基于点对匹配的印刷电路板CT图像配准方法
Point-pair Match Based Registration Algorithm for PCB CT Images
-
摘要: 针对PCB CT图像在拼接时存在较多相似区域无法有效配准的问题,提出一种基于点对匹配的方法.首先根据邻近特征点生成特征点对,并简化一种快速低维的特征描述子KPB-SIFT来描述特征点对,然后通过规定点对方向和点对分类匹配降低误匹配,最后给出一种匹配点方向约束条件,并基于该条件提出一种错误匹配点抑制算法提取正确匹配点.实验结果表明,该算法在保持较低时间复杂度的同时,匹配正确率达到95.58%,能够提取较多均匀分布的正确匹配点,从而能有效地解决PCB CT图像的配准问题.Abstract: Aiming at the problem that it is impossible to effectively registration PCB CT images that contain many similar regions,an algorithm is proposed based on point-pair match.We first generate the point-pairs from adjacent points and describe them using the simplified method of a fast low-dimensional descriptor called KPB-SIFT,then reduce mismatches by means of the prescriptive point-to-point directions and classification match.Finally a mismatch resistant algorithm is developed to pick up the correct match points based on an error constraint of matched point direction.Experimental results show that the algorithm maintains a low time complexity and keeps a high correct match ratio of 95.58%.It can effectively solve the registration problem of PCB CT image by extracting correct matches which are uniformly distributed.