局部特征信息的航拍图像匹配算法  

Aerial image matching algorithm based on local feature information

作　　者：倪佳忠 马志龙 NI Jiazhong;MA Zhilong(Zhejiang Institute of Surveying and Mapping Science and Technology, Hangzhou Zhejiang 311100, China;Hangzhou Tiantu Geographic Information Technology Company Limited, Hangzhou Zhejiang 310007, China)

机构地区：[1]浙江省测绘科学技术研究院,浙江杭州311100 [2]杭州天图地理信息技术公司,浙江杭州310007

出　　处：《北京测绘》2022年第3期228-232,共5页Beijing Surveying and Mapping

摘　　要：针对快速定向和描述符构建(oriented fast and rotated brief,ORB)算法对航拍图像匹配稳定较差、保留内点较少等问题,提出一种基于非线性尺度空间的改进ORB算法。首先借助非线性扩散滤波算法构建非线性尺度空间,相较高斯尺度空间可更好的保护图像细节信息,增强匹配算法的鲁棒性;然后借助Fast算法提取特征点,再以特征点为中心选择合适的窗口提取描述符;最后通过网格运动统计算法筛选内点。实验得出:所提算法匹配准确率相较ORB算法提高5%左右,相较加速“风”算法匹配效率提高65%左右,表明所提算法具有更高匹配正确率和效率,可更好应用在航拍图像匹配中。An improved ORB algorithm based on nonlinear scale space was proposed to solve the problems of poor stability of ORB algorithm in aerial image matching and fewer reserved interior points.Firstly,nonlinear diffusion filtering algorithm wasapplied to construct nonlinear scale space,which could protect image details better than Gaussian scale space and enhance the robustness of matching algorithm.Then feature points were extracted by Fast algorithm,and BRIEF descriptors were extracted by selecting appropriate windows centering on feature points.Finally,mesh motion statistics algorithm was explored to screen interior points.The experimental results showed that the matching accuracy of the proposed algorithm was about 5%higher than that of ORB algorithm,and the matching efficiency was about 65%higher than that of AKAZE algorithm,indicating that the proposed algorithm had higher matching accuracy and efficiency,and could be better applied in aerial image matching.

关 键 词：无人机航拍图像 非线性尺度空间 ORB算法 网格运动统计算法 

分 类 号：P231[天文地球—摄影测量与遥感]