基于交叉检验ORB和MCC的图像模板匹配算法  

Image template matching algorithm based on cross-check ORB and MCC

作　　者：丁小艳 王婷[2] Ding Xiaoyan;Wang Ting(Jiangsu Vocational College of Medicine, Yancheng 224005, China;Dalian Ocean University, Dalian 116023,China)

机构地区：[1]江苏医药职业学院,盐城224005 [2]大连海洋大学,大连116023

出　　处：《电子测量与仪器学报》2019年第5期39-45,共7页Journal of Electronic Measurement and Instrumentation

基　　金：江苏省自然科学基金(BK2015609)资助项目

摘　　要：针对传统ORB算法在进行图像模板匹配时稳健性和精确性较低的问题,提出了一种基于交叉检验ORB和最大相关熵准则(MCC)的图像模板匹配算法。首先对模板图像和目标图像分别进行ORB特征点检测,得到模板特征点集和目标特征点集,并利用各点的特征描述符建立两特征点集间的对应关系;然后对建立好对应关系的特征点对进行交叉检验,利用特征点的邻域支持点集从正逆两个方向计算对应点对的准确性;接着在剔除准确性较低的对应特征点对后,通过优化基于最大相关熵准则的代价函数实现两特征点集的精确仿射配准;最终完成图像模板在目标图像中的稳健匹配。该算法在处理大姿态形变和低分辨率下的图像模板匹配问题时具有较强的精确性和稳健性,在公共数据集上的实验验证了该算法优于当前主流的图像模板匹配算法。Aiming at the problem that the traditional ORB algorithm has low robustness and accuracy in image template matching, an image template matching algorithm based on cross-check ORB and maximum correntropy criterion is proposed. Firstly, the algorithm performs the ORB feature point detection on the template image and the target image to obtain the template feature point set and the target feature point set, and uses the feature descriptor of each point to establish the correspondence between the two feature point sets. Then the feature point pairs which have established corresponding relationship are crosschecked, and use the neighborhood support point set of the feature points to calculate the accuracy of the corresponding point pairs from the positive and negative directions. After eliminating the corresponding feature points with lower accuracy, the accurate affine registration of the two feature point set is realized by optimizing the cost function based on the maximum correntropy criterion, and the robust matching of the image template in the target image is completed. When dealing with the image template matching problem with large attitude deformation and low resolution, the proposed algorithm has strong accuracy and robustness. Experiments on public datasets show that the proposed algorithm is superior to the current mainstream image template matching algorithms.

关 键 词：模板匹配 交叉检验ORB 最大相关熵准则 仿射配准 

分 类 号：TN911.73[电子电信—通信与信息系统] TP391[电子电信—信息与通信工程]