GF-KCF: Aerial infrared target tracking algorithm based on kernel correlation filters under complex interference environment

Abstract

Aerial infrared target automatic tracking technology is the core technology of the photoelectric coun-termeasure system of infrared imaging missiles. The correlation filter (CF) tracker has attracted considerable attention in the field of aerial target tracking owing to its excellent performance in tracking accuracy and timeliness. We propose a correlation filter tracking based on Gabor filter (GF) features (GF-KCF) in the frequency domain for aerial infrared target tracking. This tracker constructs a set of frequency-domain GFs, which effectively suppress background noise and highlight target texture information. In view of the large overload maneuver and rapid deformation of aerial infrared targets in the process from near to far distance, we propose a scale estimation method in the frequency-domain based on GF features. The eigenvectors of the target spectrum scale are extracted to improve the accuracy of target scale information estimation. To address the problem of partial occlusion of targets by infrared decoy flares, we propose an anti-interference tracking strategy based on a high-confidence patch model. High-confidence patches are used to track the reliable part of the target, which provide more target information and improve the robustness of the algorithm in case of reappearance of the occluded target. Compared with other trackers, the average accuracy of the proposed tracker was improved by 15.2%, and the frame frequency reached above 110 Hz.