Abstract:

To deal with the conflict between realtime target tracking and lowpower hardware platform of television homing seeker, a binary feature descriptor using local intensity histogram is proposed, which is inspired by the offline training method. Then, a twostage feature matching scheme consisting of dissimilarity function and index value is adopted based on the proposed descriptor. The main advantage of this scheme is to convert timeconsuming feature matching to simple bitwise operations. Experiment results demonstrate that the proposed method achieves outstanding runtime performance at the speed of over 25 fps with only a little memory cost. With its robustness to scale changes, rotation and illumination changes, the method could well fulfill the realtime target recognition and tracking needs of homing seekers.