增强现实中的目标远距离识别方法研究  

Research on Far Distance Recognized the Objects in Augmented Reality

作　　者：吕淘沙 汤汶[1] 万韬阮[1] 朱耀麟[1] 武桐[1] 

机构地区：[1]西安工程大学电子信息学院,陕西西安710048

出　　处：《软件工程师》2015年第4期7-8,共2页Software Engineer

摘　　要：在以往ARTool Kit增强现实系统中,摄像机能够正确识别标识物并在计算机生成虚拟物体的距离大都过近。针对增强现实三维注册过程中传统算法标识物匹配过程中误识率较高,识别距离不够远,提出了基于轮廓特征点的目标远距离识别方法。在识别过程中,利用多边形近似算法和金字塔中的迭代Lucas-kanade算法寻找图像中要跟踪识别的对象,对寻找到轮廓角点进行亚像素定位,利用角点的精确定位来跟踪目标。实验结果验证及分析表明,采用这种方法,其识别距离有大幅度增加。In the unimproved ARToolKit system,the distance is too close that camera can recognized the markers correctly and the virtual objects generated by the computer.The insufficient of traditional algorithm of the three-dimensional registration process in augmented reality is analyzed,a method to improve the registration in the augmented reality is put forward.In the process of identification,the extraction and outline of the polygon fitting algorithm and pyramid Lucas-kanade algorithm are used to identify and track targets.The corners of the polygon target are positioned in sub-pixel resolution of the position.The polygons can be replaced with these corners which describe the goals in high precision.The experiment result shows that the recognized distance is improved greatly.

关 键 词：增强现实 远距离 目标跟踪 多边形近似 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]