Abstract: The main change detection algorithms include forming likelihood ratio algorithm and normalized image subtraction algorithm in FOPEN target detection using UWB SAR. There are respective problems for two change detection algorithms:The performance of normalized image subtraction algorithm will greatly degrade， when the correlation of sar images is reduced， and forming likelihood ratio algorithm needs enough prior knowledge. In this paper， a two stage change detection algorithm is proposed.In first stage， a change test statistic is computed based on modified LMSE estimation. In second stage， a improved rank order filter is introduced. It is used to overcome the positive tail of clutter pdf without destroying the edge information of target， which improves the performance of detection. Finally， the experimental results show that an obvious performance improvement on detect can be obtained through the new algorithm.