摘要
针对SIFT(scale-invariant feature transform)算法计算速度低、匹配精度不高的问题,提出一种结合多维定标和局部纹理特征的改进SIFT匹配算法。通过多维定标算法(multidimensional scaling,MDS)对128维SIFT特征描述符进行降维,提高SIFT特征匹配的计算速度,与其它降维算法相比,MDS保证了数据的几何拓扑性;基于特征一致性匹配规则和比值一致性匹配规则,提出一种改进的双向匹配策略;分析比对匹配点对邻近区域的LBP纹理特征,进一步降低误匹配率。仿真结果表明,与同类算法相比,该方法在匹配精度和匹配速度上都得到一定程度的改善。
Aiming at the problems of the low speed and the poor matching precision of SIFT algorithm（scale-invariant feature transform）,an improved SIFT matching algorithm combining multidimensional scaling with local texture feature was proposed.The 128 dimension feature descriptor of SIFT was reduced using multidimensional scaling algorithm（multidimensional scaling,MDS）,improving the computing speed of SIFT feature matching.Compared with other algorithms,MDS ensured the topological structure in the geometry of data.An improved bidirectional matching strategy was proposed based on the consistency of characteristics and the consistency of ratio.LBP features of the matching points’ adjacent area were analyzed and compared to reduce the error matching rate further.Simulation results show that the proposed method provides better accuracy and higher speed than other similar algorithms.
作者
程诗梦
周之平
李忠民
CHENG Shi-meng ZHOU Zhi-ping LI Zhong-min(School of Information Engineering, Nanchang Hangkong University, Nanchang 330063,Chin)
出处
《计算机工程与设计》
北大核心
2017年第11期3087-3092,共6页
Computer Engineering and Design
基金
国家自然科学基金项目(61263040)
