卫星图像传输跟踪优化识别方法仿真研究

Simulation Research on the Optimal Identification Method of Satellite Image Transmission and Tracking

对卫星图像传输进行跟踪识别方法优化,可有效提高卫星图像处理过程目标跟踪效果和准确性。对多特征目标进行跟踪,需要将不敏感特征与颜色特征进行融合,保证跟踪精度,完成目标跟踪。传统方法融合目标纹理特性、颜色特性、运动边缘特性等对目标进行表示,但忽略了对其进行迭代搜索,导致跟踪复杂度过高。提出改进的粒子滤波与均值漂移相结合的跟踪识别方法。通过粒子滤波获得目标状态的初始估计,然后利用卡尔曼滤波得到初始估计结果的最优估计,提高估计精度。在跟踪过程中利用Harris角点对目标旋转、光照及部分仿射变化不敏感等特性,与颜色特征进行融合来解决识别方法由于背景过于复杂而跟踪失败问题。然后将聚类中心作为代表粒子进行均值漂移迭代搜索,通过减少迭代的粒子降低识别方法的运算复杂度。试验结果表明,改进后的识别方法在鲁棒性和实时性方面都有所提高。

A method for tracking identification integrating modified particle filter with mean shift is put forward. Initial estimation of target state is obtained with particle filter. Then, Kalman filter is used to obtain the optimal result of initial estimation in order to improve estimation precision. During tracking process, Harris comer is used to fuse character of target rotation, sunshine, and partial insensitive affine transform with color feature to overcome tracking failure caused by too complicated background. Thus, the center of clustering is used as representative particle to carry out iterative search of mean shift, and the algorithm complexity of identification method is reduced through reducing iterative particle. Simulation results show that the method improves robustness and instantaneity.