Research And Application Of Tracking Registration Technology In Mobile Augmented Reality

Image tracking registration technology is a very important basic technology for implementing mobile augmented reality systems,and it is also the key to determining the performance of mobile augmented reality systems.There are two problems in the actual application of tracking registration of mobile augmented reality: system delay and the images cannot be identified effectively.It is the key point to study an algorithm that adapts to the mobile augmented reality system in academic research.The main research work of this article is as follows:(1)Regarding to the problem that the image cannot be identified effectively,the method of detecting and describing feature points of ORB(Oriented FAST and Rotated BRIEF)based on bilateral filtering algorithm is proposed.Firstly,this paper compares the common filtering methods by analyzing and selecting nonlinear bilateral filtering to perform image preprocessing,which has the advantage of preserving edge noise removal.Secondly,it adopts the ORB feature detection algorithm.The algorithm uses the pyramid principle to enhance the fastest detection FAST algorithm.The feature points have scale invariance,and the gray point centroid method is used to obtain the offset angles between the corner grays and the centroids so that the feature points have directions.Finally,the ORB feature description algorithm is used to establish the coordinates of the key points as the center of the circle.The system makes the BRIEF description sub-coordinate system rotate with the rotation of the picture,so that the feature points have rotation invariance.The results show that the proposed algorithm can provide better edge features while suppressing background noise,and it is beneficial to obtain high-quality feature points with scale invariance and rotation invariance,thus making the image more effective.(2)Regarding to the problem of real-time application of mobile augmented reality,a robust and efficient ORB feature matching and registration method is proposed.Firstly,feature descriptors are acquired using feature detection and description of the bilateral filtering preprocessing.Secondly,the binary descriptors are matched using Hamming distances.They are sorted in descending order by the fast sorting method,thereby reducing robust and efficient PROSAC(Progressive Sample Consensus)algorithm arranges the time,then alternating from the inner point set and the ordered sample set to select the sample point,reducing the number of iterations,quickly eliminates the ORB algorithm mis-match pair,achieving the effect of improving the real-time performance of feature matching.Finally,virtual reality fusion will be realized through the camera coordinates transformation Pose estimation.The experiments show that the matching real-time performance is improved after the edge feature information is retained and the false matching pair is removed.(3)To implement a smart education system based on mobile augmented reality.To build a system framework for the tracking registration module,and to apply the improved algorithm to the system.The boundary value analysis method is used to test the target image tracking registration in a complex environment.The experiments show that the improved image can be real-time recognized effectively in different angles,different sizes or being occluded.