基于相位一致梯度算法的无人机影像分类

UAV Images Classification Based on Phase Congruency Gradient Algorithm

传统的分水岭分割算法会产生过度分割的现象,以浙江省玉环县海山乡的无人机影像为例,利用相位一致梯度算法对分水岭算法进行了改进实验,并将分割结果与传统ESP算法进行比较,得到了较好的效果。对分割后的图像进行了面向对象的信息提取,实验区影像土地利用类型分类的总体精度达到90.11%,Kappa系数为0.88,说明分类的总体精度较好。

The traditional watershed segmentation algorithm will produce“excessive segmentation”phenomenon.On the basis of this,taking the unmanned aerial vehicle(UAV)images of Haishan Township in Yuhuan County,Zhejiang Province for example,we used the phase congruency gradient algorithm to improve the watershed algorithm.Compared with the traditional ESP algorithm,this algorithm was better.And then,we classified the segmented images by objectoriented information extraction.The classification overall accuracy of the land use types in experimentational area images reached 90.11%,and the Kappa coefficient was 0.88,which indicating that the overall accuracy of the classification was good.