印刷品缺陷检测系统的快速配准方法研究

Research on fast registration algorithm for the detecting system of the printed matter defects

根据印刷品的特点和质量要求,以药盒为例对印刷品图像配准算法进行研究,提出了一种新的针对印刷品图像的配准算法。在基于线段特征的粗配准后,在离线确定的纹理丰富的四个局部(包括待测样品和标准模板)区域采用SIFT方法估计特征点信息,综合特征点匹配对信息估计位置偏差量,经平移实现待测印刷品与模板的高精度配准。实验结果表明,该算法能够快速、有效地配准采集到的印刷图像。

According to the characteristics and the quality requirements of the printed matter, by taking boxes of the medicine as an example, the printing image registration algorithm is studied, and then a new print image registration algorithm is proposed in this paper. After the coarse registration based on the line feature, the SIFT method is used in the rich texture four local areas （including the sample under test and the standard template） determined in offline to estimate the characteristic point information, and the matching feature points are synthesized to estimate the position information deviation value, and the printing material to be tested matches the template with high accuracy through the translation. Experimental results show that the algorithm can rectify the collected print images quickly and effectively.