基于遥感影像水体信息提取的改进方法 Modified Water Information Extraction Method Based on Remote Sensing Images

本文分析了单独使用归一化差分植被指数法和归一化差分水指数法所引起的误提取和漏提取，总结了导致错误的原因及规律。耦合了归一化差分植被指数法、归一化差分水指数法和数学形态学理论，并提出了一种水域信息提取的改进方法。以鄱阳湖水域为研究对象，选取2010年3月至2011年12月的HJ-1A时序卫星影像图进行水域信息提取，并采用混淆矩阵对提取结果进行了比较和评价。结果表明：改进的方法比三种常规方法(NDVI，NDWI，监督分类)的水体提取精度平均提高了9.43%。 By analyzing the error extracted areas which is caused by using NDWI and NDVI methods indi-vidually, a modified water information extraction method was proposed and tested. The method interconnects the NDWI, NDVI and mathematical morphology methods together. The HJ-1A sequence statelier images in the Poyang Lake during 2010 to 2011 were selected as case study. The water information were recognized and extracted by the modified method and compared with three conventional methods based on confusion matrix. The results show that modified method is more accurate than three conventional methods by improv-ing 9.43% on average.