Implementation Of Visual Inertial Monocular SLAM In Mobile Phone

With the development of artificial intelligence technology,mobile augmented reality technology has been widely concerned by researchers.In order to realize the combination of virtual information and real information in the mobile terminal,it is necessary not only to let the mobile terminal understand the structure of the real world visually,but also to let the mobile terminal system understand its own location.The key technology needed is slam technology.In recent years,almost all mobile phones have been equipped with cameras and IMU sensors,which provides hardware support for the application of visual inertial slam technology in mobile phones and is highly feasible.However,there are many difficulties in the implementation of visual inertial monocular slam in mobile phone.The main work of this thesis is to analyze and study the monocular slam system.Aiming at the application of monocular slam technology in mobile phone,the visual monocular slam is combined with IMU,and the visual inertial monocular slam system and algorithm are optimized and improved.Finally,the optimized system and augmented reality engine are combined to realize real-time camera positioning and space reconstruction in mobile phone.In this thesis,the realization of visual inertial monocular slam in mobile phone contains the following work:The first is the research and improvement of the basic SLAM algorithm,which is mainly divided into three parts:front-end initialization optimization,front-end visual odometer speed optimization and back-end word bag optimization.The second is the transplantation of the system to Android mobile phone,which is mainly divided into three parts:encapsulating the optimized slam system and writing JNI interface,compiling NDK script and porting the third-party library.The third is the design and implementation of augmented reality display module,which is mainly divided into two parts:the design of augmented reality module based on unity and the combination of slam system and augmented reality engine after optimization.Finally,the camera real-time positioning and spatial reconstruction of vision inertial monocular slam system are realized on the mobile phone,and the real-time performance,adaptability and reliability of SLAM algorithm are effectively improved.