基于图像识别的视觉跟踪注册算法-中国图象图形学报.doc

中图法分类号:TP391 文献标识码:A 文章编号:1006-8961(200 ) - 论文引用格式： 结合图像识别的移动增强现实系统设计与应用 严雷1 ， 杨晓刚2 ， 郭鸿飞2 ， 陈靖1 （1.北京理工大学光电学院100081； 2.新华社通信技术局100803） 摘 要：目的：在移动互联网时代下，移动增强现实应用得到越快的发展。户外场景中存在许多相似结构的建筑，手机的存储和计算能力有限，因此应用多集中于室内小范围环境对于室外大规模复杂场景的适应性较弱。对此，建立一套图像识别的移动增强现实系统。方法：为解决相似特征的误匹配问题，Gravity-SURF和Gravity-BRISK特征描述。系统对增强信息进行有效管理，采用基于Gravity-SURF特征的VLAD方法对大规模图像进行识别；智能终端的中呈现识别图像的增强信息，并利用识别图像的Gravity-BRISK特征和光流结合的方法对相机进行跟踪，采用Unity3D渲染引擎实时绘制三维模型。结果：在包含重力信息4000张户外图像的数据库中进行实验。，能够增强具有相似特征的描述符的区分性，并提高匹配正确率。图像识别算法的识别率能达到88%以上，识别时间在420ms；光流跟踪的RMS误差小于1.2个像素，帧率能达到23FPS。结论：本文针对室外大规模复杂场景建立的基于图像识别的移动增强现实系统，能方便对不同应用的增强现实数据进行管理。系统被应用到谷歌眼镜和新闻领域上，不局限于单一的应用领域。结果表明，识别算法和跟踪注册算法能够满足系统的精度和实时性要求移动增强现实; 重力; 图像识别; 特征跟踪; 模型渲染Mobile Augmented Reality System Design and Application Based on Image Recognition Yan Lei1, Yang Xiaogang2, Guo Hongfei2, Chen Jing1 (1. School of Optoelectronics, Beijing Institute of Technology, Beijing, 100081; 2. Communication Technology Bureau, Xinhua News Agency, Beijing, 100803) Abstract： Objective: With the development of mobile Internet, the speed of mobile augmented reality application advancement is getting faster. However, because the storage and computation capacity are limited on mobile devices and the large-scale outdoor images contain lots of similar building structures images, those applications are restricted to indoor environment. To solve the problems, a mobile augmented reality system based on cloud image recognition is established. Method: To reduce mistake-matching probability, we add gravity information into SURF and BRISK feature description algorithm. In our system, those two description algorithms are named Gravity-SURF and Gravity-BRISK. In cloud, large-scale images and augmented information are managed by the augmented reality management system. VLAD algorithm based on Gravity-SURF is applied to the large-scale images recognition. On intelligent terminals, different augmented reality application’s user interfaces are designed to display the augmented inf