基于SUKF与SIFT特征的红外目标跟踪算法研究

针对复杂红外背景下单一跟踪算法难以准确定位运动目标的问题,提出了基于尺度无迹卡尔曼滤波(SUKF,scale unscented Kalman filter)与尺度不变特征变换(SIFT,scale invariant featuretransform)相结合的红外运动目标跟踪方法。首先,通过SUKF算法对状态空间进行滤波估计,确定运动目标的初步位置,并以此建立局部SIFT特征检测域。其次,SIFT算法在该局部检测域内对运动目标进行特征提取与匹配,最终实现对目标的准确定位;同时,利用定位结果更新并校正SUKF的状态模型。实验结果表明,本文提出的基于SUKF-SIFT的跟踪策略与相关算法相比,体现出较好的跟踪效果与实时性能。

Research on infrared object tracking based on SUKF and SIFT

Aiming at the problems that accurate motion tracking of infrared object is hard in complex background by using a single algorithm,a new infrared object tracking method which combines scale unscented Kalman filter(SUKF) and scale invariant feature transform(SIFT) is proposed.Firstly,SUKF can estimate the state of unknown system,and the position of object center is located,and then the local detection area with SIFT feature is identified.Secondly,the features of object are obtained and matched by SIFT algorithm in this local area.Finally,the object center is accurately located by SIFT algorithm,and the state model of SUKF is updated and corrected by the final location.Compared with the related methods,the proposed tracking strategy based on SUKF-SIFT has better tracking results and real-time performance.