根据桥梁水体相连的知识,提出了一种基于知识的高分辨率遥感影像水上桥梁提取的实用方法,即在水体提取的基础上将图像细化,通过宽度限制检测桥梁主干线,并对其进行矢量化和特征表达来实现桥梁提取,通过实验证明了该方法的有效性。实验证明此方法不仅适合于水体与陆地灰度差别大,即单峰直方图的影像,还适合于水体与陆地灰度差别不大,即直方图多峰或峰顶不明显的影像。
According to the fact that bridge over the water is connected with water in remote sensing image, this paper proposes a practicable knowledge-based bridge extraction method from high resolution remote sensing image. The steps are: water extraction based on GMRF-SVM classification ; image thinning; main line of detection by width; vectorization; feature expression. In the end, this paper gives the test with IKONOS image. The experimental results show that the method presented is suitable for images both with single-peak histogram ( the contrast between water and land is great ) and with multi-peak histogram (the gray value of water is close to land).