摘要
由于复杂的成像机制,对于存在丰富地表附属物的城镇、村庄等区域,从不同角度获取的雷达图像间存在明显的非线性辐射畸变和阴影特征差异,使得现有的图像配准方法不再适用。为了解决这一问题,本文在辐射不变特征变换(Radiation-Variation Insensitive Feature Transform,RIFT)的框架下分别从特征计算和特征向量生成两个方面进行改进,提出了一种基于多特征描述和阴影感知的多视角雷达图像配准方法。在特征计算阶段,算法利用log-Gabor滤波器实部和虚部的空间域性质对雷达图像中两种不同特征进行提取,使单一的特征变化不再影响特征描述,增强算法对非线性辐射畸变的鲁棒性。在特征向量生成阶段,利用log-Gabor滤波器对阴影区域响应低的性质,实现对图像阴影区域的检测,并为受阴影影响严重的子邻域所生成的特征向量施加掩膜,有效消除了阴影特征差异对图像配准的影响。实验结果表明,针对多视角雷达图像配准问题,本文方法在均方根误差、成功匹配概率和正确匹配点数上优于现有算法,能够实现多视角下雷达图像的精确配准。
Due to the complex imaging mechanism,there are obvious nonlinear radiation distortion and shadow feature dif⁃ferences among radar images obtained from different angles for areas such as towns and villages with abundant surface ap⁃pendages,making the existing image registration methods no longer applicable.In order to solve this problem,this paper proposes an image registration method based on multiple feature description and shadow perception,which is improved from feature calculation and feature vector generation under the framework of radiation-variation invariant feature transfor⁃mation(RIFT).In the feature calculation stage,the algorithm extracts two different features from the radar image using the spatial domain properties of the real and imaginary parts of the log-Gabor filter,so that the single feature change no lon⁃ger affects the feature description,and enhances the robustness of the algorithm to nonlinear radiation distortion.In the feature vector generation stage,the low response of log-Gabor filter to the shadow region is utilized to detect the shadow re⁃gion of the image,and the mask is applied to the feature vector generated by the sub-grid which is seriously affected by shadow.This method effectively eliminates the influence of shadow feature difference on image registration.The experi⁃mental results show that the proposed method is superior to the existing algorithms in terms of root mean square error,prob⁃ability of successful matching and number of correct matching points,and can achieve accurate registration of radar images from multiple views.
作者
李焱磊
刘静博
刘文成
刘云龙
郭宇豪
王明明
梁兴东
LI Yanlei;LIU Jingbo;LIU Wencheng;LIU Yunlong;GUO Yuhao;WANG Mingming;LIANG Xingdong(National Key Laboratory of Microwave Imaging Technology,Aerospace Information Research Institute,Chinese Academy of Sciences,Beijing 100190,China;School of Electronic,Electrical and Communication Engineering,University of Chinese Academy of Sciences,Beijing 100049,China)
出处
《信号处理》
CSCD
北大核心
2023年第9期1633-1650,共18页
Journal of Signal Processing
基金
国家自然科学基金(62271471)。