基于改进AKAZE算法的快速图像配准方法  

Fast Image Registration Method Based on Improved AKAZE Algorithm

作　　者：赵卫东 刘俊德 王慢慢 李丹 Zhao Weidong;Liu Junde;Wang Manman;Li Dan(School of Electrical and Information Engineering,Anhui University of Technology,Maanshan 243032,Anhui,China)

机构地区：[1]安徽工业大学电气与信息工程学院,安徽马鞍山243032

出　　处：《激光与光电子学进展》2023年第6期80-86,共7页Laser & Optoelectronics Progress

基　　金：安徽省自然科学基金(2108085MF225);安徽普通高校重点实验室开放基金(PEMC1902)。

摘　　要：针对无人机影像匹配容易出现匹配速率低、鲁棒性差的问题,提出一种改进AKAZE(accelerate-KAZE)算法的快速图像匹配方法.首先,在特征提取阶段,使用AKAZE算法对非线性尺度空间进行构建,采用fast retina keypoint(FREAK)描述符对特征点进行有效描述;之后,利用基于网格的运动估计(GMS)方法对所获得特征点进行预匹配,并进行鲁棒性优良的区分;最后,在随机抽样一致性(RANSAC)算法的基础上对匹配结果进行进一步筛选.为了验证所提方法的有效性,使用Oxford标准图像数据集和RSSCN7遥感图像数据集进行实验,对所提方法与改进AKAZE、ORB、KAZE、SIFT+FREAK算法进行对比,确保所提方法在保持较高准确率的同时能够实现快速的图像配准.在图像光照变化、模糊变换及压缩变换下,所提方法能够保持较好的鲁棒性,可以满足无人机影像实时匹配的需求.A fast image matching method based on the improved accelerateKAZE(AKAZE)algorithm is proposed to address the issues of low matching rate and weak robustness in UAV image matching.The proposed method first constructs the nonlinear scale space during the feature extraction stage using the AKAZE algorithm,and then efficiently describes the feature points using the fast retina keypoint(FREAK)descriptor.Later,the obtained feature points are prematched using the gridbased motion statistic(GMS)method to distinguish them with high robustness.The matching outcomes are then further screened using the basis of random sample consensus(RANSAC)algorithm.Experiments are conducted on an Oxford standard image dataset and an RSSCN7 remote sensing image dataset to verify the effectiveness of the proposed method.The proposed method is compared with the improved AKAZE,ORB,KAZE,and SIFT+FREAK algorithms.Continuous testing can guarantee that the proposed method can achieve fast image registration while maintaining high accuracy.It can maintain a high robustness under image illumination change,fuzzy transformation,and compression transformation and can meet the needs of UAV image realtime matching.

关 键 词：图像处理 特征匹配 FREAK 网格运动估计 随机抽样一致性算法 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]