基于特征线匹配的城市建筑物识别方法研究

随着手机GPS位置测定、导航以及摄像功能的普及，对于移动状态下的位置定位以及都市空间建筑物实时搜索等应用功能存在日益增长的需求。介绍了图像处理方式的城市建筑物识别方法，通过对两种图像匹配算法--SIFT和Local Search的比较分析表明，构造物轮廓线及其组合是一种相对稳定的几何特征，在匹配时受图像的仿射变换、画质变化以及镜头畸变等因素干扰较小，此外，基于轮廓的图像匹配指数还能反映摄影位置的变化。因此，对Local Search进行了改进,利用实时图的建筑物的特征轮廓线与从三维数据库中提取的建筑群特征轮廓线进行匹配，然后选择匹配指数高的记录作为识别结果。结果表明基于Local Search算法的建筑物识别技术可以很好地适应移动条件下建筑物快速识别的要求。

Research of Urban Building Recognition Method based on Line Features Matching

With the popularity of cell phones having functions of GPS positioning,navigation and camera,there exist increasing demands for real-time identification of urban buildings with mobile terminals.In this paper,image processing algorithms for urban buildings recognition,including SIFT and Local Search,were discussed and compared in details to show that contour lines and their combinations are stable geometric features of buildings,so the algorithms based on contour lines immune to affine transformation,low picture qualities and lens distortion.Besides,the match index based on contour lines can reflect the camera position change.As a result,Local Search was improved;firstly getting the building contour lines by abstracted from pictures photographed by mobile terminals and generated from the 3D-database,secondly utilizing it to find some of the best matches between  those contour lines.The improved algorithm can meet the demand of fast building recognition in mobile terminals .