The Business Card Recognition AR System Based On SURF Algorithm

In the age of information,mobile Internet and mobile terminals have become the most important part of people's lives.The image as one of the basic but essential information bearer form,is the way people get the ma-jority information in daily life.Among them,the user can transfer and ob-tain a large amount of information by taking photos and sweeping functions through the terminal.How to further analyze and interact with the image information obtained by this way has become an important topic in the field of computer vision.The difficulty is how to identify the target that the user is interested in from the image acquired by the mobile terminal.The topic of this thesis is to explore and practice real-time detection systems for im-ages.C ombining the frontier AR(augmented reality)effect,3D rendering is performed on the identified target.The user can quickly identify the busi-ness card while scanning the image,and obtain the information dynami-cally displayed on the business card rendered by the AR system.To deal with the scene of mobile phone scanning to identify business cards,this thesis deeply studied the efficiency of traditional matching al-gorithm in the mobile terminal.First,we optimized the SURF algorithm.In the offline production of database and online real-time recognition,my algorithm adopts respective improvements.For the offline production of data sets,the scale space is fully constructed,and the features that are sig-nificant at each scale are extracted to ensure the scale invariance.And the feature point set is purified by RANSAC algorithm,which reduces the amount of calculation when matching the image.For the online real-time recognition scene,the ROI of the business card is firstly calculated from the entire image,and then the calculation of the scale space is simplified,which greatly reduces the amount of data of the image to be extracted.Then,the extracted features are matched with the data set to obtain the final matching result.At the same time,the matching algorithm is improved,LRU cache algorithm is introduced in the matching process to reduce the number of samples in the data set that need to be traversed in the matching process to further reduce the matching process time.In the end,this thesis combines AR technology to perform AR rendering on the recognized busi-ness card using Unity3D.In this thesis,the original image collected by An-droid is superimposed with the 3D model view.The model moves dynam-ically with the change of the business card's position;thus,the business card transform from a single image to an augmented reality information carrier with multimedia display capabilities.The research in this thesis fi-nally landed into an Android mobile phone application.After the accuracy and response time test,the feasibility of the algorithm and model was proved.