| 一种改进SIFT算法 |
| |
| 引用本文： | 师艳伟,杨晶东,杨敬辉.一种改进SIFT算法[J].江南大学学报(自然科学版),2014,13(4):403-410. |
| |
| 作者姓名： | 师艳伟 杨晶东 杨敬辉 |
| |
| 作者单位： | 1. 上海理工大学光电信息与计算机工程学院,上海,200093
2. 上海第二工业大学经济与管理学院,上海,201209 |
| |
| 基金项目： | 国家自然科学基金项目，上海市教委科研创新项目，上海市研究生创新基金项目 |
| |
| 摘 要： | 针对图像特征提取算法-SIFT,特征描述器维数较高,特征匹配耗费时间较长,匹配过程中存在相同图像不能匹配和不同图像能够匹配等问题,提出了一种改进SIFT算法与KD-tree搜索匹配算法相结合的新方法。采用KD-Tree算法替代传统链表式搜索方法降低特征点匹配时间;把特征点间距离和特征描述子内积同时作为匹配标准,加入相应匹配阈值减少匹配错误率,并通过理论和实验证明采用欧几里德距离作为相似性度量具有更高的匹配成功率。实验结果表明,在图像特征匹配中,该算法能够有效减少特征匹配错误率,大幅度降低匹配时间,具有较好的实时性和鲁棒性。
|
| 关 键 词： | 尺度不变特征变换 特征提取 KD-维树搜索 相似性度量 匹配阈值 |
Research on an Improved SIFT Algorithm |
| |
| SHI Yanwei,YANG Jingdong,YANG Jinghui.Research on an Improved SIFT Algorithm[J].Journal of Southern Yangtze University:Natural Science Edition,2014,13(4):403-410. |
| |
| Authors: | SHI Yanwei YANG Jingdong YANG Jinghui |
| |
| Institution: | SHI Yanwei,YANG Jingdong,YANG Jinghui( 1.School of Optical Electrical and Computer Engineering,University of Shanghai for Science and Technology,Shanghai 200093, China; 2.School of Economics and Management, Shanghai Second Polytechnic University, Shanghai 201209, China;) |
| |
| Abstract: | The scale invariant feature transform（SIFT） algorithm is widely used in the 3D reconstruction,image registration and object recognition,etc.There are some problems in the study of SIFT,its characterization instruments have high dimensions,which increase the time-consuming; and another problem is that different images can match but the same image can not match.According to this,this paper chooses k-dimension search instead of the traditional chain table search method,proves by experiments that using the Euclidean distance as the similarity measurement is more accurate and reliable,and the improved SIFT joined distance and inner product matching threshold can solve mismatch.The results indicates that the improved SIFT algorithm can solve the fitting errors,and has less time consuming,higher accuracy and strong robustness. |
| |
| Keywords: | SIFT feature extraction KD-Tree search similarity measurement matching threshold |
| 本文献已被 CNKI 维普 万方数据 等数据库收录！ |
|
