基于区域增长的海岸带黑暗像元自动提取算法研究

The algorithm to extract automatically dark objects in coastal zone based on region growing

针对海岸带区域独特的地表状况设计了一个近海岸的黑暗像元自动提取算法。首先利用修订型归一化水体指数(RNDWI)和归一化植被指数(NDVI)对黑暗像元备选区域进行确定,然后从备选区域中找到初始黑暗像元,最后利用多阈值的区域增长法对黑暗像元的选取进行优化,最终计算得到黑暗像元值。应用此算法对Landsat-5 TM影像遥感数据进行处理,结果表明,大气校正之后的典型地物光谱曲线更接近于真实的光谱曲线,说明此算法设计合理,其大气校正结果与用ENVI软件进行FLAASH大气校正和黑暗像元法大气校正的结果相比,此算法更接近于FLAASH模型法,校正效果较好。同时利用黑暗像元自动提取算法得到大气校正后的结果大大减少了反射率值为负的情况,提高了数据的利用率。

The coastal surface features are considered and an algorithm to extract automatically dark objects in coastal zone is developed. Firstly, the optional dark pixel areas are determined by the revised normalized different water index （RNDWI） and normalized difference vegetation index （NDVI）. Secondly, the initial dark pixels from the optional areas are identified. Finally, optimum selection of dark pixels is achieved by the multi-threshold re- gion growing algorithm and the values for dark pixels are calculated. The algorithm is applied with remote sensor data from Landsat-5 TM. The result shows that the typical ground objects＇ spectral curves with atmospheric correc- tions can effectively reveal real ground objects＇ spectral characteristics, and that the algorithm is acceptable. Com- pared with other atmospheric corrections of dark pixels, this algorithm is much closer to the FLAASH model method and correction is much acceptable. Further, the atmospheric correction by this algorithm reduces largely negative reflectance and improves data utilization.