中国余数定理(Chinese remainder theorem,CRT)方法是多基线相位解缠绕技术的一个基本方法,但是其较差的噪声鲁棒性问题限制了该方法在多基线相位解缠绕中的应用,然而基于聚类分析的多基线相位解缠绕技术能够克服传统的CRT算法噪声鲁棒性差的问题,本文根据CANOPUS算法中的聚类方法,提出用L∞-norm的距离测度定义两点之间的距离,从而减少特小类的产生和降低噪点数,进而提高聚类分析的精度,并且改进CANOPUS算法的算法流程以提高聚类分析算法的执行效率,进而大幅度降低聚类分析的运算时间。通过用仿真数据和实测数据验证可得,本文提出的改进聚类方法的聚类分析精度和算法执行效率更高,有效性通过实测数据实验得到了验证。
As a basic technique against the multi-baseline phase unwrapping problem, Chinese remainder theorem (CRT) restricts itself in the application of multi-baseline phase unwrapping due to its worse noise ro- bustness. However, multi-baseline phase unwrapping algorithms based on the cluster-analysis are able to over- come the drawbacks of the traditional CRT method. Based on the cluster analysis method in the CANOPUS al- gorithm, an L∞-norm is employed to define the distance between two elements to decrease the production of very small clusters and the number of noise points so as to improve the cluster-analysis (CA) performance. Besides, the procedure of the CA method in the CANOPUS algorithm is changed to improve the efficiency of the method and decrease the consuming time of the algorithm. According to the experiments on a simulated and repeat-pass real interferometric synthetic aperture radar (InSAR) dataset, the effectiveness and efficiency of the improved cluster-analysis method are tested.