基于双目视觉系统的三维目标识别研究  

Identification Technique for 3D Target Based on Binocular Visual System

作　　者：张驰[1] 靳伍银[1] 刘昊[1] 冯瑞成[1] 

机构地区：[1]兰州理工大学机电工程学院,甘肃兰州730050

出　　处：《机床与液压》2010年第1期22-23,126,共3页Machine Tool & Hydraulics

基　　金：国家自然科学基金(30670529)资助;甘肃省自然科学基金项目(3ZS042-B25-019)

摘　　要：双目视觉系统一般是由一个或两个摄相机从不同角度获取周围景物两幅数字图像,基于人类视觉的视差原理重建周围景物的三维几何信息。通过开源计算机视觉库(OpenCV)重建周围景物的三维几何信息,并提取三维目标的特征信息。采用支持向量机(Support Vector Machine)的软件包LIBSVM进行小样本的学习,从而达到对特定目标的识别。The theory of binocular visual system is that two digital images of surrounding scenery are got by one or two cameras from different point of view, then 3D information of surrounding scenery is reconstructed based on the principle of binocular visual sys- tem. One identification technique for 3D target was proposed. The characteristic information of target was extracted and reconstructed from two digital images based on Open Source Computer Vision Library （OpenCV）. The recognition of given target was realized based on LIBSVM of Support Vector Machine.

关 键 词：三维重构 模式识别 视差 OPENCV 支持向量机 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]