大比例地形图中房屋目标的自动识别

AUTOMATIC RECOGNITION OF HOUSES IN LARGE-SCALED TOPOGRAPHICAL MAPS

城市GIS发展迅速，各种大比例地形图是其基础数据．流行的地图数据数字化方法效率很低．一定程度上已经成为城市GIS发展的障碍．扫描输入技术有望解决这个问题．扫描输入技术的核心是地图符号的自动识别．而房屋目标在城镇地区大比例尺地形图中占有极大的比重，但现有大多数识别方法采用的都是先分割后识别的策略，使得分割难以充分利用目标符号的有关知识约束．本文提出并实现了大比例尺地形图中房屋目标自动识别的分割识别一体化方法．

Inputing various current maps into GIS database by scanning is a hopeful approach to solve the problem of automatic GIS data acquisition. And the kernel of this method is automatic recognition of object symbols in maps. Houses are the main objects in urban topographical maps and the 1: 10000 scaled topographical maps are the most common used data in rapidly developing Urban GIS. But most of present methods for map recognition are based on the policy of first-segmentation-then-recognition,as a result,segmentation could not fully use the knowledge about objects. In this paper,a method for recognizing houses objects in 1: 10000 scaled topographical maps based on the integration of segmentation and recognition is proposed.