Constructing and visualizing three-dimensional sea bottom models to test AUV machine vision systems

Abstract

This paper describes an algorithm for constructing a procedural sea bottom model, which can be used for testing and debugging machine vision systems of autonomous underwater vehicles (AUVs). The algorithm consists of three main stages: generating a low-frequency heightmap (used by the designer to define the basic form of a water area), constructing a three-dimensional model (based on the heightmap and fractal noise), and visualizing the three-dimensional model (refined by means of hardware or manual tessellation). The sea bottom model has the following features: it is detailed accurate to a screen pixel, each of its sections is absolutely unique, and its size is adequate for any tests.