Automated contact calling visual aid using sequential mathematical processes: Using textural analysis for mine-like contact detection

Abstract

This paper presents an automated process for finding mine-like contacts by incorporating sequential mathematical processes: textural analysis, polygon discrimination, and dimensional filtering. These processes define a methodology that requires a minimal amount of user input to visually aid the side scan interpreter. The initial focus of this project was to find mine-like objects in high clutter density areas while maintaining a high probability of detection. In order to meet this requirement, a methodology had to be established to combine multiple analytic algorithms. There was also a concerted effort to create a methodology that could be used on a variety of side scan sonar systems. However, in order to discriminate contacts on multiple sonar systems, certain system specifications would have to be incorporated into the process. The key differences between side scan sonar systems necessary for this study are the optimum operating altitude and the along and across track resolutions of the imagery. Since altitude dictates the pixel area encompassed by contacts and resolution determines distortion of the contacts in the imagery, the iterating texture measurement window was built around these two concepts. Implementing a polygon discrimination algorithm became necessary to locate and find the centroid of polygons in the imagery that met textural analysis threshold requirements for a mine-like contact. The application of these methods has shown the potential for an automated process capable of ingesting different types of side scan sonar data with minimal user input.