Researching The Core Technology Of The Outdoor Augmented Reality System In Real Time

Augmented Reality(AR) is a technology that it can integrate the computer-generated virtual information into the real scene.It has a very broad application prospects. And it is quickly used in various fields,conveniencing and enriching people’s lives.The tracking registration problems is one of the difficulties for researching the technology of Augmented Reality. Currently, more mature technology is based on special markers. Due to the method based on the special markers is simple, and without impacting by environment.It is widely used in the relatively stable scene, such as indoors. However, the trend of technology development of augmented reality is to be freely used in every scene, confined indoors, outdoor use is inevitable. Due to technical limitations of special markers, it is no longer suitable for outdoor applications in the augmented reality scene.For outdoor scenes of location registration method for augmented reality systems, it needs unmarked technology to solve. This article is based on computer vision technology, research and design a markerless augmented reality systems for outdoor scenes, to identify target objects in natural scenes, such as book covers, building construction, etc. System use of feature extraction and matching algorithms, look for the object, again according to the camera’s parameters corresponding relationship between virtual objects and objects, through the transformation of the coordinate system, and eventually, registered on the target virtual objects, complete enhancements to reality. Completed research work includes:1) In the detection and exteaction of feature,The accuracy of the single classic algorithm of recogniting outdoor scene is not high or poor real-time.eventually, system combines using ORB and BRISK algorithm, experiment proved, both of combines, target of looking for more efficient, full reflected has algorithm of scale not degeneration and rotating not degeneration, target real 360 ° rotating, are can found, and speed also meet enhanced reality of real-time sex requirements;2) For the feature matching algorithm, the real-time feature matching has the problem that it is matching slower and mismatch. This paper give Hamming distance matching algorithm, which sampled sample binary bit measured at their similarity; relative Euclidean distance matching algorithm to calculate speed, matching rate is high; using RANSAC algorithm to remove the mismatching points, the match point constrained to a straight line, and then greatly improve the matching speed.3) In the calibration and registration, for the problems of the registration result from the target center. This paper presents two methods for camera calibration, resulting in accurate camera parameters. By comparing the two sets of parameters, get more precise camera parameters, thus enabling registered results more accurate. System as much as possible the result so convincing results.The system realizes indoors, outdoors enhance the effect of the virtual objects of reality, even if part of the block, but also stable operation; real-time systems meet the basic requirements.