一种用于中国空间站增强现实维护应用的自动跟踪框架OACSTPCD

Tracking registration is a critical technology of augmented reality(AR).Marker-based and image-based registration methods have been widely used in augmented assembly systems.However,the space station encounters distinctive challenges due to weak product texture,symmetrical structure,limited data,and difficulty in attaching markers.This paper presents a novel 3D object tracking method specifically designed for induced maintenance applications in the Chinese space station,aiming to replace traditional paper manuals with the AR technology to provide astronauts with more intuitive operational guidance.We propose a marker-less approach for intelligent maintenance of the Chinese space station.A point pair feature method that combines curvature information for improving efficiency is employed to estimate the initial frame pose,obviating the need for manual adjustment to achieve the corresponding pose.This is crucial for astronauts,as precise movement in the microgravity environment is a significant challenge.Precise restrictions on the position of astronauts are eliminated,and the tracking robustness of space symmetric products is further enhanced by incorporating both texture and region information.The method utilizes both point cloud and 2D image.It leverages point cloud matching to estimate the initial pose of the first frame and recalculates the pose after loss of tracking.Once the pose of the previous frame is obtained,the tracking is calculated solely according to the region and texture information of the 2D image to obtain real-time tracking.The experimental results show that our method has the same pose trend as the marked-based method.And the error difference between the measurements based on electronic vernier calipers is in the millimeter level.The successful replacement of filters using specialized tools designed specifically for space products demonstrates the practicality and potential of implementing induced maintenance procedures on space station.