面向对象的池塘养殖用海信息提取

摘要

针对SPOT5卫星遥感数据,以面向对象的图像分析理论为基础,通过多尺度图像分割,获取不同空间尺度结构下的海域使用地物斑块,并综合分割对象的光谱、形状和语义特征,建立分类规则集,实现池塘养殖用海信息提取.结果显示,分类精度优于94％,表明对于地物混杂度较大的海岸带地区,采用面向对象的图像分析技术能有效实现较高精度的池塘养殖用海信息提取,在海域遥感监测领域中具有较好的应用前景.%Based on the theory of object - oriented image analysis, this paper carried out a multi - scale segmentation of SPOT5 image to obtain sea use objects at different spatial scales. By integrating objects' spectral, shape and context features to formulate detailed procedures of information extraction, areas of pond aquaculture were identified with classification accuracy better than 94% , indicating that the method of object - oriented image analysis is effective in the extraction of pond aquaculture information in coastal areas. It would promote the operational monitoring and management of sea use.