Abstract: Forward looking infrared（FLIR）target recognition is an important factor that influences the hitting accuracy of precision guided weapons. A forward looking infrared target recognition algorithm based on edge matching is proposed for the decrease in mask matching performance due to the dissimilarity between base image and real-time image. For each candidate target chosen by mask matching, the neighborhood information is used to eliminate the false alarms, the matching confidence is computed by using a matching cost matrix, and the recognition results of continuous foregoing frames are utilized to provide information for current frame recognition, thereby improving the accuracy and robustness of target recognition. The experimental results of two FLIR videos of different target types show that the proposed algorithm has an excellent performance.

Key words: information processing, scene matching, edge matching, distance transform, matching confidence