基于改进Chamfer匹配的台标识别  

TV Symbol Recognition Based on Improved Chamfer Matching

作　　者：胡星火[1] 姚剑敏[1] 林志贤[1] 郭太良[1] 

机构地区：[1]福州大学物理与信息工程学院,福州350002

出　　处：《计算机工程》2013年第1期195-199,共5页Computer Engineering

基　　金：国家"863"计划基金资助项目(2008AA03A313);福建省自然科学基金资助项目(2010J01333;2011J01347);福建省教育厅基金资助项目(JA10033;JA09003)

摘　　要：已有电视台台标的识别技术多基于视频流处理,实时性较差,且对半透明台标识别效果不佳。为此,提出一种基于改进Chamfer匹配的台标识别方法。利用Canny算子提取台标边缘颜色及梯度方向,采用RANSAC算法对台标边缘进行优化,通过距离变换计算边缘梯度方向的距离图,并基于距离最短准则对其识别。实验结果表明,该方法能实现台标的单帧识别,克服半透明台标的识别难题,平均识别率达97.7%,平均识别时间为801 ms。The real-time of existing TV symbol recognition method based on streaming video processing is not good, and has a bad result of recognition to semitransparent one. In order to solve this problem, this paper proposes a recognition method of TV symbol based on improved Chamfer matching. It extracts the TV symbol color and edge gradient direction to represent the shape of it by Canny operator, and uses RANSAC algorithm to optimize and fit the edge. It computes the distance image of edge gradient direction by distance transformation, and identification is based on the principle of shortest distance. Experimental results show that the method can realize the recognition of TV symbol in a single frame and overcomes the difficulty of the recognition of semitransparent TV symbol, while its recognition rate is 97.7% and average recognition time is 801 ms.

关 键 词：网格简化 边折叠 三角形法向约束因子 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]