Interference Image Registration Based on FPDE-SIFT for Sonar
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.