印品检测过程中基于SIFT算法缩小匹配范围的方法  

Method to Reduce Matching Range in Print Detection Based on SIFT Algorithm

作　　者：赵立辉[1] 杨红喆[2] 郭栋[1] 霍春宝[1] 

机构地区：[1]辽宁工业大学,锦州121001 [2]北京理工大学,北京100083

出　　处：《包装工程》2013年第17期104-107,共4页Packaging Engineering

基　　金：辽宁省科技攻关项目(2011201035);辽宁省教育厅项目(L2012230)

摘　　要：针对多个CCD采集多幅图像会产生一定的重叠区域,为了实现印品在线检测的要求,提出了一种结合SIFT和缩小匹配范围的图像检测方法。该方法基于SIFT提取特征点,改进了局部搜索范围,利用RANSAC算法计算图像坐标变换矩阵,采用多分辨率融合方法对拼接图像进行融合处理。结果表明,采用该方法可以减少匹配和图像检验时间,降低估算概率,完成检测图像拼接。A print detection method based on SIFT to reduce matching range was proposed to solve image overlap region when acquiring images with the multiple CCD for online printing detection requirement. The method improve local searching range using image extraction feature of SIFT. RANSAC algorithm was used to calculate the image coordinate transform matrix and the multi resolution fusion method was used to fuse stitching images. The results showed that the method can effectively reduce the matching and test time of image, reduce the error probability, and complete image stitching.

关 键 词：SIFT CCD 图像拼接 印品检测 

分 类 号：TB807[一般工业技术—摄影技术]