一种基于投票法融合的ASTER遥感影像水体提取方法

遥感影像在水资源调查和洪涝灾害监测中发挥着重要作用,但从遥感影像中提取水体通常面临着阴影和狭小水体漏提等难题。针对单一方法在水体提取中的局限性,引入分类器集成的思想,提出一种基于投票法融合的水体提取方法,首先利用Bagging、Random Forests和神经网络(NN)分类器对遥感影像进行分类,然后采用多数投票法从决策层融合3个分类结果,得到研究区水体提取结果。试验结果表明,该方法能够有效去除阴影且能较好地识别狭小水体,具有良好的应用效果。

A Method of Water Extraction Based on Voting Method Fusion for ASTER Remote Sensing Image

Remote sensing imagery plays an important role in water resources surveying and floods monitoring.Extracting water from remote sensing images are often faced with some problems,such as the perturbation of shadow,the leakage of narrow water et al.According to the limitations of single approach,a thought of classifier integration was introduced and a novel water extraction algorithm is proposed based on voting.Firstly,Bagging,Random Forests and Neural Net classifiers were used to classify remote sensing images.Then,the majority voting was adapted to fuse the above three classification results.Finally,the final map of water was obtained from the decisive fusion result.The experimental results show that the proposed method can effectively remove the shadows and identify the small water well.