fainter than previous comprable photographic sur­ veys. We estimate that there are on the order of 10 7 galaxies 10 9 stellar objects (including over 10 5 quasars) are detectable in this survey. This data set will be the most comprehensive large-scale imaging survey produced to date and will not be surpassed in scope until the com­ pletion of a fully digital all-sky survey.

The purpose of SKI CAT is to enable and maximize the extraction of meaningful information from such a large database in timely manner. The system is built in a modular way, incorporating several existing algorithms and packages. There are three basic functional compo­ nents to SKI CAT, serving the purposes of sky object cat­ alog construction, catalog management, and high-level statistical and scientific analysis.

2 Classifying Sky Objects

The first step in analyzing the results of a sky survey is to identify, measure, and catalog the detected obvalues of the attributes measured from the lower resolu­ tion plate image. Hence, in principle, the classifier will be able to classify objects in the photographic image that are simply too faint for an astronomer to classify by in­ spection. Using the class labels, the learning algorithms are basically being used to solve the more difficult prob­ lem of separating the classes in the multi-dimensional space defined by the set of attributes derived via image processing. This method is expected to allow us to clas­ sify objects that are at least one magnitude fainter than objects classified in photographic all-sky surveys to date.

By effectively defining robust features, we were able to obtain classifiers with an accuracy exceeding that of humans for faint objects [13] Since faint objects consti­ tute the majority of objects on any plate, this results in a dramatic increase in the number of classified objects available for further scientific analysis. In effect, this shows that the pixels contained important information that the human visual system could not extract. Projec­ tion of the high-dimensional pixel space onto a powerful lower-dimensional feature space allowed us to transform the problem into one that is easily solvable by a super­ vised learning algorithm. By defining additional "normal­ ized" image-independent attributes, we were able to ob­ tain high accuracy classifiers within and across photo­ graphic plates.

3 Conclusions

The implications of a tool like SKICAT for Astronomy may indeed be profound. One could reclassify any por­ tion of the survey using alternative criteria better suited to a particular scientific goal (e.g. star catalogs vs. galaxy catalogs). This changes the notion of a sky cat­ alog from the classical static entity "in print", to a dy­ namic, ever growing, ever improving, on-line database. The catalogs will also accommodate additional attribute entries, in the event other pixel-based measurements are deemed necessary. An important feature of the survey analysis system will be to facilitate such detailed interac­ tions with the catalogs. The catalog generated by SKI­ CAT will eventually contain about two billion entries representing hundreds of millions of sky objects. Unlike the traditional notion of a static printed catalog, we view our effort as targeting the development of a new gener­ ation of scientific analysis tools that render it possible to have a constantly evolving, improving, and growing catalog. Without the availability of these tools for the first survey (POSS-1) conducted over four decades ago, no objective and comprehensive analysis of the data was possible. In contrast, we are targeting a comprehensive sky catalog that will be available on-line for the use of the scientific community.

For future work, we are sudying the application of clustering algorithms (e.g. AutoClass [2]) to automate discovery of new classes in the large catalog databases created by SKICAT [13; 14; 3]