MEASURING ARTIFICIAL REEFS USING A MULTI-CAMERA SYSTEM FOR UNMANNED UNDERWATER VEHICLES
Keywords: Underwater Photogrammetry, ROV, Multi-Camera System, Automated Image Masking, Transfer Learning, Convolutional Neural Network
Abstract. Artificial reefs provide an efficient way to improve marine life abundance in the oceans, including growth on the structure itself. Photogrammetric methods provide suitable tools to measure marine growth. This paper focusses on cubic reefs placed in Western Australia. The capturing platform featured a photogrammetric multi-sensor system for unmanned underwater vehicles attached to a low-cost vehicle BlueROV2. The multi-sensor system and its photogrammetric data captured was calibrated, adjusted and analyzed employing a structure-from-motion processing pipeline. Novel automated image masking techniques were developed and applied to the data to significantly reduce noise in the derived dense point clouds. Results show improvements of signal to noise ratio of more than 50 %, while maintaining a complete representation of the observed artificial reef.