主流城市不透水面信息提取方法评价及分析

Evaluation and Analysis Mainstream Methods of Urban Impermeable Surface Information Extraction

选取库尔勒城区为研究区,以2014年RapidEye为数据源,在基本数据处理的基础上,采用主流的不透水面提取方法(线性光谱混合模型:LMSA;主成分分析方法:PCA、NDVI二元法)提取2014年库尔勒市城区的不透水面信息。然后整理地理国情普查成果指标,转换为实际的不透水面信息作为检验样本,评价3种方法的提取精度,确定适宜新疆典型城市的不透水面提取方法。结果表明:LSMA总体精度、Kappa系数及提取效果最佳,NDVI二元法次之,PCA最低。LSMA法虽然局部识别效果较破粹、小图斑居多,但能提取出PCA及NDVI二元法敏感性较差的地块,推广性较强。

Based on the year of 2014 Rapid Eye imagesand mainstream impervious surface extraction methods, this paper aimed at extract the impervious surface information of the study area, Korla. First, use LSMA,PCA,NDVI binary method to extract information ofimpervious surfacein Korla in 2014, and then reorganize information of impervious surface as an actual sample which is based on theresults of geographical census to evaluate the extraction accuracy and confirm the appropriate method to extract impervious surface intypical Xinjiang cities. Result shows that： LSMA’ s precision, Kappa coefficient and the extraction effect are the best,NDVI binarymethod as follow,PCA is worst. From the recognition effect, although LSMA has more broken local recognition effect and small plots,this method can extract less sensitive plots where PCA and NDVI binary method less sensitive. So, the promotion is stronger.