Performance Evaluation of a Low-Cost Stereo Vision System for Underwater Object Detection

Abstract

This paper describes the development and performance assessment of a low-cost stereo vision system for underwater object detection. The system has been conceived as a prototype to investigate the performance, power consumption, and thermal dissipation tradeoffs involved in designing an embedded stereo vision unit for underwater operation. The embedded system has been experimentally assessed in underwater object detection tasks. The system has proven thermally stable and capable of guaranteeing a level of autonomy of at least two hours of video acquisition. Several algorithms for mono and stereo image processing have been evaluated to assess their effectiveness in the underwater environment along with their suitability in presence of constrained computational and energy resources. Evaluation of the stereo vision system in detecting simple objects has shown strong limitations of commodity off-the-shelf sensors when used in underwater perception. Nevertheless, the prototype described in this work provides insights for development of more advanced vision systems suitable for underwater vehicles.