Visual features extraction and types classification of seabed sediments

摘要

The purpose of this research is to define and extract the visual features of the seabed sediments to improve the autonomous ability of a underwater vehicle while implementing exploring missions. A scheme of seabed image classification is proposed to identify three types of seabed sediments. The texture features of images are stable and robust visual features in underwater environment comparing with general visual features, and which are described by using gray-level co-occurrence matrix and fractal dimension. Subsequently, for purpose of evaluation, a supervised non-parametric statistical learning technique, support vector machines (SVMs), is applied to verify the availability of extracted texture features on seabed sediments classification. The presented results of seabed type recognition justify the proposed features extracted method valid to seabed type recognition.