Research On Key Technology Of Mobile Augmented Reality

Recently,augmented reality is getting more and more attention.People are gradually realizing that AR may replace the mobile platform as the future products which consumers care about.With the development of hardware technology and the popularity of mobile devices,the application scene of augmented reality is more extensive.Therefore,mobile augmented reality will become the mainstream development trend.Solving the real-time problem and improving performance have become a focus in current mobile augmented reality research.This thesis works on mobile augmented reality,including the image feature matching,the target object tracking and registered information displaying.A sensor-based feature descriptor RBRIEF(Rotational-aware Binary Robust Independent Elementary Features)is proposed.Due to the high computing time and large hardware space required by traditional image feature matching algorithms,mobile devices can not meet the real-time requirement.Besides,these descriptors do not use hardware features of mobile devices.So the descriptor in this thesis introduces the inertial sensor data to make it rotational-aware.Because it is a binary descriptor,its described and matching process takes less time and less memory.An improved Efficient Second-order Minimization tracking method is proposed.The traditional tracking method do not consider the motion blur,and requires high-performance hardware.In this regard,the algorithm in this paper introduces a motion blur model.During the tracking process,the template images are divided into sub-grids,and the high-gradient sub-grids are selected for tracking.Experimental results show that this tracking method reduces the computational load and improves the real-time performance.A brightness modulation algorithm is proposed to improve the saliency of foreground images.As a actual environment changing,the contrast of foreground registered images may decrease.This situation results in recognizing difficultly.in this paper,the saliency of image information that needs to be registered is calculated.The contrast of the virtual information is adjusted only by discrete brightness levels to reduce the amount of computation.For a given augmentation object,the chromatism is carried out using a representative pixel.