摘要
该文提出一种高分辨率遥感图像目标自动提取方法,该方法首先使用分类器实现目标的快速检测,然后利用图像色彩模型和平滑性先验知识建立分割代价函数,并最小化此代价函数实现目标的精确提取,最后在后处理步骤中加入目标的形状先验知识,进一步提高精度。以油罐提取为例进行了实验,结果证明了该方法的有效性和鲁棒性。
In this paper, a method for automatic object extraction in high-resolution remote sensing image is proposed. First, a robust multilayer classifier is employed to detect the object efficiently. Secondly, a cost function based on the color model and the smoothness prior knowledge is built up and minimized to segment the object accurately. Lastly, in the post processing stage, the shape prior knowledge of the object is utilized to eliminate the false positives and improve the extraction precision. As an example of objects in remote sensing images the oil tanks are extracted. Experimental results demonstrate the robustness and effectiveness of the proposed automatic object extraction method.
出处
《电子与信息学报》
EI
CSCD
北大核心
2008年第11期2732-2736,共5页
Journal of Electronics & Information Technology
