Development of Automated Tools for Underwater Photographic Analysis

As part of the ongoing collaboration between the Woods Hole Oceanographic Institution and the PIFSC and the NWFSC we have developed an AUV for conducting high resolution optical imaging surveys. This platform is now routinely collecting thousands of images per dive. This proposal is a first attempt to standardize and develop the tools required to deal with such datasets within the constraints and requirements of the end users. The emphasis for this particular project will be on basic research in the area of automated analysis and classification. Within the context of the standard tools already in development, this project will focus on the use of principal component analysis and the Fisher linear discriminant for automatically extracting and classifying organisms of interest within the AUV imagery. In addition it will investigate the use of Varma-Zisserman method for background habitat classification.