面向对象的海岸带养殖水域提取

在海岸带遥感地物自动信息提取工作中，传统基于像元的图像处理技术难以解决由于地物混杂带来的椒盐噪声现象。提出了一种面向对象的养殖水域提取方法，该方法分为4步：① 结合eCognition分割算法的基本思想，设计并使用了多精度图像分割算法获得分割图斑；② 计算图斑的多类特征，包括光谱、形状、纹理、空间关系等，为后续分析服务；③ 通过基于图斑的多特征分析，提取出水域全图以及其派生的面状水域、线状水域；④ 在所有水域中剔除面状水域和线状水域，获得养殖水域。在SPOT\|5高分辨率遥感影像上的提取实验证明该方法是可行的。

An Object-oriented Approach for Extracting |Farm Waters within Coastal Belts

A feasible object-oriented farm water extraction method is proposed.This approach includes the following steps.First,a multi-resolution image segmentation method combining spectral and shape features is implemented to extract image parcels.Second,a set of parcel features including spectral,shape,spatial relationships are obtained for the succeeding analyses.Third,the whole water areas and then the polygonal water areas and linear water areas are extracted with multi-feature analyses.In the last step,the farm waters are extracted by eliminating polygonal water area and linear water areas.Our method is validated with experimental analyses.