基于分形纹理的遥感图像岩性识别方法

Recognition of Lithology with Remote Sensing Image Based on Fractal Texture

根据分形理论和遥感影像的特征 ,本文提出一种图像分形纹理计算方法。分形纹理能够准确刻画图像的空间特征信息 ,在遥感图像分类中作为空间特征信息的补充。利用该方法在河西走廊酒泉盆地的红柳峡地区进行了TM图像的岩性分类实验 ,当只采用原始TM六个波段图像分类时 ,精度仅为 71 .6 2 % ,而采用分类纹理信息时分类精度则提高到 79.5 4 % ,说明分形纹理对岩性空间特征信息有较好的描述 ,将分形纹理信息配合原始波段进行岩性识别时 ,总体识别精度达到 87.93% ,提高了 1 6 .3个百分点。因此 。

The image classification is the main theme of Remote Sensing. By analyzing the characteristic of remote sensing image, this paper presents a classification method based on fractal theory. The fractal texture can describe the space characteristic information of image correctly. The results of experiments in Hongliuxia area indicate that the presented method can enhance the classification precision remarkably. With the fractal texture only, the precision increases from 71.62% to 79.54%, with the fractal texture and the TM image, the precision increases to 87.93%.