摘要
针对SIFT算法在图像配准过程中存在遗漏掉大量的正确匹配对且产生误匹配对等问题,本文提出一种基于尺度、方向和距离约束的改进SIFT配准方法(SODC-SIFT).首先用SIFT算法对图像进行初匹配,利用特征点的尺度和主方向的差异剔除误匹配对;然后用最小二乘法求出两幅图像的几何关系;最后利用距离约束,迭代求解,找出正确匹配对,从而实现图像的精确配准.实验结果表明,本文方法与SIFT算法相比具有较强的鲁棒性,能够提升正确匹配对的数量并提高图像的正确匹配率,增强了算法匹配的精确性.
Traditional SIFT method has certain false matching points and will miss a large number of correct matching points in image registration process. To solve this problem, an improved SIFT algorithm based on scale, orientation and distance constraint (SODC-SIFT) is proposed in this paper. First, the images are matched with original SIFT algorithm, and the scale and orientation joint restriction are used to remove mismatching pairs. Then, the least square method is employed to calculate the geometric relationship between the two images. Finally, iteration is conducted to select the correct matching points by distance constraint, so as to achieve accurate image registration. Experimental results show that the proposed method has greater robustness and achieves more correct matching pairs than the traditional SIFT method. Furthermore, the correct matching rate is increased and the accuracy of the algorithm is enhanced.
出处
《纳米技术与精密工程》
CAS
CSCD
北大核心
2017年第1期36-43,共8页
Nanotechnology and Precision Engineering
基金
国家自然科学基金资助项目(61363049)
中国博士后第八批特别资助基金资助项目(2015T80155)
江西省自然科学基金资助项目(20161BAB212033)
中国科学院复杂系统管理与控制国家重点实验室开放基金资助项目(20140101)
南昌航空大学无损检测技术教育部重点实验室开放基金资助项目(ZD201429007)
江西省教育厅科学基金资助项目(GJJ150751)
关键词
图像配准
SIFT
特征提取
特征匹配
image registration
SIFT
feature extraction
feature matching