中文摘要：

研究单幅图像与2D地理空间数据的融合方法，可推动其在GIS中的深入应用。在阐述单幅图像空间化概念的基础上，提出了一种将单幅图像中的地平面映射至2D地图的空间化方法。首先，采集大量的单幅图像并基于决策树进行训练，分为地面和非地面两类，建立分类器；然后，基于近似水平的约束，选择摄像机中心坐标及方位角，实现图像中地面区域的坐标到地理空间坐标的转换。以大学校园拍摄的大量单幅图像为例，对该方法进行了验证，并分析了地平面空间化的精度和敏感性。结果表明，该方法对于主光轴近似水平或灭线位置估计较准确的图像的空间化效果较好，可为单幅图像在GIS中深入应用提供借鉴。

英文摘要：

Single image geo-spatial mapping is very important and the research can promote the thorough application of the single image in GIS. Firstly, the concept of single image geo-spatial mapping is briefly introduced. Then, a method is proposed. Its main steps are as follows： A large quantity of images should be collected and the principal optic axis is approximately horizontal. Then each image is divided into many super-pixels, and lots of features are selected such as color, texture and location. All super-pixels are calculated one by one and the classes include ground and no ground. Through training, a decision tree can be obtained. On the basis of the horizontal constraint as well as the center coordinate of the camera and azimuth, the image coordinate of the ground area can be transferred into the projected coordinate. Through shooting many images on a university campus, the method is verified. The mapping accuracy is analyzed. The result is good. The study is useful for single images application in GIS.