【摘要】 为了稳定跟踪空中具有强机动特点的战斗机目标,提出了一种图像梯度特征相似性度量算法,并利用空中目标梯度空间分布特点构造了一种自适应调整模板尺寸的方法。在传统MAD相关匹配算法的基础上,通过引入梯度匹配权值模板,实现了对梯度特征的相似性度量。在跟踪过程中利用目标梯度空间分布局部占优的特点,构造能够反映目标尺寸的特征分布曲线来估计目标大小,调整匹配定位时的目标模板尺寸。仿真结果表明,跟踪算法能够适应飞机在做强机动时产生的快速形变,以及由形变带来的目标自身灰度的剧烈变化,实现了对目标的稳定跟踪。更多还原

【Abstract】 In order to achieve robust and efficient tracking of a battleplane with strong flexibility, a novel tracking method based on a new similarity measure of gradient features and adaptive sizing of target is proposed in this paper. The similarity measure of gradient features is based on Mean Absolute Difference (MAD) by the introduction of matching power template, which is obtained by the transformation of target template. Using the characteristic of local superiority of target gradient feature, the probability distributing of gradient transformation is used to estimate target size. And then adjusting the size of target template is performed efficiently according to the estimate result while tracking. The results of experiment over real-world sequences show that the proposed tracker is robust to the rapid changes in viewpoint, pose and scale by the battleplane while moving flexibly.更多还原