a pa pr m s
Feature Based Pattern Recognition - Video Detection

Mixed multibeam (colored) and side-scan (grey scale) map of the study site showing the ROV tracks.



Still images containing anomalies, from the video.



With funding from the Army Pentagon Office for Environment, Safety and Occupational Health, Prometheus Scientists have used noncommutative group harmonic analysis to develop an anomaly detector to automatically detect discarded chemical munition canisters on the sea floor, in data gathered from an ROV-mounted video camera. Please note that linked video is 118 MB.

The resulting Prometheus Object Detection Software (PODS) significantly reduces video imagery review efforts. PODS is a fast, cost-efficient tool for reducing the data volume to image frames containing targets of interest. PODS can be used as an autonomous data reduction tool or as an interactive cuing tool. Input to the software is uncompressed red/green/blue (RGB) video imagery. The underlying mathematics is presented in the monograph Group Filters and Image Processing (Psypher Press, 2003) written by two Prometheus Principal Scientists, Myoung An and Richard Tolimieri.




© Prometheus Inc. All rights reserved. Web Privacy Statement