Abstract: The authors present an effective approach to extracting coastline from TM imagery, integrating texture and spatial relationship of imagery. Firstly, the image area is classified into two main categories (one is water, the other is land), exploiting the texture and MNDWI, which are extracted from the input image. Secondly, the segmentation above is dealt with locally using mathematical morphology and the small heterogeneous regions in water and land are filled using the region growing method, relying on the spatial relationships and expert knowledge about the coastline. Finally, vectorizing the segmented image generates the final coastline, and the coastline is assessed by the proposed evaluation system, which is found on the matching of two lines. The experimental results demonstrate that the proposed method is able to accurately extract coastlines from TM image of Liaodong Peninsula. Moreover, texture and spatial relationships are useful information for coastline extraction.

Key words: texture, spatial relationship, coastline, automatic extraction, lines matching, evaluation

摘要： 针对以往研究中简单运用影像单一特征(光谱特征或者空间关系)提取海岸线的不足, 在提取过程中逐步融入区域纹理特征、空间关系和经验知识, 提出一种融合纹理特征和空间关系的海岸线自动提取方法: 首先利用影像的水体指数和纹理特征, 对影像水、陆区域进行二值分割; 然后结合空间关系和经验知识, 对分割结果进行优化和后处理, 并生成矢量海岸线; 最后, 利用基于线匹配的岸线精度评价准则, 分析岸线提取结果。将该方法运用到辽东半岛复杂背景条件下TM影像的海岸线提取实验中, 取得了良好的实验效果, 证明纹理和空间关系在海岸线提取中具有重要作用。

关键词: 纹理特征, 空间关系, 海岸线, 自动提取, 线匹配, 精度评价