Interferometric Synthetic Aperture Sonar Signal Processing Algorithms Research

Interferometric Synthetic Aperture Sonar (InSAS) provides a means of obtaining 3-dimensional images of scene. Three-dimensional information is derived from the phase difference between two (or several) complex images of SAS taken from separate receiving array across the track. It can be used in sea-route measurement, terrain mapping, underwater archeology, salvage and oil industry etc. It is useful to improve mine's identify probability, too. The thesis involves all-aspect of the signal processing of interferometric synthetic aperture sonar, especially in the challenging problems such as phase-preserved SAS imaging algorithm, image registration, interferogram denoising filtering and two-dimensional phase unwrapping. In allusion to difficulties and emphases, some new methods are proposed. The study is important to prompt the development of interferometric synthetic aperture sonar in China.The thesis firstly introduces the basic principle of InSAS, and reviews the history of its developing, and depicts briefly its actual state.Secondly each step in InSAS signal processing is studied as following. The author discusses the phase-preserved SAS imaging algorithm, which is the precondition of interferometric processing. Several classical SAR algorithms used in SAS are analyzed. Because of the contradiction of the unambiguous range and along-track sampling requirements, multi-aperture technique is widely used in SAS. At the same time, the "stop-hop-stop" approximation is no longer valid in SAS due to the low speed of sound. The receiver has unavoidable movement, when echoes get to receiver from different distances. In order to overcome the problem, an image forming algorithm of multi-aperture SAS is presented in the thesis, which is suitable for the case of non-"stop-hop-stop".Because of the unstability of platform and the complexity of the propagation characteristics of sound in ocean, the motion compensation problem of the SAS is more serious than SAR. The thesis studies deeply the motion compensation problem of SAS based on echoes data.The image registration was discussed the following three steps: initial registration, pre-filtering, sub-pixel registration. An adaptive range- filtering algorithm was proposed, which filters the different parts of object spectrum in master and slave synthetic aperture sonar images.The fundament of flat terrain phase effect and its removal are stated. The interferogram consists, for each pixel, of the complex conjugate product of two SAS images directly, which can't reflect the actual terrain change. The general phase trend in range can be considered as the phase generated by an ideally flat terrain and is often subtracted from the interferogram before further processing.After investigating statistical property of the interferogram, the problem of interferogram phase denoising filtering is discussed. The filtering can be aimed at not only the complex image, but also the phase image. A terrain adaptive interferogram filtering method is proposed based on the classical adaptive filtering algorithm. The new method can suppress noise and keep the useful information at dense fringe areas in the same time.The author digs into various phase unwrapping algorithms and gives a thoroughly compare through analyses and simulation. These algorithms involve Branch-Cut Algorithm, Region-Growing Algorithm, Minimum Discontinuity Approach Algorithm, Least-Squares Algorithm, Minimum Lp -norm Algorithm and Network-Flow Algorithm. The thesis compares the effectivity, the operation time and memory requirement of the algorithms. Because of the high operation efficiency and the more valuable project application, the network-flow algorithm is studied more deeply. But there are some disadvantages in the algorithm. The author remedies these flaws. The improved algorithm can lead to more accurate unwrapped solutions.The traditional single-baseline InSAS offers only measure capability, and has not resolution capability yet. At the same time, the phase unwrapping problem is very difficult i...