为解决姿态剧烈变化的火箭目标跟踪问题, 提出了一种新的基于纹理特征的跟踪方法。相对于传统的模板匹配方法, 这种方法能对助推器和子级分离的捆绑式运载火箭进行跟踪。在对火箭目标特性进行分析的基础上, 采用局部二值化模式(LBP)纹理特征结合Hu不变矩特征描述进行跟踪。跟踪过程中特征匹配采用简单的相似性度量准则, 实时模板更新算法为加权更新。测试结果表明, 这种方法对发生位移、旋转和尺度变化的火箭目标同样具有比较好的跟踪效果。

A new texture feature-based tracking algorithm is proposed for tracking rocket targets with attitudes changing violently. Compared with the conventional algorithm of template matching, the new algorithm can track strap-on launch vehicles with separation of boosters and stage engines. Based on the characteristics analysis of rocket targets, local binary pattern (LBP) texture feature is combined with feature description of Hu moment invariants to make target tracking. A simple similarity measurement criterion for template matching in tracking and a weighted updating algorithm for updating reference template are adopted. Experimental results certificate the new algorithm can also make excellent tracking on rocket targets under geometrical changes, such as displacement, rotation and scale variation.E-mail: