An Underwater Micro Cable-Driven Pan-Tilt Binocular Vision System With Spherical Refraction Calibration

Abstract

To improve the operation performance of underwater robots, a novel micro cable-driven pan-tilt camera system is developed, which can play an important role in underwater environmental perception and visual measurement. The designed camera system encompasses three parts: a miniature mechanical structure, a real-time target tracking and location module, and an underwater spherical refraction calibration module. Originally, a cable-driven model is adopted to establish a micro and flexible pan-tilt structure, for which the corresponding active target tracking and location algorithms are proposed. More importantly, for eliminating the measurement error caused by the underwater spherical refraction, an advanced nondominated sorting genetic algorithm (NSGA-II)-based underwater calibration algorithm is presented with the aid of the comprehensive geometric spherical refraction model. By integrating the flexible pan-tilt camera and the proposed calibration algorithm, the performance of underwater environmental perception and visual measurement can be greatly improved. Finally, extensive experiments in air and underwater scenarios are conducted. According to the utilized evaluation index, the deviation of target tracking and underwater location errors can be maintained within 0.06 rad and 21 mm, respectively. The proposed vision system sheds light on precise perception and interaction tasks of underwater vehicles.