Research On 3D Tracking And Registration Technology Of Aero-engine Oriented To Mixed Reality

As an emerging technology,mixed reality can provide users with an immersive experience that seamlessly connects virtual and reality,and the application of this technology can improve the overall production efficiency of the industry to a certain extent.However,the application research of this technology in the domestic aero-engine maintenance and training field is still in its infancy,and the three-dimensional tracking and registration of the engine model is the key factor restricting its development.This thesis focuses on the problems existing in 3D tracking and registration of engines,and conducts research work with the goal of integrating virtual and reality to realize mixed reality scenarios.This thesis firstly proposes a computer vision-based solution of registering the model of the engine entity and then using the algorithm to track the camera trajectory.And according to the experimental conditions,the technical routes of model registration based on QR code,pose tracking based on SLAM algorithm,and mixed reality effect verification using Hololens2 are determined.The research content and results of the thesis are as follows:1.With the help of the engine manual and 3D scanning technology,the 3D modeling of the appearance of a certain aero engine was completed,and the Unity package of the engine model was developed on this basis.2.For the spatial positioning of large objects,a solution of using three QR codes for positioning is proposed,which realizes the initial registration of the entity by the three-dimensional model of the engine.3.Optimized the RGB-D image data of the engine model tracking algorithm,deleted the pixels without distance information and created a new data set.Based on the optimized RGB-D data,the original SLAM image feature point extraction and matching algorithm was improved.The ORB key point detection algorithm was combined with the BRISK key point description algorithm,which improved the accuracy of image matching and reduces the time-consuming of the algorithm.Secondly,according to the characteristics of the engine maintenance training scene,the screening criteria of the matching feature points of the original algorithm were optimized,the number of successfully matched feature points was increased,and the number of feature matching point pairs was guaranteed for the back-end optimization.The camera trajectory solved by the improved SLAM algorithm and the original algorithm was compared,and the error analysis was carried out with the actual trajectory.The final result showed that the improved SLAM trajectory estimation accuracy was better than the original algorithm in the engine scene.4.The design of the virtual-real fusion system based on the Hololens2 aero-engine was studied.The system solved the virtual-real fusion problem of the aero-engine and the 3D model within a reasonable error range.It was of great significance to the application of mixed reality technology in the field of aero-engine maintenance and training.