Intelligent identification of ship types from visual geometric features in maritime safety projects

Abstract

This study addresses the enhancement of maritime navigational safety through intelligent ship type identification based on visual-geometric features, particularly under conditions of reduced visibility and dense traffic where human-factor risks increase. The research proposes a software approach that classifies vessels using structured descriptive or catalog data while ensuring interpretability of results in terms aligned with navigational expert reasoning. The objective is to achieve a stable and reproducible classification accuracy of approximately 65 – 70%, supporting its application in simulator training, maritime education, and decision-support systems, and contributing to the development of explainable AI tools in maritime safety contexts.