IEEE Journal of Oceanic Engineering · 2019 · 18 citations · 27 references
EngineeringMarine EngineeringOcean EngineeringComputer VisionOcean AcousticsOcean InstrumentationChange DetectionMotion AnalysisMotion CorrectionMotion EstimationSonar SystemsUnderwater NavigationChange Detection From Synthetic Aperture Radar ImageryCoherent DetectionMotion CompensationPhase CoherenceSynthetic Aperture SonarUnderwater Change DetectionUnderwater ImagingUnderwater Navigation Systems
The capability to detect changes in an underwater scene has many applications, including environmental monitoring, surveillance of strategic maritime waterways, and naval mine countermeasures. This paper examines the automated detection of changes in repeat-pass synthetic aperture sonar (SAS) imagery by exploiting the loss of phase coherence caused by differences in the scene. A renavigation method based on track registration is developed for performing the coregistration of images. The method first determines the corrections to be applied to the repeat-pass navigation data by optimizing a linear track model using the estimated across-track pixel displacements. A second optimization is then performed using the along-track pixel displacements to determine a velocity correction for each ping location of the repeat-pass image to compensate for residual motion errors. The corrected navigation solution is then used to rebeamform the repeat-pass image onto the same focal points as the reference image. The method is demonstrated on a pair of repeat-pass SAS images acquired using the AquaPix INSAS2 sonar, where it is shown that it is not necessary to have access to raw element data to coregister images accurately enough for coherent change detection requirements. This technique is then applied to the second data set obtained using the HISAS 1030 sonar for the purpose of comparing both coherent and noncoherent (amplitude-only) approaches to change detection in scenes with low and high clutters. False alarms caused by the acoustic shadows of objects in high-clutter areas are mitigated using the single-pass interferometric coherence as a reference mask. The results show the successful use of the repeat-pass coherence to detect changes between SAS images with temporal baselines of several days on scenes consisting of mud, gravel, and sand, and demonstrate that a coherent approach is capable of detecting large-scale differences (caused by the introduction or removal of targets) as well as subtle scene changes that are not detectable using noncoherent methods.
27
Object recognition from local scale-invariant features
David Lowe · 1999 · 16.2K citations
Robust Real-Time Face Detection
Paul Viola, Michael Jones · International Journal of Computer Vision · 2004 · 14.2K citations
Synthetic aperture radar interferometry
Richard Bamler, Philipp Hartl · Inverse Problems · 1998 · 2K citations
Image change detection algorithms: a systematic survey
Richard J. Radke, Srinivas Andra, Omar Al-Kofahi et al. · IEEE Transactions on Image Processing · 2005 · 1.9K citations
Coherence estimation for SAR imagery
R. Touzi, A. Lopes, J. Bruniquel et al. · IEEE Transactions on Geoscience and Remote Sensing · 1999 · 716 citations