基于改进SIFT匹配方法的货架乳制品识别  

Improved SIFT Matching Method for Milk Beverage Recognition in Grocery

作　　者：郑建彬[1,2] 白雅贤[1,2] 詹恩奇[1,2] 汪阳[1,2] 

机构地区：[1]武汉理工大学信息工程学院,武汉430000 [2]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

出　　处：《计算机科学》2017年第9期315-319,共5页Computer Science

基　　金：国家自然科学基金(61303028)资助

摘　　要：利用尺度不变特征变换(SIFT)算法识别盒装乳制品时易产生误匹配,从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SIFT误匹配点剔除方法。根据盒装乳制品图像形变较小、多数为刚性变换的特点,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提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,结果表明,所提方法在匹配准确率和误剔除率方面有明显改善。Using scale-invariant feature transform （SIFT） 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 abnormal ratios, and the mismatching points are removed by majority vote method. The proposed method is compared with SC-RANSAC（spatially consistent random sample consensus） algorithm and graph transformation matching method through experiments. The results show that the proposed method could improve the recognition accuracy effectively.

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 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]