基于知识规则的数字扫描图像目标边界提取  

Edge Extraction in Scanned Map Based on Knowledge Rule Analysis

作　　者：谭衢霖[1] Aldred A 胡吉平[1] 

机构地区：[1]北京交通大学土木建筑工程学院,北京100044 [2]Department of Geography,the University of Western Ontario

出　　处：《应用基础与工程科学学报》2013年第3期553-561,共9页Journal of Basic Science and Engineering

基　　金：国家自然科学基金项目(51078020);中央高校基本科研业务费专项项目(2012JBM078)

摘　　要：为提高扫描图像数字化的效率,提出了一种基于知识规则的数字扫描图像目标区域边界提取方法.该方法通过对数字扫描图像进行分割和区域合并,将输入的扫描图像分割成多个待分类的图像区域;构建多层次知识规则,应用面向对象分析的分类方法,从形成的多个图像对象层中逐层分类识别,提取出目标对象区域;然后矢量化分类提取的目标对象轮廓,最终获得目标边界矢量图像.试验结果表明,该方法能有效地进行扫描图像边界数字化,不仅能提取复杂的目标区域边界,而且具有较高的自动化程度.A new method based on knowledge rules was developed in the paper to improve the efficiency of edge extraction for digitizing scanned map. Firstly, segmentation and region merging were performed to form image objects in the digitally scanned image. Secondly, multi-level knowledge rules were constructed. The object-oriented image analysis was applied to the classification of image objects at multi-layer image merging regions. The target-object areas were recognized layer by layer. Finally, the boundaries of the classified interest objects were extracted by vectorization and generalization. Experimental results showed that the proposed method was efficient in the edge digitalizing process. The method cannot only extract complex target area boundaries,and the workflow process of the same classification tasks can be automated to a large degree.

关 键 词：扫描地图 数字化 图像分割 知识规则 边界提取 

分 类 号：TN215[电子电信—物理电子学] TP391[自动化与计算机技术—计算机应用技术]