基于模糊集的土地利用变化自动发现方法探讨

传统的逐像素比较法忽略了地理对象的模糊性,缺乏区分噪声与真实变化信息的能力,难于处理空间数据的复杂性和不确定性。基于模糊集理论,根据邻域像元的空间位置关系确定变化的程度,并利用滑窗和景观指数表达比较不同时相遥感数据(TM、ETMI、KONOS)的分类专题图。试验结果表明,模糊集的描述更接近真实情况。

Land Use/Cover Change Auto-Detection Method Based on Fuzzy Set

The conventional pixel-by-pixel post-classification comparison method has some limitations while handling geographic change comparison,which often leads to fake change detection results.This paper presents an improved post-classification comparison method based on fuzzy set theory with a case study of Wuhan.During the comparison,two sources of fuzziness are considered,I-fuzziness due to vague distinctions between categories,II-fuzziness due to a gliding scale of severity of spatial error.Instead of one single category or value per pixel,each pixel is characterized by a membership vector.Each element in the vector declares,with a value between 0 and 1,the degree of membership for one category.Pixels within a certain distance(the neighborhood) of a central pixel influence the fuzzy representation of that pixel by a distance-decay membership function(exponential).In this way,thematic and geometric aspects of uncertainty are treated separately.Two kinds of comparison results,using this method and the conventional method separately,are compared.Groups of two thematic classification maps of different time periods are compared,using sliding windows.The change information is analyzed by the method of landscape index.The results show that the improved method is more capable of capturing both the complexity and the patterned quality of spatial changes of land use,it can get more reliable results and is more feasible in practical use.But some defects of post-classification method still remain,such as low accuracy and being insensitive to small changes,etc.So,in further work,the ability to synthesize the results from different change-detection methods needs improving.