基于改进sift匹配方法的货架乳制晶识别-计算机科学.pdfVIP

Vo1. 44 No. 9 第 44 卷第9 期 计算机科学 2017 年 9 月 COMPUTER SCIENCE Sep.2017 基于改进 SIFT 匹配方法的货架乳制晶识别 郑建彬白雅贤詹恩奇汪阳 (武汉理工大学信息工程学院 武汉 430000) (武汉理工大学光纤传感技术与信息处理教育部重点实验室 武汉 430000) 摘 要 利用尺度不变特征变换(SIFf)算法识别金装乳制品时易产生误匹自己，ρ、而影响识别的准确率。为了消除误 匹配点的影响并精确识别商品的种类和数量，提出了一种改进的 SIFf误匹配点剔除方法。根据金装乳制品图像形 变较小、多数为刚性交换的特点，首先利用粗匹配对的主方向角度差进行筛选，再计算出模板图和测试图各自特征点 两两之间的距离比，标记距离比出现异常的匹配点，最后通过投票剔除误匹配点。在自建商品图像数据库上将所才是方 法与改进的随机抽样一致性算法、基于图的消除误匹配点方法进行对比测试，结采表明，所提方法在匹配准确率和误 剔除率方面有明显改善。 关键词 商品识别，尺度不变特征变换，误匹配，空间一致随机抽样一致性，图转换匹配算法 中固法分类号 TP391 文献标识码 A DOI 10. 11896/j. issn. 1002-137)♀ 2017.09.059 Improved SIFf Matching Method for Milk Beverage Rec句~ition in Grocery ZHENG Jian-bin BAI Ya-xian ZHAN En-qi WANG Yang (School of Information Engineering ,Wuhan University of Technology ,Wuhan 430000 ,China) (Key Laboratory of Fiber Optic Sensing Technology and Information Processing ,Ministry of Education , Wuhan University of Technology ,Wuhan 430000 ,China) Abstract Using scale-invariant feature transform (SIFf) to identify boxed milk beverages is vulnerable to errors caused by mismatching ,which will affect the accuracy of identification. To reduce the impact of mismatching and in crease the recognition rates ,a new method was put forward to eliminate the mismatching points. Since the deformation of boxed milk beverage images is mostly rigid transformation ,the angle differences of the key-points between reference image and observation image are calculated in order to remove some mismatching points. Then the ratios of pairwise dis- tances between reference image and observation image are used to mark the