Automatic mine detection by textural analysis of COTS sidescan sonar imagery
Abstract:Sidescan sonar imagery of the sea floor is difficult to interpret visually, and classification techniques are now increasingly used to supplement the interpreter with reliable, quantitative results. Active high-resolution sonar has been shown to be the only sensor able to detect man-made objects and, in particular, mine-like objects on the sea bottom with a high probability of detection and an acceptably low false-alarm rate and a high area coverage rate. The actual distinction between the image of a mine and an object that physically resembles a mine is very complex, however, and relies on the recognition of subtle differences in shapes and textures. This study aimed to combine two different advances in sidescan sonar applications. Recent trials at sea have demonstrated that the latest generation of commercial off-the-shelf sidescan sonars was able to image mine types, even stealthy mines with low acoustic signatures. Concurrently, a method of advanced image analysis has been developed, based on the quantification and recognition of acoustic textures. This method has been extensively calibrated and ground-truthed in complex terrains, and the results presented here show that it can be applied successfully to the detection of mines.
Document Type: Research Article
Affiliations: Department of Physics, University of Bath, Claverton Down, Bath BA2 7AY, England, UK
Publication date: November 10, 2000