The Research And Implementation Of Web Augmented Reality System Based On Natural Markers

Augmented reality is a technology which calculates the position and angle of the camera in the scene and augments the images,videos or 3D models to the scene in real time.Its goal is to try to augment the virtual information in the real world and make them interact so that people can understand the real world better.Although mobile augmented reality has received more and more attention due to its portability,good interactivity and excellent user experience,it cannot be run across mobile phone systems.In contrast,web-based augmented reality technique is platform independent.We designs and implements a web-based augmented reality system based on natural markers,to make web augmented reality be able to be used in more complicated environment.The system mainly includes camera calibration,object tracking and model rendering.The paper deeply researches the key issues of camera calibration and object tracking,and proposes some improvements of such algorithms.The accuracy of the corner detecting influences the result of camera calibration directly.The paper analyses the issues of using the Harris algorithm to detect the corners of the chess board during camera calibration and proposes an improved method.The algorithm improves the accuracy of the corner position according to the geometry information of the chess board corner,and the improved Harris algorithm has a higher accuracy.The accuracy of camera calibration has also improved.Object detection is the key of consistent and stable 3D registration.We mainly use the ORB algorithm and the RANSAC algorithm to achieve target detection.Through analysis,it is found that the RANSAC algorithm repeatedly selects some subsets from the dataset to estimate the model and test the estimation model,and finally finds the optimal model.The RANSAC algorithm is very time-consuming and makes the system run inefficiently.To accelerate the tracking process,the paper proposes an optimized method which takes advantage of the sequence similarity of feature points to get rid of unnecessary matching processes.The proposed method effectively reduces the number of times to calculate the best adaptation model,reduces the time to calculate the best adaptation model and improves the real-time performance of the system.The paper utilizes the two improved algorithms in the web augmented reality system,and improves the code efficiency to achieve the effect of augmented reality.The results of the experiment tests show that the web augmented reality system can deal with the object zooming,rotating and partial sheltering,and has good robustness.