This paper proposed an approach to building extraction from aerial image using Bayesian nets. Firstly,it represented texton dictionary by training aerial image sets. Secondly,acquired classification of a new image by mapping the textons to texton dictionary and obtained the scene categories in the whole image. Finally, it selected naive Bayesian nets to represent spatial constraints between objects and scene classes, and computed the posterior probability of category node to building extraction. The test on the dataset shows that the proposed approach yields substantial improvement over others on building extraction from the aerial image.%提出一种基于贝叶斯网的建筑目标提取算法.该算法通过多场景航拍图像进行训练后建立纹元字典,将实际图像中的纹元映射到纹元字典获得图像的场景类信息；然后使用朴素贝叶斯网建模建筑目标与场景类空间上下文的关系约束,将建筑目标提取转换为求解贝叶斯网类别节点的后验概率问题.与同类方法的对比实验表明,提出的算法能有效地提取航拍图像中的建筑目标.
展开▼
