Photo-identification of marine mammals using curve and patch-matching techniques

Individual identification of marine mammals is of major interest to marine mammal biologists. This work aims at recognition of individual marine mammals using edge and patch information in their images. The images are captured under various geometric transformations and hence the identification strategy involves using invariants under an appropriate group of transformations. The developed identification or curve/parch matching techniques have been applied to databases of dolphins, gray whales, humpback whales, and sea lions. Performance results show that these matching techniques considerably reduce the number of images to be searched manually and hence are quite useful to marine biologists in their identification studies.