Automatic Change Detection (ACD) compares new and stored terrain images for alerting to changes occurring over time. ACD techniques, long used in airborne radar applications, are just beginning to be applied to sidescan sonar. In Coherent Change Detection (CCD) the cross-correlation of multi-temporal complex data collected from coherent imaging sonars detects changes in the transduced amplitudes and phase of image pixels which, under the right conditions, can be used to detect new objects or disturbances on the seafloor. Synthetic aperture sonars (SAS) produce range-independent, fine resolution seafloor images. With centimetric resolution demonstrated out to hundreds of meters, these coherent systems can classify small manmade objects at long ranges, and should be suitable for CCD. This paper describes experiments testing CCD with data from synthetic aperture sonars mounted on autonomous undersea vehicles and actively navigated tow bodies. A noncoherent example carried out with data collected from an AUV-mounted SAS demonstrates the utility of correlation-based automatic change detection. CCD tests were carried out with repeat pass data collected using a SAS mounted on a dynamically controlled tow vehicle. While simple image pair co-registration procedures failed to provide sufficient coherence in the overall scene required for CCD, preliminary tests of image warping techniques used for airborne radar applications show promise of transitioning successfully into the SAS signal processing chain.
展开▼
