基于知识与规则的红树林遥感信息提取

Remote sensing information extraction of mangrove based on knowledge and rules

针对难以将红树林同陆地植被,尤其是同水体与陆地植被混合像元有效识别的现象,结合TM影像提取了能有效反映红树林湿地特征的绿度指数和湿度指数,同其他常用的NDVI、TM3/TM5、TM5/TM4等指数相比:绿度指数和湿度指数更能有效地提高红树林同陆地植被,尤其是同水体与植被混合像元的可分性.采用知识与规则方法提取红树林遥感信息,与其他学者常采用的分类特征及分类方法相比,识别精度有明显提高,Kappa系数提高0.10,错分率降低16.1个百分点.

The classification accuracy of mangrove is always low due to the similarity of spectra between mangrove and land vegetation, especially water-vegetation mixed pixels. Greenness index and wetness index were extracted based on TM imagery, which can effectively reflect the wetland characteristics of mangrove. The greenness index and wetness index can significantly improve the separability between mangrove and water-vegetation mixed pixels by comparison with NDVI, TM3/TMS, TMS/TM4, which always were employed by other researchers. Knowledge and rules method can significantly increase the classification accuracy of mangrove, compared with conventional classifi- cation features and method employed by other researchers. And the Kappa coefficient increased 0. 10 while commis- sion error of mangrove class decreased 16. 1 percent by using decision tree method.