区域梯度特征耦合运动约束的图像匹配算法  

Image matching algorithm based on region gradient feature and motion constraint

作　　者：鲁明珠 张思祥[2] 周围[2] 

机构地区：[1]沧州师范学院机械与电气工程学院,河北沧州061001 [2]河北工业大学机械工程学院,天津300130

出　　处：《计算机工程与设计》2017年第12期3390-3395,共6页Computer Engineering and Design

基　　金：国家重大科学仪器开发基金项目(2012YQ060165);河北省自然科学基金项目(B2014202260);河北省科技支撑基金项目(15210336)

摘　　要：为解决当前基于图像灰度特性的匹配算法在面对灰度变化较大或存在较多的干扰时,易出现较高误配率的不足,提出区域梯度特征耦合运动约束的图像匹配算法。利用FAST算子检测图像的特征点,借助Hessian矩阵剔除伪特征点;构建圆形区域,求取其梯度特征,改进SIFT机制,生成特征描述符。建立双向匹配机制,完成特征点的匹配;根据特征点运动方向的特点,构造图像运动约束规则,对匹配特征点进行提纯。仿真结果表明,与当前图像匹配算法相比,所提算法具有更高的抗干扰能力以及最低的匹配耗时。To solve the defects of high mismatch rate caused by large change in image gray level change or many interferences existing in current image matching algorithms based on gray characteristic,an image matching algorithm based on region gradient feature and motion constraint was proposed.The FAST operator was used to extract the feature points of the image,and the Hessian matrix was used to eliminate the false feature points.The circular region and the gradient feature in the region were constructed to improve the SIFT mechanism for generating the feature descriptor.The bidirectional matching mechanism was established to match the feature points.According to the characteristics of the image and the moving direction of the feature points,the image motion constraint rules were constructed to purify the matching feature points.The simulation results show that the proposed algorithm has higher anti-jamming capability and minimum matching time compared with current image matching algorithms.

关 键 词：图像匹配 区域梯度特征 图像运动约束 FAST算子 HESSIAN矩阵 双向匹配 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]