Planar Object Tracking Algorithm Based On Control Points For Augmented Reality Application

Planar object tracking is a kind of 3D-tracking technology for plane targets in the field of AR.It requires real-timely and accurately calculation of the 3D pose of a certain plane target in camera image.Then it will provide a transformation matrix for transmission projection of AR render system.Through lots of investigations to the existing planar object tracking algorithms,these methods are generally inadequate due to the poor real-time performance and weak robustness for challenging conditions.As for these urgent problems,this thesis makes the corresponding research,which is as follows:Firstly,this thesis proposes a novel planar object tracking algorithm based on control points.The core of this method is three-step: 1)the spatial-temporal relationship between control point and their gradient contexts is modeled;2)the 2D coordinate of control point can be computed according to the correlation of the spatial-temporal model and priori model;3)four sets of 2D-3D coordinate correspondences are able to be obtained through four control point trackers,so that it is easy to estimate the pose of the current planar object by PnP algorithm.Apart from this,this thesis proposes a control point relocation algorithm based on key frame and feature match for those object with enough texture.This method makes the planar object tracking algorithm get self-correction ability.Several of experiments show that the planar object tracking algorithm this thesis proposed not only holds well real-time performance and highly accuracy but also is robust to many challenging tracking conditions.Secondly,this thesis designs an AR chart let application based on the above method.This application can real-timely register the augmented image on the surface of planar object according to the pose of the planar object to camera.In addition,this application is able to expand to poster campaign,wall advertising,guidance information view,and so on.