基于FPDE-SIFT的声呐干涉图像配准方法

Interference Image Registration Based on FPDE-SIFT for Sonar

图像配准是声呐进行高精度干涉测量的保障,该文针对水下目标的声呐图像配准,提出了一种基于4阶偏微分方程尺度不变特征变换的声呐干涉图像配准方法。该方法聚焦声呐图像配准的难点,首先基于4阶偏微分方程构建尺度空间,在保持图像细节的前提下滤除噪声,提高特征提取的准确度;对于残余噪声造成的特征点误检,借助特征点的相位一致性信息加以筛选,精简特征点样本集;最后对特征点匹配策略进行优化,提出改进的快速样本一致性匹配策略剔除特征点的误匹配。算法增加了匹配点对的数量,提高了匹配点对的准确度,实现了声呐干涉图像的精确配准。水池实验和外场试验表明,该文所提出的算法相较现有算法对声呐图像有着更好的适用性,配准后的均方根误差与留一法均方根均小于1像素,达到了亚像素配准精度。

Image registration is the cornerstone of sonar for high-precision interferometry.This study presents an innovative method for registering sonar interference images,utilizing the Fourth-order Partial Differential Equation(FPDE)in conjunction with the scale-invariant feature transform.This technique is specifically tailored for underwater sonar targets.This method specifically addresses the challenges associated with sonar image registration.First,we establish the scale space by employing the FPDE.This process filters noise while preserving image details,resulting in an improved accuracy of feature extraction.The proposed method utilizes phase congruency information to counter false feature point detection due to the residual noise,thereby screening and simplifying the sample set of feature points.Ultimately,the features point matching strategy undergoes optimization,with an enhanced fast sample consensus matching strategy proposed to rectify feature point mismatches.The algorithm increases the number of matching point pairs and augments their precision,ultimately achieving precise registration of sonar interference images.Rigorous tests,both under controlled conditions and lake environments,demonstrate the algorithm’s superior applicability to sonar images compared with existing approaches.The root-mean-square-error and mean-square-error are calculated post-registration using leave-one-out analysis,both are under one pixel,attesting to the algorithm’s achievement of sub-pixel registration accuracy.