Research On Key Technologies Of Aviation Medical Simulation Training Based On Mixed Reality

Various hidden dangers and risks of accidents in my country overlap,and it is urgent to strengthen emergency medical rescue capabilities.It is imminent to establish an aviation medical emergency rescue system with helicopter as the main carrier suitable for national conditions.In recent years,Mixed Reality,mixed reality(MR)technology has advanced by leaps and bounds,and mixed reality technology has entered many industries,and has developed the most rapidly in the medical field.At the same time,the development of mixed reality technology has also accelerated the replacement of medical-related mobile devices,aviation rescue and medical training.Aspects and mixed reality technology also give people hope.This paper mainly studies the key technology of aviation medical simulation training-virtual and real registration technology.This technology is mainly divided into two types,point cloud registration and vision-based marker registration.In this regard,this paper verifies two virtual and real registration methods.On the basis of predecessors,an innovative virtual-real registration algorithm based on markers is proposed.It can meet the high dynamic environment and real-time performance required for simulation training.The main work is as follows:(1)first of all,this paper tries to use the point cloud registration way to complete the actual registration,by studying the actual registration of the spatial positioning technology,and monocular vision based on RGB SLAM technology,design the unmarked based on point cloud registration process of experiment,by comparing the characteristics of the local library and ICP algorithm were used to detect feature matching,The experimental results show that the unlabeled ICP algorithm can match the virtual model and the real object well,but the real-time tracking effect is not good.(2)In order to solve the problem of poor real-time tracking effect,the matching algorithm based on markers in virtual and real registration is studied.The QR code with strong fault tolerance is selected as the marker code.According to the target detection network YOLO V4 and YOLO V4-TINY network structure,the lightweight network suitable for transplantation to Holo Lens2 mobile terminal is found.An innovative algorithm is proposed to improve YOLO V4-TINY,and two CSPBlock modules in backbone network are replaced by Res Block-D module with less computation,so as to further reduce computation and increase detection and tracking speed.Finally,it is concluded that the model capacity is reduced by 13.1%,the data capacity is further reduced by 13%,the data volume is significantly reduced by 24%,and the complexity is greatly reduced,which solves the problem of low computing power of the mobile terminal and makes it more suitable for the experimental device Holo Lens2.(3)Set up the experimental environment of the system,integrate the innovative algorithm proposed in this paper into Unity development tool,and finally realize the virtual reality registration based on markers.The virtual model can be detected and tracked with the real object.Compared with the registration based on point cloud mentioned above,this registration technology has better robustness,enabling trainees to better complete simulation training and solving the problem of dynamic registration and tracking in the aviation rescue environment.