Director of Research

Ocean Imaging Informatics

Many oceanographic projects at WHOI work with still images and moving images. Projects like HABCAM, SeaBed, and the National Deep Submergence Facility, and the Marine Mammal Laboratory hold terabytes of images. Current research in ocean imaging informatics includes a project funded by Gordon and Betty Moore Foundation.

Rapid advances in imaging technologies – from nanoscale imaging sensors to earth observation satellites to characterizing 1000km of the seafloor – coupled with the equally rapid progress in digital storage and transmission technologies, have resulted in an exponential growth in image data. The field of Image Informatics provides the tools necessary to extract relevant information from images based on end user requirements. Steps in extraction of information from images include segmentation, object detection, tracking, pattern recognition, supervised and unsupervised classification, machine learning and datamining.

Areas of interest for Ocean Imaging Informatics include: Multi-scale imaging and image informatics of the seafloor using hardware mounted on towed, AUV, and ROV platforms. Stereo pair imaging is required for accurate measurements in 3-dimensions. Real-time mosaicking is required to establish registration of contiguous images and nearest neighbor analysis of biological and geological components. Image processing must be done in real time to allow immediate response to events and to avoid becoming swamped in unprocessed data. Real-time products include color correction based on local spectral characteristics of the water, segmentation of targets from background substrate, texture analysis of substrate, feature extraction and classification of targets, statistical representation of spatial distributions, and feedback to the imaging platform to allow for adaptive sampling. Other requirements include hardware implementation of processing code, miniatureization and high density integration of imaging and processing hardware, archiving, accessing and visualization of very large GIS encoded image data sets.