基于多视InSAR相干性估计误差模型的相干性迭代估计

文章全面分析了引起InSAR多视相干性估计误差的因素,建立了与区域增长相干性估计方法对应的多视InSAR相干性估计误差模型,提出了一种新的基于该误差模型的相干性迭代估计方法.该方法首先利用基于强度图像的区域长原理计算相干性,区域增长保证窗里的所有像素属于同一分布,从而能够消除不同分布样本导致的相干性估计误差;后利用所建立的多视相干性与真实相干性的非线性模型对所得的相干性估计值进行高斯牛顿迭代,迭代可以减少由于相性太低和估计样本太少导致的相干性估计误差,得到更加准确的相干性估计.

The Iterative Estimation of the Coherence Based on the Multi-Look InSAR Coherence Estimation Error Modeling

The paper roundly analyzes the biases of space multi-look InSAR coherence estimation and constitutes the Multi-Look InSAR Error model of the Coherence estimation estimated by Region Growing Theory. Based on this InSAR Error model, the paper proposes a highly accurate iterative estimating method. Firstly, it chooses the moving window by region growing theory which ensures pixels in the window belong to the same distribution, then the bias resulting from the pixels belonging to different distributions is decreased. Secondly the bias resulting from too low coherence and too small number of pixels is decreased by using Gauss-Newton Iteration algorithm. The Iteration algorithm bases on the equation between the. multi-look coherence and the real coherence and the equation is educed from the complex Hermit product Speckle Noise Modeling. Experiment results validate the validity of this algorithm.