The Finnish Meteorological Institute operates five digital all-sky cameras, which routinely monitor the auroral emissions in Northern Finland, Sweden, and Svalbard; each camera records an image of the fall sky at 20-s interval at a resolution of 512 x 512 pixels. As a result, over 2.5 million images are recorded each winter. We demonstrate a syntactic pattern recognition algorithm for searching auroral arcs in the all-sky images. The algorithm operates in two phases: a quicksearch and a detection phase. The quicksearch is based on simple correlation to a model axe and is utilised to scan through the database to locate potential auroras. In the detection, an auroral image skeleton is determined, and, based on the shape skeleton, auroral forms known as "arcs" are located. A quality estimate is calculated for each arc. Trial runs indicate that the algorithms developed can be implemented in real-time at the auroral all-sky stations. (C) 2001 COSPAR. Published by Elsevier Science Ltd. All rights reserved.