基于双目视觉的嵌入式三维坐标提取系统  

Embedded 3D Coordinate Extraction System Based on Binocular Vision

作　　者：夏茂盛[1] 孟祥磊[1] 宋占伟[1] 贺琨[1] 

机构地区：[1]吉林大学电子科学与工程学院,长春130012

出　　处：《吉林大学学报（信息科学版）》2011年第1期61-66,共6页Journal of Jilin University（Information Science Edition）

基　　金：国家科技型中小企业创新基金资助项目(09C26212200450)

摘　　要：针对汽车辅助倒车和泊车系统的需求,提出了一种基于嵌入式处理器ARM11和Linux操作系统的双目立体视觉三维坐标提取系统,给出了系统的硬件和软件设计方案,并详细论述了坐标提取算法。在搭建的测试平台上,测试了坐标提取算法以及测量误差和测量范围的关系。实验结果表明,在0.5～1.6 m范围内,深度测量误差小于1.2%,能满足应用的精度要求。与基于PC机的双目视觉测量系统相比,该系统的价格和功耗明显降低,且体积也明显减小。该系统的实现对双目视觉技术的实用化有一定的推动作用。In order to meet the demands of automobile assisting parking and astern running system,we propose a binocular vision 3D coordinate extraction system based on embedded processors ARM11 and Linux OS.We first provide the designs for both hardware and software,and then thoroughly describe the algorithm for coordinate extraction.On the fabricated testing system,experiments were carried out for testing the algorithm performances and the relationship between measurement errors and distance.The results exhibit that the measurement errors are less than 1.2% in the distance between 0.5 m and 1.6 m.The size,cost and power consumption of the system based on embedded processors are much smaller than those in the system based on PC.The implement of this system somehow promotes the substantiation of binocular vision technique.

关 键 词：嵌入式 三维坐标提取 双目视觉 

分 类 号：TP274[自动化与计算机技术—检测技术与自动化装置]