基于投影的快速模板匹配算法  

Fast Template Matching Algorithm Based on the Projection

作　　者：何新鹏[1] 黄英[1] 刘奇[1] 刘云峰[1] 潘琦[1] 

机构地区：[1]广东工业大学自动化学院,广东广州510006

出　　处：《自动化技术与应用》2011年第7期72-75,共4页Techniques of Automation and Applications

基　　金：国家自然科学基金资助项目(编号60272089);"可拓检测的物元机制研究;广东省自然科学基金资助项目(编号04009464);"可拓检测的要素处理及焦点显形的机理研究"

摘　　要：针对传统模板匹配算法效率不高,计算时间长等缺点,本文提出一种基于投影的快速匹配算法,利用对图像色度的投影到一水平轴,使图像数据空间的维度降到一维,并用内积计算相关度,根据已计算的相关度,决定下一次匹配的步长,过滤一些可忽略的子图投影;在计算水平域投影时,不需要从新计算,只需在上一次的结果中,增减一行色度,而且,一次投影,可以进行多次匹配。这样大大减少计算量,获得更快的匹配速度。实验结果表明,本算法准确率高,能满足高实时性的要求。The traditional template matching algorithm is not efficient： long computing time and other shortcomings.The paper propose a fast matching algorithm based on the projection.In this paper the date of the image will project to a horizontal axis to get the dimensions of the image data down to one-dimensional space and caculate the relevance with the inner product.the calculated relevance,to decide the step of matching in order to filtering some of the sub-map projection;when caculating the projection of the image area,new calculation is not necessan,and only use the results in the last with exchange the image date of the first and last line.Projecting one time can do match many times.It can greatly reduce the computation and get faster matching speed.Experimental results show that high accuracy of the algorithm meeting high real-time requirements.

关 键 词：图像匹配 机器视觉 投影 降维 向量内积 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]