The Application Of Dynamic Viewpoint Technology With Multiple Screens In Helicopter Simulator

Helicopter equipment is rapidly evolving in military and civilian applications with its flexible maneuvering characteristics,terrain-free vertical take-off and landing,and the unique advantages of hovering in the air.However,the dynamic coupling of the helicopters in various directions is serious,and the operation is difficult.In the occasions where the working space is small,the "ship effect" and the "steep wall effect" are likely to occur,which make the training task not fully suitable for the real field training.In recent years,with the rapid development of virtual reality technology,the helicopter simulator has become a new way to train drivers in professional fields.It not only trains the pilot's conventional helicopter maneuvering skills,but also simulates special flight situations such as special training and extreme weather,serious aircraft failures,and has the advantages of safety,economy and efficiency.Based on the BELL206 helicopter flight simulator platform built by the research group,this paper deeply studies the dynamic viewpoint technology with multi-screen view stitching to improve the visual presence of the simulator.The helicopter simulator screen display unit is arranged in a curved shape by using a plurality of upper and lower screens for splicing that can provde large angle view of virtual environment.Dynamic view technology is improved over fixed viewpoint technology.Fixed viewpoint technology means once the geometrical dimensions and the placement position of the display screen are determined,the parameters of the viewpoint projection matrix in the virtual scene remain unchanged during the simulation.The dynamic viewpoint technology uses sensor technology to track the head position information of the driving pilot in real time,which is used to dynamically update the projection matrix of the viewpoint in the virtual scene.The imaging content and imaging effect are similar to the effects when the pilot observe the external environment through the cockpit window.In this paper,the point cloud data of the pilot's head is collected by the Kinect sensor in a non-contact manner,and the posture information of the pilot's head is estimated in real time.At the same time,use two Kinect to solve the self-occlusion problem of the head.The main work of the paper is as follows:(1)Built a multi-screen dynamic view system overall framework.Introduced the structure of the helicopter simulator platform,analyzed the selection principle of the RGB-D camera,calibrated of the internal and external parameters of the Kinect camera which were used to unify the Kinect coordinate system,lastly gave an effective head point cloud filtering algorithm.(2)Designed head pose estimation algorithm which was used to implement dynamic viewpoint technology.Firstly,a random forest algorithm with high real-time and large recognition range was selected for rough estimation,and finished feature selecting and trained classification regression decision tree.Secondly,the point cloud registration scheme was used for accurate estimation,which referred to register the point cloud collected by the Kinect camera to a predefined head face template.Improved the strategy in the nearest point selection stage,optimized the cost loss function,and gave an approximate solution of the rotation matrix.Finally,an interframe real-time tracking system based on Kalman filter was proposed,which could quickly located the head region in the next frame depth image.It helped advance the fps and accuracy of the original random forest algorithm,and provided a good initial transformation matrix for accurate estimation.(3)Used eight high-definition LCD screens to achieve a low-cost multi-screen splicing solution compared to expensive dome projection and achieved a high degree view of virtual environmen.Based on the fixed-view multi-screen splicing scheme,the global coordinate system modeling was used to update the attributes of the virtual viewpoints which was the keypoint in implementing dynamic viewpoint.At the same time,based on Accela filtering,the stability of the projection parameters was enhanced,and the jitter that may occur in multi-screen splicing was removed.(4)Built a multi-screen dynamic viewpoint system platform and did experimental analysis.The acquisition of the head position parameter was the core of the dynamic viewpoint technology,so the rough estimation algorithm based on Kalman filter enhancement,the accurate estimation algorithm of point cloud registration,the algorithm workspace and the running speed of different algorithms were firstly tested and effected.Then the large field view of the simulator was realized by testing the multi-screen splicing technology.Lastly the projection parameter stability enhancement algorithm based on Accela filtering was analyzed experimentally.