基于离散曲率的扫描线条图快速圆弧检测  

A fast arc detection method for scanned line-drawing recognition based on discrete curvature

作　　者：王飞[1] 赵季中[2] 何永健[1] 

机构地区：[1]西安交通大学人工智能与机器人研究所,西安710049 [2]西安交通大学计算机系,西安710049

出　　处：《中国科技论文在线》2008年第2期108-113,共6页

基　　金：国家自然科学基金资助(60021302);国家科技支撑重点项目资助(2006BAK31B04)

摘　　要：本文提出一种从圆弧假设到圆弧验证模式的扫描线条图的圆弧检测方法。该方法首先提取扫描线条图的图像骨骼,使用分段线性多边形对线条图像骨骼进行近似表示,来达到简化计算和减少数据量的目的;再通过对骨骼图像的局部离散曲率的计算和统计来假设图像局部存在圆弧;最后根据计算出的圆弧的圆参数将假设的圆弧映射至原始图像进行比对来验证局部圆弧的存在。该方法可以有效地克服圆弧断裂、重叠、交叉等噪声的影响,具有较好的鲁棒性和计算快速的优点。This paper presents a fast arc detection method which is used for scanned line-drawing graphics recognition. It is a kind of method which is from arc supposition to arc verification. Firstly extract the skeleton of the line-drawing image, then use the piecewise linear polygon to approximate the skeleton for more simplification and reducing the quantum of data. After that, the local discrete curvature of every dominant point of the linear polygon is calculated and the discrete curvature is used to assume some arcs exist locally. Finally the hypotheses are verified by mapping the hypothetical arcs＇ back into the original image according to the arcs＇ circle parameters. This proposed method has the advantages that it can deal with the conditions of rupture, intersection and overlapping and it is robust and fast.

关 键 词：计算机应用 图形识别 圆弧检测 图像骨骼 离散曲率 线条图 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]