Processing and Analysis of Underwater Acoustic Images Generated by Mechanically-Scanned Sonar Systems

The processing and analysis of images generated by mechanically scanned sonar systems have received poor attention despite their widespread application. In this paper, some efficient methods for acoustic image enhancement and automatic object detection are presented and assessed using a large set of experimental data collected at sea with commercial sonar systems. Specifically, a set of methods for increasing the quality of the gray-level images produced by a fan-shaped-beam sonar is introduced. Such a set includes a dynamic brightness assignment, a precise interpolation, a speckle-reduction filter, and a contrast-enhancement block. Two versions of a template-matching-based method that allows the automatic detection of a simple object contained in a region scanned with a pencil-beam sonar are also proposed and assessed. The main difficulty to be coped with in this field is related to the sparseness of the binary maps generated by this sonar system. The performance and robustness of the proposed techniques have been evaluated using real data that provided satisfactory results for both the image-enhancement and the object-detection tasks. Moreover, the computational burden of most of the proposed techniques turned out to be quite limited, and their real-time implementation with a standard computer architecture could be estimated.