分形理论在遥感影像地物空间结构及其变化规律研究方面具有很大优势。本文使用三折线拟合法确定"无标度区间",改进传统的分线分维数计算方法,使分线法的计算结果具有实际意义并且更为准确。利用改进后的分线法对多源遥感影像进行分形特征研究,发现城区、田地、山体植被、水体等典型地物的分维特征在不同空间尺度遥感影像上比较相似,在同一尺度遥感影像上又具有比较明显的方向性特点。这些特征为更有效地进行遥感影像地物识别提供了重要参考。
Fractal theory has a great advantage in studying the spatial structure and its variation rule of the objects in remote sensing images.In this paper,the three-line fitting algorithm is utilized to determine the "no-scale part",and therefore the traditional line-divider method of fractal dimensions computation is improved,which makes the values of fractal dimensions of practical significance and more accurate.Based on the improved line-divider method,the fractal features of multi-source remote sensing images were studied.The results show that the fractal features of typical land covers,such as city zone,farmland,mountain plants and water,are quite similar in different spatial scales of remote sensing images,and have obvious directional characteristics in the same scale.These features provide more information for land covers identifying.