基于改进SIFT的遥感图像匹配方法
发布时间：2019-01-07 18:16
【摘要】：针对SIFT算法处理遥感图像时存在计算量大、时间代价高的问题,从极值点检测和相似性度量两个方面对SIFT算法进行优化改进。改进算法首先利用距离检测点越近的像素点对其影响越大的特点,在极值点检测时选取距离检测点更近、权重更高的14个相邻点来替代SIFT算法中的26个邻域点,减少极值检测的计算量。其次,在SIFT特征向量匹配的相似性度量方面利用更简单的曼哈顿距离与切比雪夫距离的线性组合来替代欧氏距离,减少特征匹配的计算复杂度,提高匹配效率。最后通过实测遥感数据验证所提方法的有效性。
[Abstract]:Aiming at the problems of large computation and high time cost in processing remote sensing images with SIFT algorithm, the SIFT algorithm is optimized and improved from two aspects: extremum detection and similarity measurement. The improved algorithm firstly takes advantage of the fact that the pixel point closer to the detection point has greater influence on it, and selects 14 adjacent points which are closer to the detection point and higher weight to replace the 26 neighborhood points in the SIFT algorithm. Reduce the computation of extremum detection. Secondly, in the aspect of similarity measurement of SIFT feature vector matching, a simpler linear combination of Manhattan distance and Chebyshev distance is used to replace Euclidean distance, which reduces the computational complexity of feature matching and improves the matching efficiency. Finally, the validity of the proposed method is verified by measured remote sensing data.
【作者单位】： 海军航空工程学院研究生管理大队;海军航空工程学院信息融合研究所;海军航空工程学院电子信息系;
【基金】：国家自然科学基金(61501487,61531020,61471382,61401495) 山东省自然科学基金(2015ZRA06052) “泰山学者”建设工程专项经费
【分类号】：TP751
本文编号：2403976
[Abstract]:Aiming at the problems of large computation and high time cost in processing remote sensing images with SIFT algorithm, the SIFT algorithm is optimized and improved from two aspects: extremum detection and similarity measurement. The improved algorithm firstly takes advantage of the fact that the pixel point closer to the detection point has greater influence on it, and selects 14 adjacent points which are closer to the detection point and higher weight to replace the 26 neighborhood points in the SIFT algorithm. Reduce the computation of extremum detection. Secondly, in the aspect of similarity measurement of SIFT feature vector matching, a simpler linear combination of Manhattan distance and Chebyshev distance is used to replace Euclidean distance, which reduces the computational complexity of feature matching and improves the matching efficiency. Finally, the validity of the proposed method is verified by measured remote sensing data.
【作者单位】： 海军航空工程学院研究生管理大队;海军航空工程学院信息融合研究所;海军航空工程学院电子信息系;
【基金】：国家自然科学基金(61501487,61531020,61471382,61401495) 山东省自然科学基金(2015ZRA06052) “泰山学者”建设工程专项经费
【分类号】：TP751
【相似文献】
相关期刊论文 前2条
1 吴一全;陈飒;;基于Contourlet域Krawtchouk矩和改进粒子群的遥感图像匹配[J];宇航学报;2010年02期
2 ;[J];;年期
相关博士学位论文 前1条
1 胡修兵;中波红外与可见光遥感图像匹配技术研究[D];武汉大学;2014年
相关硕士学位论文 前3条
1 田影;基于局部特征与拓扑约束的遥感图像匹配与目标定位[D];东北大学;2014年
2 赵恒;数字高程（DEM）差异性检测及校验方法研究[D];天津大学;2014年
3 王宇帆;基于局部特征的遥感图像匹配方法[D];上海交通大学;2013年
，本文编号：2403976
本文链接：https://www.wllwen.com/guanlilunwen/gongchengguanli/2403976.html
