多点最小二乘匹配的可变权阵阵列松弛算法

MULTI-POINT LEAST SQUARES MATCHING WITHARRAY RELAXATION UNDER VARIABLE WEIGHTS

本文详细介绍了多点最小二乘匹配的常规算法和由阵列代数表示的阵列松弛算法，并把Rauhala限于等权的情形推广到不等权（变权）的情形，给出了实现阵列松弛的步骤，最后从计算效率、匹配精度对常规算法，等权阵列松弛和不等权阵列松弛进行了比较、分析，揭示了变权阵列松弛法巨大的实用潜力。

:The algorithm of Multi-point Least Squares Matching with Array Relaxation by Rauhala has been discussed in detail, and the new formulation of array relaxation under variable weights was presented. Array relaxation was also developed into the case of first ordeV derivative constraints. The realizable com puted steps were drawn out. The new algorithm was investigated using the ()(上接177页)loess plateau aerial images, SPOT images and ISPRS Working Group Ⅲ/ 4 Test A('WALL'). The matching results of the conventional algorithm, the array relaxation unde'r optimal weights and the array relaxation under variable weights were compared and analysed. The solution speed of the array relaxa tion algorithm of image matching was found to be near one hundrad of nodes per second in our software implementation. It has shown, that the method of array relaxation under variable weights has great potential for multipoint least squares matching.