基于图像配准的微小型自主水下机器人视觉全局定位方法

Vision Global Positioning Method for Micro Autonomous Underwater Vehicle Based on Image Registration

随着科技的进步,微小型自主水下机器人将代替潜水员进行水下结构目视检查。本文介绍了一种基于图像配准的有效视觉同步定位与映射(SLAM)算法,可应用于水下结构物的自主检测。考虑到典型水下结构物表面的视觉特征稀疏,提出的视觉SLAM算法采用了一种由关键帧选择和关键对选择组成的选择性图像配准方案。通过仅使用潜在有效的图像和图像对进行基于特征的图像配准,与传统方法相比,视觉SLAM的计算负担可以大大减少。本算法在微小型自主水下机器人的实验结果验证了所提方法的实用可行性。

With the progress of science and technology, divers’ visual inspection of underwater structures is a very dangerous task,which is no longer suitable for the development of modern society. Therefore, it may become a potential application of micro autonomous underwater vehicles. This paper introduces an effective visual synchronous location and mapping(SLAM) algorithm based on image registration, which can be applied to the autonomous detection of underwater structures. Considering the sparsity of visual features on the surface of typical underwater structures, the proposed visual SLAM algorithm uses a selective image registration scheme consisting of key frame selection and key pair selection. By using only potentially valid images and image pairs for feature based on image registration, the computational burden of visual SLAM can be greatly reduced compared with traditional methods.The experimental results of the proposed algorithm in a micro autonomous underwater vehicle verify the practical feasibility and performance of the proposed method.