摘要
在成像制导过程中需要实时处理大量的信息,为了尽可能在保留有效信息情况下降低计算量,提出了一种基于人眼非均匀采样特性和粒子滤波相结合的红外目标跟踪方法。首先应用对数极坐标模型的旋转、缩放及灰度分布不变性来压缩信息量以提高计算速度,然后采用基于目标灰度特征的粒子滤波跟踪方法来抑制匹配点漂移和目标的非刚性形变,从而实现对非刚性变形目标的稳健跟踪。与角点匹配跟踪方法及传统的Mean Shift跟踪方法相比,该方法具有跟踪稳定,计算速度快等优点。实验结果表明,该方法能够有效抑制匹配点漂移和目标的非刚性形变。
Large quantity information needs to be dealt during image guidance. In order to compress calculation quantity, a novel method for tracking infrared target based on particle filtering theory and human eye non-uniform sampling characteristic was proposed. First, Log-polar coordinate transform model was used to compress calculation quantity and increase calculating speed. And then, particle filtering tracking method based on target intensity feature was presented to suppress matching point drift and local deformation, thus achieved non-rigid deformed target tracking. Compared with comer matching method and traditional Mean Shift tracking method, this method is stable and efficient. Experimental results show the proposed method is effective for suppress matching point drift and non-rigid deformed.
出处
《宇航学报》
EI
CAS
CSCD
北大核心
2008年第5期1637-1642,1647,共7页
Journal of Astronautics
基金
航天支撑基金(N7CH0009)
航空支撑基金(05C53005)
西北工业大学研究生创新实验中心资助
关键词
非均匀采样
对数极坐标变换
非刚性形变
粒子滤波
匹配点漂移
Non-uniform sampling
Log-polar transform
Non-rigid deformed
Particle filtering
Matching point drift
