天基光学传感器动态成像配准方法研究

Research on registration for dynamically spaced optical sensor imaging

图像配准是天基光学动态成像处理中的关键技术,在遥感侦察等领域有着重要应用。对星空背景和地球大气背景下天基目标探测中的图像配准问题进行研究。针对星图配准建立了部分平均Hausdorff距离(PMHD),并定义了一种新的相似性匹配测度,利用序贯岭回归估计方法实现快速配准;针对地球大气背景图像配准,通过改进的Harris角点检测亚像素定位提取特征点集,基于PMHD利用粒子群优化(PSO)算法实现图像的精确配准。仿真实验表明:提出的星图配准方法具有实时性和亚象元配准精度,联合PMHD和PSO的大气图像配准具有良好的配准精度和较强的抗噪干扰能力。

Image registration is the key for the dynamically spaced optical imaging processing,which is widely implemented such as in remote sensing reconnaissance.Research is done on the image registration for the spaced target detection under the background with stars and earth atmosphere.As to register the star images,partial mean Hausdorff distance（PMHD） is built and a similarity measurement is defined,which are followed by the sequential mountain regressive estimation to complete the registration quickly.For the earth atmosphere image registration,firstly the feature sets are extracted with the improved Harris corner detection and sub-pixel location.And then the PMHD and particles swarm optimization（PSO） algorithm are implemented to achieve precise registration.As the Monte Carlo experiments show that the method presented for the star images registration is real-time with sub-pixel precision and the method for atmosphere image registration is accommodative to noise with good registration precision as well.