Supervised Classification of Underwater Optical Imagery for Improved Detection and Characterization of Underwater Military Munitions
Abstract
Optical images of the seabed could benefit surveys for underwater military munitions (UWMM). Due to the need for human interpretation, however, analysis is currently a bottleneck for quantitative assessment of underwater images. In this project, a recently developed image classification algorithm was tested for identifying UWMM and seabed types. Also, an extension to the algorithm using seabed microtopography, or roughness, features was developed and tested. The image classifier by itself distinguished munitions from non-munitions (background) with generally > 80% accuracy. Discrimination of environments was high for the major seabed types. For example, sand and mixed sand-seagrass were classified with 80-100% accuracy in both shallow and deep water. Extending the algorithm to also use height data derived from stereo reconstruction greatly improved the classification results. Improved accuracy with the height features was observed not only on the basic, binary munitions / non-munitions classes, but also improved the capability to discriminate different types of munitions from one another.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jun 01, 2015
- Accession Number
- ADA624851
Entities
People
- A. Shihavuddin
- Arthur Gleason
- Brooke Gintert
- Greg Schultz
- Nuno Gracias
Organizations
- University of Miami