基于改进ORB算法的遥感图像自动配准方法

Automatic registration method for remote sensing images based on improved ORB algorithm

针对遥感图像自动配准的问题,提出了一种基于改进定向二进制简单描述符(oriented brief,ORB)算法的遥感图像自动配准方法。该方法主要由3个步骤组成:首先是特征匹配,利用改进的ORB算法提取特征点,并建立描述符进行匹配,获取初始控制点;然后采用随机采样一致性方法,结合变换参数估计,剔除可能的错误匹配;最后利用最小二乘法估计的变换参数,对图像进行几何纠正。分别利用2组卫星光学遥感图像和1组SAR图像进行基于改进ORB算法的自动配准方法试验,并与基于尺度不变特征变换(scale-invariant feature tramsform,SIFT)算法和加速鲁棒性特征(speeded up robust features,SURF)算法的自动配准方法进行了比较。试验结果表明,该方法能获得与SIFT算法和SURF算法相当或者更高的配准精度,并在配准效率上有较大提高。

Aiming at reliable registration of remote sensing images,the authors present in this paper a remote sensing image registration method based on improved ORB（oriented brief） algorithm.The proposed method mainly includes three stages： The first stage is feature matching,the improved ORB algorithm is used to detect features and build descriptors,and the descriptors are matched to obtain initial control points.The second stage is to employ RANSAC（random sample consensus） processing via transformation parameters estimation to remove possible wrong matching points.The third stage is to rectify the image based on the transformation parameters calculated by the least square method.The proposed method is evaluated based on two sets of optical and SAR remote sensing images,and is compared with the registration methods based on SIFT and SURF algorithm.The results show that the method proposed in this paper can provide the same accurate remote sensing image registration result as or even the higher result than the methods based on SIFT and SURF algorithm,and can obtain improved efficiency.