Research On Optical Flow Based Mobile Terminal Markless Augmented Reality

At present,the research of Augmented Reality technology in mobile terminal is still in the initial stage,compared with the implementation of the PC terminal,the mobile terminal has weaker computation ability.How to use the limited resources of mobile devices to realize the efficient operation of augmented reality system is an urgent problem to solve.Registration by tracking is the kernel module of the Augmented Reality system,this paper focuses on how to achieve stable and efficient motion tracking in mobile terminal.First,the overall framework of Augmented Reality is introduced,and the usage scenarios and function of the tracking registration in the system are illustrated,Then,a feasible tracking registration method suitable for mobile terminal is proposed.Finally,in order to realize the efficient and stable markless tracking registration on mobile platform,this paper has mainly done the following works:(1)The main features tracking algorithm of tracking registration technology are analyzed,and the performance of the algorithm is analyzed from two stages of feature extraction and matching.Although the feature points extracted by FIST and SURF algorithm are multiple and stable,the computational complexity is high.For the Augmented Reality system,real-time tracking is the guarantee of accurate registration of virtual information,and feature extraction and tracking is only the most basic part of the system,after which the pose calculation,virtual information rendering are needed.Under comprehensive consideration,the FAST algorithm based on accelerated segmentation is selected for feature extraction.While FAST algorithm can only extract feature points but cannot track features.Considering the sparse optical flow method has stable tracking function to the specific points of the sequence images,LK optical flow is used to track the FAST features extracted before,and good results are obtained.(2)Aiming at the problem of poor tracking of large displacement and light variation in motion tracking based on LK optical flow,we deeply study on how to make optical flow tracing more stable and proposed a diamond template improved Pyramid Lucas-Kanade optical flow algorithm.Diamond template is used to correct the error of optical flow calculation and then combined with Gauss Pyramid camera tracking the large displacement of camera.Through theoretical analysis and experimental comparison,it is confirmed that the improved Pyramid Lucas-Kanade optical flow algorithm is suitable for the camera motion tracking of non ID Augmented Reality.(3)Finally,a mobile mark-less augmented reality system is designed and implemented,which uses the multithreading support of Android to put the time-consuming tracking operation on the single thread and accelerate the efficiency of the system,the direction sensor and acceleration sensor are introduced to realize the hybrid tracking.In the end of this work,the system implementation in the real machine is illustrated,which verifies the method proposed in this paper has real-time performance,accuracy and stability.