Research On 3D Registration Algorithm In Mobile Scene Based On Augmented Reality

Augmented Reality Technology(AR)is a technology based on computer graphics and computer vision.It combines target tracking,three-dimensional registration and virtual reality to achieve the fusion and interaction between virtual objects and real scenes.When the transaction is not completed,AR technology can provide more intuitive and detailed virtual effects,bringing great convenience to people's production and life.In recent years,with the in-depth study of AR technology by scholars and the popularity of smart mobile devices,AR technology based on mobile scenes has become a key technology affecting the user experience of augmented reality.The motion of objects in a moving scene is more complicated,and the target object is prone to deformation and occlusion during motion.At the same time,real-time performance is also the key to augmented reality technology in mobile scenes.This paper focuses on the delay problem,target matching misalignment and even target loss in augmented reality applications in mobile scenes.An augmented reality three-dimensional registration algorithm for mobile scenes is proposed to improve this situation.Firstly,the TLD tracking algorithm is studied.Secondly,the traditional feature point detection matching algorithm used in the augmented reality technology is studied and compared.Then the research on the target tracking algorithm and the feature point matching algorithm is proposed.The augmented reality 3D registration algorithm is used to verify the robustness and real-time performance of the proposed algorithm in mobile scenarios.The main contents of this paper are as follows:1)Introducing and researching the traditional TLD tracking algorithm,mastering its working principle and process,aiming at the disadvantages of the traditional TLD algorithm in the mobile scene,combined with the ViBe algorithm to improve the TLD algorithm,establish a background model for the pixel,and pass Matching the background model reduces the number of scan windows of thealgorithm,which in turn speeds up the target tracking.2)Several feature point detection and matching algorithms based on natural feature points are studied,namely: SIFT algorithm,SURF algorithm and ORB algorithm,and the advantages and disadvantages of the three algorithms are obtained through experimental comparison,combined with SURF algorithm.The traditional ORB algorithm is improved,the feature points of the Hessian matrix determinant are obtained,and the feature points at different scales are obtained by Gaussian image pyramid.Then the direction information of the feature points is given by the gray centroid method.Finally,the RANSAC algorithm is used to eliminate the feature points.The algorithm matches the point pair incorrectly.Under the premise of ensuring the same advantages,the robustness and accuracy of the ORB algorithm are enhanced and verified by experiments.3)A three-dimensional registration algorithm based on improved TLD and improved ORB algorithm is proposed.The improved TLD algorithm tracks the target and reduces the range of detection of feature points.The semi-learning supervision method is used to predict the motion trajectory of the target.The improved ORB algorithm is used.The feature points of the target are detected and matched,and the transformation matrix between the virtual and real scenes is quickly calculated.By simulating the possible occlusion and target deformation in mobile scenes,the author experiments with the three-dimensional registration algorithm proposed in this paper.The experimental results show that the proposed three-dimensional registration algorithm shows the above scenarios.Good robustness and accuracy,but also a relatively acceptable operating speed.The research of text can enhance the accuracy and robustness of the application in the more complex external environment for the augmented reality application of the smart mobile device,improve the application enhancement effect,and satisfy the demands of user's application accuracy and real-time.