Automatic Classification of Incidents in Coastal Zones

Abstract

The increase of seasonal population on coastal areas, in certain periods of the year, as well as the diversity of events to be monitored by the authorities, raise the likelihood of marine incidents. These occurrences can have a diverse nature and severity amenable to ban the access to the beaches. This work describes the development of a low-cost system based on UAV for real time detection, recognition, and classification of several types of incidents in the coastal area and inland waters in an efficient manner. The system provides, to maritime authorities, a faster and more effective capacity for intervention in controlling maritime incidents, contributing to greater protection of public health and safety of the populations and the activities developed ashore. The system implemented a machine learning algorithm and a mobile app that help human operators monitoring maritime incidents. The development of the system was based on usability principles in order to tailor the system’s graphical interface to the first responder’s users (e.g. lifeguards, coastguard officers).