Increasing the reliability of ISOCAM cross-identifications by use of a probability pattern

The built-in play at the wheels of the ISOCAM instrument affects the accuracy of the astrometry of sources from the ISOCAM Parallel Catalogue. We present the results of different methods for cross-identifying ISOCAM sources with reference catalogues in the optical (USNO-A2.0) and in the near infrared (2MASS). For a sample of 7197 ISOCAM sources, the fraction having a real counterpart in the USNO and 2MASS catalogues was estimated to be 95% and 91% respectively. Using a nearest neighbor method for cross-identification, only 74% of the possible associations with USNO (79% with 2MASS) are retrieved. With the use of a so-called probability pattern to perform the associations, these values increase to 77% for USNO and 83% for 2MASS. In addition, the overall reliability of associations increases from 77 to 87% with USNO, and from 81 to 90% with 2MASS.