Classification of underwater color images with applications in the monitoring of deep corals reefs

Coral Reefs ecosystems have been impacted by natural and anthropogenic effects resulting in a decline of coral communities worldwide. This decline in coral reefs has an ecological and an economical impact in tourist areas and marine ecosystems due to beach activities, scuba diving, and fishing. Monitoring of coral reefs is made by marine biologists using traditionally diving techniques. The main purpose of this work was to develop a classification algorithm for digital benthic images with applications in the monitoring of deep coral reefs in order to calculate the percent of living coral area on the sea bed. At depths beyond approximately 30 meters the absorption and scattering properties of the water do not allow the use of remote sensing. In these cases, other imaging platforms, such as, autonomous underwater vehicles (AUV) are needed. The AUV images present objects with greater spatial details. The classification challenges, however, arise from the common color statistics, non-uniform illumination, and the high concentration of noise introduced by the attenuation and scattering of the light by the water column. The Image-Spatial-Coefficients-Classification-Algorithm (ISCCA) developed during this research obtained constant overall classification accuracy over 87%. The classification algorithm combines segmentation, color and texture to perform the image discrimination. This automated classification system will replace many hours of manual photo interpretation by a marine biologist involved in corals studies.