遥感图像建筑物识别及变化检测方法

摘要

For the low accuracy when only using pixel-level or feature-level change detection for high-rise building ,a recog-nition and change detection method combined image features and experience knowledge is presented to detect changes of high-rise building in multi-temporal remote sensing images .We adopted a proposed registration algorithm combined with ratio gradient and crossing accumulative residual entropy to register SAR and panchromatic images with two different phases ,identified building re-gions respectively using knowledge rules and utilized pixel ratio method for building change detection in recognition building re -gions .Experimental results show the method can effectively improve the accuracy and reduce the false acceptance rate and reject rate .%针对单独使用像素级变化检测或特征级变化检测对于高层建筑物检测精度低的问题，提出了一种图像特征和经验知识结合的建筑物识别及变化检测方法，用于检测多时相遥感图像中高层建筑物的变化情况。首先采用本文提出的 Ratio 梯度与交叉累积剩余熵相结合的配准算法配准两个不同时相的 SAR 和全色图像，分别利用知识规则识别 SAR 和全色图像建筑物区域，在识别的建筑物区域上，采用像素比值法进行建筑物变化检测。实验结果表明，该方法可以有效提高建筑物的检测正确率，降低虚检率和漏检率。