基于SPOT-5图像的城市水体自动提取模型研究

A MODEL FOR AUTOMATIC EXTRACTION OF WATER INFORMATION FORM THE CITY ZONE BASED ON SPOT-5 IMAGE

以南京市主城区为对象,研究基于SPOT-5图像的城市地区水体信息的提取方法。分析表明,对SPOT-5的SW IR波段进行简单的阈值处理,可以清晰地将水体与阴影提取出来。在分析水体和阴影光谱特征和空间特征(形状指数等)的基础上,建立了基于SPOT-5图像的城市地区水体的自动提取决策树模型。精度验证表明,该模型的水体提取精度较监督分类的提取精度提高2.5%,尤其在具有许多建筑物阴影的局部区域,本模型的水体提取精度提高11.6%。此外,本模型还具有很好的移植性,只是在阈值大小的确定上会有部分差异。

With the main city zone as the research area, the authors studied the method for extracting water information in SPOT- 5 image. The water and shadow information was obtained from the image by setting the division line in band SWIR. By using spectral, spatial and temporal characteristics, such as the index of shape, the deci- sion tree model for automatic extraction of water information in city zones from SPOT - 5 image can be set up , so as to get the water information from the area. For estimating the precision of the model, a comparison was made between the model and the supervised classification method in the whole area and in a special zone with much building shadow. The results show that in the whole area the precision of the model is about 2.5% higher than that of the supervised classification method, and in the special zone the precision improvement reaches 11.6%. Besides, the model has good transplanting function, with only slight differences in the division line in different cases.