利用面向对象分类方法提取冬小麦种植面积的研究

应用陆地卫星TM数据和遥感图像处理软件eCognition5.0和ENVI4.3软件,以面向对象的方法和监督分类波谱角法分别提取泰安市2005年冬小麦种植面积及其分布信息。逐像素分类的结果存在"椒盐"效应,而且很难克服同物异谱、同谱异物现象,面向对象的分类方法可以有效的集成专家知识和各种辅助数据,克服逐像素分类的弊端。分类结果表明,利用面向对象的分类方法可以获得比传统的像素级分类方法更高的分类精度,为冬小麦种植面积的自动提取提供了广阔的前景。

Extraction of winter wheat planting area by object-oriented classification method

With digital satellite remote sensing TM data and image processing software eCogntion5.0 and ENVI4.3, the winter wheat planting area and distribution information is extracted based on object-oriented classification method and Spectral Angel Mapper respectively in 2005 in Tai’an City. The result of per-pixel classification has the issues such as "salt and pepper", and can’t overcome the phenomena of same feature with different spectral character and the same spectral character of different features; object-oriented classification can integrate expert knowledge and auxiliary data and overcome the issues of per-pixel classification. The result of this paper shows, compared with traditional per-pixel classification, object-oriented classification can obviously increase the accuracy of classification. Object-oriented classification provides a prosperous future in the field of automatic satellite image interpretation.