| Synthetic Aperture Sonar (SAS) provides a large improvement in azimuth resolution compared with conventional side-scan sonar. This improvement is achieved by combining data from a number of pings to synthesise a long virtual receive array. Difficulties in forming a SAS image arise from the need for complex processing techniques and accurate knowledge of platform position. This thesis presents work to develop processing algorithms for forming images from raw SAS data. Methods of receiver calibration, motion estimation and imaging are presented, along with the development of a computer model for producing realistic simulated SAS data. The availability of a large amount of real SAS data to the project has allowed comparisons of the performance of techniques developed to be demonstrated using both simulated and real data sets. A new time domain imaging (inversion) method. Fast Factorised Back-Projection (FFBP), was recently proposed for processing Synthetic Aperture Radar (SAR) data. FFBP overcomes many of the problems associated with other inversion methods in that it can cope with general aperture geometry, is fast, and has very low memory requirements. The application of FFBP to SAS is presented, along with a number of modifications to the original method. A computer model for producing simulated SAS data is developed. This model is able to predict shadow in three dimensional scenes using ray tracing and represents realistic seabed textures using fractal surfaces. The program is used to verify the performance of the imaging algorithms developed. A technique for forming interferometric images is developed, based on radar interferometry. A new technique for combining height estimation and image co-registration is presented. |
