改进配准测度的SUFT红外图像快速配准算法  

A Fast Registration Algorithm for Infrared Images Based on SUFT with Improved Registration Measure

作　　者：魏新[1] 马丽华[1] 李云霞[1] 徐志燕[1] 李大为[1] 

机构地区：[1]空军工程大学信息与导航学院,西安710077

出　　处：《电光与控制》2012年第11期77-80,共4页Electronics Optics & Control

摘　　要：为方便电路板卡故障诊断,实现红外图像快速、有效配准,提出一种基于SUFT(Speeded-Up Robust Features)和相似四边形的红外图像快速配准算法。该算法首先对红外图像进行特征点检测,生成SUFT特征点描述子;然后采用欧氏距离进行相似性度量,提取粗匹配特征点对,再利用相似四边形进行精匹配,去除误配准点对;最后依据精匹配点对求解变换模型参数,实现红外图像的配准。实验表明,改进后的算法能有效剔除误匹配点对,提高配准精度,配准结果较理想,与同类算法相比耗时较短。因此该算法具有快速性、稳定性等优点,且配准精度较高,有很好的实用价值。In order to facilitate the fault diagnosis of circuit boards and realize fast and accurate registration of infrared images,this paper presents a fast registration algorithm for infrared images based on Speeded-Up Robust Features（SUFT） and similar quadrilateral.Firstly,the algorithm detects feature points of infrared image and generates SUFT feature point descriptor.Then,the Euclidean distance similarity measure is used to extract feature points of coarse matching,and the similar quadrilateral accuracy matching is used to remove mistaken matching on schedule.Finally,the accurate matching points are used for obtaining the model parameters,thus can realize the registration of infrared image.Experimental results show that the improved algorithm can effectively eliminate the error of the matching points and improve the registration precision.The registration result is more ideal and it consumes shorter time compared with other similar algorithms.The algorithm has such advantages as rapidity,high stability and high registration precision,which has good practical value.

关 键 词：图像配准 SUFT 相似四边形 精匹配 

分 类 号：V271.4[航空宇航科学与技术—飞行器设计] TP391[自动化与计算机技术—计算机应用技术]