Commercial marine aquaculture operators face many operational hazards including disease, predators, husbandry operations, and environmental changes. Most of these risks are only identified with constant surveillance and physical presence at a farm site. However, human observation of risk factors is expensive, slow, and sometimes ineffective. Sensors are available to monitor individual environmental parameters, but comprehensive monitoring of all operational risks is currently infeasible or cost-prohibitive. This project seeks to develop a single, inexpensive tool, CERBERUS (Camera-based Examination of Risk via Behavioral Evaluation with Remote Underwater Surveillance), to detect and alert operators to the presence of multiple types of operational hazards through the use of low-cost hardware and intelligent software processing. CERBERUS will enable fish farmers to remotely and automatically monitor their stock for responses to such hazards, helping them reduce reaction time in rectifying the causal issues, improve outcomes, and decrease overall operational risk. SUMMARY OF
Anticipated Results: Phase II will build upon the advances made in Phase I and result in the development a cloudbased, computer vision framework which will facilitate the acquisition, segmentation, and classification of continuous, real-time video, rapid training and testing of neural network models, and the management of the entire process from a web-enabled dashboard.
