深空探测巡视器视觉导航定位方法研究与地面试验场验证

Research and ground validation based on the visual positioning methods for rover in deep space exploration

地外天体表面巡视器就位探测是深空探测研究的重要手段,其自主导航定位能力是任务开展的重要前提。本文主要对视觉导航方法中的视觉同时定位与地图构建(SLAM)方案开展应用研究。介绍了当下经典的视觉SLAM框架,并以具有代表性的ORB-SLAM3算法为例开展研究;分别利用公开数据集、深空探测地面模拟场数据进行试验验证,探索视觉导航定位算法在深空探测场景下的有效性。本文研究为我国后续开展长距离连续巡视任务提供参考。

Rovers are commonly used in deep space exploration missions and it is important to develop the ability for autonomous navigation and positioning of rovers.In this paper,a typical visual simultaneous localization and mapping(SLAM)technique for navigation and positioning is studied.Firstly,the general visual-SLAM framework is summarized and the well-known ORB-SLAM3 method is introduced to evaluate its effectiveness for the deep space exploration missions.In addition to the public datasets,a simulated ground test field for extraterrestrial environment is built and the corresponding datasets are further used to validate the possible application for visual-SLAM methods in Lunar and deep space explorations.This study can help to expand the applications of the state of art SLAM methods and also support the following long-distance rover exploration tasks for China's missions.