Research And Implementation Of Virtual Camera Locating Based On Visual Inertial Locating

In recent years,with the improvement of computer hardware,research on virtual reality,unmanned aerial vehicle,self-driving cars and other fields have come to the era of rapid development.As one of the core issues in these areas,the location problem is attracting more and more attention.Vision based localization is widely used because of its high efficiency and accuracy,but its dependence on visual characteristics makes it unable to work properly under conditions of missing environmental features and moving objects too fast.The inertial measurement unit can make up for the problems existing in the vision location method.In this paper,a location method based on monocular vision and inertial measurement unit is designed,and a virtual camera location system is implemented by this method.The smart phone is the client of the system,and the computer is the server side of the system.We use the single camera image acquisition as a visual input and use the visual odometry based on feature point to estimate the position and rotation of the camera.At the same time,we improved the storage mode of bag-of-words dictionary in the closed-loop detection to improve the speed of the dictionary loading.On the other hand,we use the data collected from the inertial measurement unit on the smart phone as input and implement two attitude calculation methods of integrating gyroscopes and accelerometers data based on complementary filtering and gradient descent.Finally,we propose a loosely coupled pose fusion method.The vision location part is combined with the inertial measurement unit to get the final pose,and we use the pose to control the camera in the virtual scene.At the same time,based on the C/S mode,the vision inertial location system is implemented on the client and server side,and data transmission is done through TCP,which ensures that the method in our paper can run on smart phone in real time.Finally,through the experimental analysis and the operation in real scene,it verifies that the method and system in our paper can satisfy the requirement of real-time and effectiveness for the virtual camera location problem.We also provide a new idea for low-cost mobile phone virtual reality or augmented reality system.