Small Infrared Targets Recognition with New Evidence Combination Approach

There are two practical problems in image fusion by using Dempster-Shafer (DS) combination rule: one is the classical DS rule involves counter-intuitive behaviors when deal with high conflict information, another is that it is difficult to determine the basic mass function of pixels. A method of image fusion with modified DS evidence theory is proposed to recognize small infrared targets in image sequences. Based on evidence distance, some inherent characters of evidences are extracted. The Basic Probability Assignment (BPA) of pixels can be obtained from the characters of infrared images. According to the experimental results of using proposed method to an infrared image sequence with targets, the new proposed method is more effective and reliable with good convergence efficiency and high detect-rate.