基于SURF算子的遥感图像配准方法研究  

Research on Remote Sensing Image Registration Method Based on SURF

作　　者：李维 王瑞[2] 邓传斌 Li Wei;Wang Rui;Deng Chuanbin(School of Zhongxing Communication,Xi’an Traffic Engineering Institute,Xi’an 710300,China;Xi’an Mordern Control Technology Research Institute;ZTE Corporation)

机构地区：[1]西安交通工程学院中兴通信学院,西安710300 [2]西安现代控制技术研究所 [3]中兴通讯

出　　处：《单片机与嵌入式系统应用》2023年第6期61-64,69,共5页Microcontrollers & Embedded Systems

基　　金：国家自然科学基金(52202491);西安交通工程学院中青年基金项目(2022KY-38)。

摘　　要：为了解决遥感图像配准的可靠性和实时性问题,提出了一种基于SURF算子的图像自动配准方法。SURF算法提取图像中适应旋转尺度变化的局部不变特征角点时使用积分图像的计算,可有效提高检测速度,用于对实时性要求高的场合,选择基于欧氏距离的最近邻与次近邻的距离之比作为相似度测量方法。多种不同场景图像对比实验结果表明,该方法可实现多种复杂图像的精确配准,自动化配准程度高。In order to solve the reliability and real-time problems of remote sensing image registration,an automatic image registration method based on SURF operator is proposed.SURF algorithm can effectively improve the detection speed by using integral image calculation when extracting locally invariant feature corners adapted to rotation scale changes.For situations with high real-time requirements,Euclian-based distance ratio of nearest neighbor to second-nearest neighbor is selected as the similarity measurement method.The experiment results show that this method can achieve accurate registration of complex images with high degree of automatic registration.

关 键 词：多传感器 遥感图像 SURF算子 图像配准 积分图像 RANSAC 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]