Planar Synthetic Aperture Sonar Generalized to Irregular Sampling Grids

Since its introduction in sonar (1975), the major application of synthetic aperture is in side-scan geometry (bottom imaging and mine detection) where azimuthal compression can only be achieved along the sonar trajectory (1D). More recently (2000), other scanning geometries (2D) have been proposed (circular or planar), in order to take advantage of SAS processing in two directions, in particular for object identification or buried objects detection. In previous works, we have shown the possible extension of SAS techniques to planar scanning (longitude, latitude and depth). P-SAS was applied to sub-bottom profiler geometry. The concept was first validated on simulation and tank data and then on sea data obtained in a dump site in the Baltic sea (SITAR project). Extending SAS to a planar geometry is faced to the errors on a trajectory both within a track and between the tracks. Due to the additional degree of freedom, conventional auto-focusing techniques and even trajectory correction techniques are quite difficult to implement because of the 3D structure of both raw and processed data. In previous works, for processing sea data, we have proposed to use oversampled $(\times 4)$ regular planar grids and to precompute focusing laws for such grids ahead of processing, and later on, to apply them to real data “re-positioned” on those grids. This, so-called Projective PSAS (PPSAS), consists thus in a rearranging of raw data followed by “conventional PSAS“ processing on a regular oversampled grid (with positioning error $ ). Repositioning is achieved thanks to the use of a long baseline for positioning the sonar system. In this paper, we propose to use the same navigation data and compute, for every case, the actual focusing delays required for planar synthesis in a generalized geometry, Generalized PSAS(GPSAS). Both methods (PPSAS and GPSAS) will be compared on simulation and tank experimental data (where all parameters and geometries are fully controlled). The complexity of both processing approaches is presented. • PPSAS ○ Pros: the focusing delays are pre-computed once for all and data is rearranged to fit on existing grid; PSAS can be split into SAS along one axis followed by SAS along the other axis (can be achieved partly in real-time along one direction). ○ Cons: spatial oversampling is required; part of the data whose position is too far from the ideal position is not used. • GPSAS: ○ Pros: oversampling is not required; all data can be used. ○ Cons: focusing delays must be computed for each geometry taking into account the actual trajectories in 3D; PSAS processing cannot be simplified and must all be achieved after the end of acquisition. First the complexity of both solutions will be investigated in details and computation times will be compared on the same data sets. Then the output performances will be compared, in terms of main lobe resolution and side lobes level. Several simulations using different configurations (single scatterer, multiple scatterers) and different mismatches (between ideal and actual geometries) will be presented. Simulations show that PPSAS can be used for a rapid, “lower quality” imaging (partly in real-time along track) then “suspicious areas” can be processed in details with GPSAS for higher resolution lower side-lobe (but slower) reconstruction of buried objects.