Abstract: A new algorithm of urban area automatic recognition is proposed based on recognition idea of step extract from coarseness to fineness，aimed at the characteristic of visible remote sensing images and complexity of urban physiognomy.The algorithm builds hierarchical recognition model including classifying，elementary recognition，boundary extracting accurately.It first uses fuzzy classification，then elementary recognition is realized by voting-based method with multi-features，at last utilizes self-adaptively searching urban area boundary.The experiment calculates very fast and the boundary is accurate with high speed，robustness and high correct rate comparing with traditional recognition approach based on single-source visible remote sensing image.

Key words: multi-features criterion voting, fuzzy classification, urban area recognition, remote sensing images

摘要： 针对可见光遥感图像的特点和城市区域地貌的复杂性，提出了一种由粗到细、逐步求精自动识别城区的新算法。该算法建立了预分类、初识别、边界精确定位分层结构模型，首先利用模糊分类技术对城区进行预分类，在此基础上通过多特征判据投票实现城区初识别，最后采用自适应边缘搜索细化算法精确定位城区边界。与传统的单源可见光遥感图像城区识别方法相比，该算法能更快速、准确地自动识别城区，具有计算量小、鲁棒性强、识别正确率高的特点。

关键词: 多特征判据投票, 模糊分类, 城区识别, 遥感图像