耕地信息是重要的农业信息,是分析土地资源利用程度的基础。论文通过面向对象的方法提取了泰国的水稻种植区耕地信息。研究中以TM影像作为数据源,首先考虑多光谱特征、紧密度和光滑度等几何特征,通过区域合并方法进行影像的多尺度分割,生成同质的影像对象多边形。然后选取影像中样地的光谱标准差、形状指数、对称度和密度作为耕地类别的识别特征,并采用模糊函数方法对各特征进行了定义。最后利用最相近匹配方法,对每个对象多边形进行类别判别。在分类的基础上,进行同类别的合并和统计,得到耕地的分布和比例信息。通过野外实测样地和目视解译结果检验,耕地类别符合率为90.25%,面积相同率为90.25%,形状一致性为90%,最终土地利用程度分析结果精度为90%。

Ploughed field information is an important information for farming,and serves as the foundation for studying land use and land cover change.In this study,the Landsat TM image is data resource while the geometric features including characteristics of spectrums,compactness and smoothness are taken into account first.The use of regional growing method allows the multi-resolution segmentation of an image into highly homogeneous image objects polygon.Based on ground observed information,standard deviation of spectrums,shape index,density and asymmetric of objects polygon are chosen as identification characteristics and fuzzy function is used to define the types of ploughed field.In multidimensional feature space,we adopt nearest neighbor method to classify every object polygon.Based on classification,the same class of adjacency is combined,and then the area and the proportion of the ploughed field are calculated.The object-oriented method is adopted to extract boundary of Thailand farmland.The field examination of ground observation and analytical result indicate that class match rate is90.25%,area equal rate,90.25%and shape consistency,90%.