| Popis: |
Synthetic aperture sonar (SAS) can produce images with centimetre-level resolution and area coverage of better than one square kilometer per hour. This makes SAS an ideal sensor for detection and classification of small targets over large areas. Fully automated target analysis allows improved autonomy when using autonomous underwater vehicles (AUVs) and saves a tedious manual analysis in post-mission analysis. Recognition of small targets in sonar imagery is, however, a difficult task. SAS imagery preserves wavenumber information. This gives the possibility for extra products in addition to high resolution imagery. We propose a two-stage processing where regions of interest are generated from reduced resolution SAS imagery and subsequently post processed images are used to generate relevant target analysis information. In this paper, we concentrate on the types of information available and their significance rather than the choice of intermediate resolution and initial detection methods. The extra processing products discussed in this paper are target-enhanced images by autofocus, shadow-enhanced images by fixed focusing, multi-aspect images, frequency-selective information and 3D shape from interferometry. We show examples of each of the additional products using data collected by the HISAS 1030 interferometric SAS carried by the HUGIN 1000-MR vehicle. |
