基于混合像元分解的辽河口滨海湿地分类

摘要

滨海湿地高精度的地物分类可以为湿地监测与保护提供数据支持和决策依据.以辽河口湿地为研究对象,以Landsat8 OLI多光谱影像为数据源,结合研究区域实际地物情况,采用像元纯度指数和均值波谱法确定端元光谱,并利用全约束最小二乘混合像元技术和决策树技术制定分类规则,最后将研究区域分为芦苇、翅碱蓬、水稻、滩涂、水体(海水、虾池水、河水等)和人工建筑(包括路面、人工设施、房屋等)六大类.结果表明:该算法分类精度高于90%,结合目视判读与野外实地调查,发现分类结果符合实际地物情况.%To classify the Liaohe Estuary, a new classification strategy was prosposed by the fully constrained least squares (FCLS) unmixing model and decision tree for the Landsat 8 OLI data. The accuracy assessment which adopts traditional error matrices with man-made region of areas is higher than 90%. Meanwhile, comparison between the traditional supervised and the unsuperviesed classificational algorithms (maximum likelihood method and ISODATA algorithm) proved that the new classification scheme can get more accurate estimates of coastal wetland landscape classification. The classification can provide data and decision support for the monitoring and protection of the regional feature.