基于SIFT算法在遥感图像匹配中的研究

发布时间：2018-03-15 22:03

  本文选题：SIFT算法　切入点：特征匹配　出处：《电视技术》2016年09期 　论文类型：期刊论文

【摘要】：针对遥感图像在光照和几何差异等复杂因素上引起的匹配误差,深入研究了SIFT特征描述符的改进方法。利用特征点圆形区域来构造特征描述符,采用自适应量化策略用来局部区域的划分和梯度直方图的计算,并对每个描述子采用一种插值法重新确立主方向,改进SIFT算法的描述符。降低特征点维数的同时,又保证特征点描述符的独特性和鲁棒性。研究结果表明,改进的SIFT描述符在遥感图像几种复杂环境下都取得了预期的结果,证实了改进算法的可行性。
[Abstract]:Aiming at the matching error caused by complex factors such as illumination and geometric difference, the improved method of SIFT feature descriptor is studied in detail. The feature descriptor is constructed by using the circular region of feature points. Adaptive quantization strategy is used to partition the local region and calculate the gradient histogram. An interpolation method is used to re-establish the principal direction of each descriptor, which improves the descriptor of the SIFT algorithm and reduces the dimension of the feature points at the same time. The research results show that the improved SIFT descriptor has achieved the expected results in several complex environments of remote sensing images, which proves the feasibility of the improved algorithm.
【作者单位】： 重庆邮电大学;重庆信科设计有限公司;
【基金】：重庆市研究生科研创新基金项目(CYS15166)
【分类号】：TP751
，

本文编号：1617004