Improved supervised classification of underwater military munitions using height features derived from optical imagery

A supervised classification routine was used to classify munitions targets and basic seabed types from underwater images. Additional features that were based on the local relief, or height, of the seabed were then added to the classifier and new results computed using the expanded feature set. The height data were generated from the input images themselves using structure-from-motion computer vision techniques. The initial image classifier was shown to distinguish munitions from non-munitions (background) with generally > 80% accuracy except that many false positive matches for munitions were observed. Extending the algorithm to also use height data derived from stereo reconstruction showed that incorporating such “2.5-D” data greatly improved the classification results. Using the 2.5-D information reduced the number of false positives. Furthermore, improved accuracy was observed not only on the basic, binary munitions / non-munitions classes. Adding 2.5-D information also improved the capability to discriminate different types of munitions from one another.