Precise geopositioning of marine mammals using stereo photogrammetry

Precise positioning of whales and other species in space and time is a key requirement for marine mammal research. It has been an elusive goal for years. We have developed a stereo camera based measurement system to meet the requirement. We have obtained preliminary results, and will describe ongoing improvements. The sounds of marine animals can be localized using multiple hydrophones. If these hydrophones are part of tags (like the DTAG) attached to individual animals, sometimes it is possible to identify which call is made by which individual. However, when social animals like pilot whales are very close together, it becomes very difficult to identify which individual is vocalizing. This is a critical problem for studies of marine mammal communication: we are not able to link the acoustic and tag data with the behavioral observations because we cannot accurately pinpoint where specific animals are in space, either on their own or relative to other animals. Precisely positioning the whales in space and time is also necessary to measure social cohesion, the critical variable for assessing the impact of anthropogenic sound on many vulnerable marine mammals. Current thought suggests that social whales, such as pilot whales, adopt a social defense strategy, grouping closer together under threat. Thus, of the dozens of sound and noise impact studies conducted on marine mammals throughout the world attempt to assess changes in cohesion during exposure to sound. However, they all estimate inter-animal distance by eye, something that is notoriously difficult and imprecise. In short, there is currently no accurate way to measure the fundamental variable that these millions of dollars of fieldwork are trying to assess. Positioning individual body parts instead of whole animals in space and time would allow precise mensuration of body part ratios, an essential statistic for assessing health and fat reserves that is currently difficult to measure in the field. Numerous techniques have been tried to address the geopositioning requirement, none have been wholly satisfactory. We developed a battery powered stereo camera system, integrated with a GPS receiver, an attitude reference system, and a laptop computer, and collected calibrated stereo imagery from a surface vessel. The stereo camera we used initially was an off the shelf firewire based system, originally intended for machine vision purposes. It was selected in part because of time pressures on development, and proved to have too short of a baseline for the precise work demanded by the scientific requirements. Other constraints of the off the shelf system made it difficult to accommodate lighting conditions in the bright marine environment, and we have since moved to a custom system. This custom system has many features in common with stereo systems we have developed for underwater use, shortening development time and testing. These common features include camera models and interface, calibration techniques and software elements, all of which will be described. Custom software has been developed for geopositioning of targets in the stereo overlap area. By differencing of positions of multiple targets, it becomes possible to achieve precise mensuration of body parts and sizes. These measurements can be made using both monoscopic viewing of two simultaneously collected images, or if three-dimensional viewing hardware is available, in stereo.