Research On Building Recognition Algorithm Applied To Campus Reinforcement Reality System

With the increasing escalation of mobile hardware and the rapid development of mobile Internet, mobile augmented reality applications started from initial position searching browser; and gradually extended to education, entertainment, product promotion, tourism and many other areas. The range of augmented reality system has gradually expanded from indoors with limited space to more complex outdoor environment. This article aims to carry out appropriate research on building recognition algorithm used in outdoor augmented reality system, which involves feature matching, camera calibration, motion recovery, scene rendering, system implementation, and other aspects. In order to facilitate the experiment, we choose campus buildings as study object. The goal is to accurately identify a building and provide personalized, visual building information. The main contents are as follows:Firstly, extracting visual features from the real-time image obtained and comparing these features to features stored in the database; and thus the building corresponding to the current image is identified. In addition to classic feature point methods, this paper plans to optimize parameters based on line features after solving camera motion parameters. So this paper compares two line feature extraction methods and selects the method based on fragment curvature estimation method for extracting line features.For camera calibration, we improve calibration processes using vanishing point based on classic Zhang Zhengyou plane calibration method. Our method can alert the user to change the image plane posture with respect to the calibration plate, which can reduce the image number without affecting the accuracy of calibration.For motion recovery, we use matched feature points to compute motion parameters of the camera and then convert them to posture of building model. To set up Unity Camera Rotation parameters, external camera rotation matrix R is converted to Euler angles. We then optimize Euler angles based on line features to further enhance the fusion scene rendering effect.For fusion scene rendering, we analyze how the building model is rendered and registered into real scenes in Unity.Finally, we develop mobile augmented reality system on Android platform relying on cross-platform called Unity. The system integrates rendering modules in Unity, image processing modules in Android, Unity and Android interaction modules and so on.