AMI: a 3-D imaging sonar for mine identification in turbid waters

Thomson Marconi Sonar is developing the Acoustic Mine Imaging (AMI) high frequency sonar (/spl lambda//spl sim/0.55 mm) for the Royal Australian Navy to assist in mine identification in turbid water. Phase 1 of the project developed an engineering prototype. This prototype was able to demonstrate that sparse ultrasonic hydrophone arrays could be used to form high-resolution images. However, this prototype was unable to process the images in real-time and was also too large to be readily fitted onto a remotely operated vehicle. In Phase 2 of the project the sonar will be miniaturized to integrate into the mine disposal vehicle on the RAN's minehunter coastal and real-time signal processing and a man-machine interface will be integrated into the existing systems onboard the minehunters. A new array based on a sparse random pattern has been developed with a higher hydrophone density than the prototype. This enables reduction in the array size while still achieving similar imaging performance. Several experiments were conducted on array patterns before selection of an optimized pattern based on performance and manufacturability. AMI uses searchlight transmitters to illuminate the target. Each transmitter is segmented so that the beam can be confined; this will reduce the impact from strong specular reflections upon image quality. Beam-forming and image forming in real-time has required the use of a number of approximations. This has been implemented on a highly parallel fixed-point processing chain in field programmable gate arrays (FPGA). The techniques were initially simulated using both artificial and trials data to validate the algorithm approximations. Initial testing has included validation of a slice of the processing chain. Image presentation takes advantage of the 3-D image produced by the sonar system to assist in identification of characteristics in images. This also enables the determination of feature dimensions. A number of image enhancement processes have been developed to significantly improve high-resolution images particularly where there is a poor signal-to-noise ratio.