车载序列影像直接相对定向质量  

Quality analysis of vehicle-based sequence images direct relative orientation

作　　者：徐振亮[1,2] 闫利[1] 段伟[1] 陈林[1] 刘华[1] 

机构地区：[1]武汉大学测绘学院,湖北武汉430079 [2]辽宁工程技术大学测绘与地理科学学院,辽宁阜新123000

出　　处：《辽宁工程技术大学学报（自然科学版）》2013年第3期321-325,共5页Journal of Liaoning Technical University (Natural Science)

基　　金：国家自然科学基金资助项目(40971192/D0106)

摘　　要：为验证地面大转角、小面幅序列数码影像直接相对定向质量.推导了一套基于SIFT特征匹配的计算机视觉分析方法,解算过程简洁明了.研究结果表明:基于计算机视觉方法能够有效解决大转角影像、特别是车载序列数码影像相对定向中的匹配难题,解算结果比较稳定;从目前的结果来看,直接相对定向质量还不高,但研究方法对室内移动机器人环境感知、无人驾驶等导航领域具有一定的借鉴价值;最后分析了制约其质量的主要原因,为今后分析和提高视觉量测质量提供了一条思路.In order to verify the direct relative orientation quality of narrow frame sequential images with big rotation angels,this study derives a computer vision analysis method based on SIFT,which simplifies the calculation process.The comparison result between orientation parameters from POS and that based on computer vision shows that the method can overcome the matching problem of images with big rotation angles.In particular,for vehicle based sequential images,the result is quite stable.Currently the quality of direct relative orientation using this method is not very high,however,the method provides a useful reference for many surveying and mapping related fields,such as indoor robot mobile environment awareness and unmanned automobile.Furthermore,the study analyzes the factors that affect the quality of orientation based on computer vision and provides a solid foundation for future study.

关 键 词：相对定向 计算机视觉 矩阵分解 直接地理参考 位姿系统 序列影像 移动测图 尺度不变性特征变换 随机取样一致性估计 

分 类 号：P234.1[天文地球—摄影测量与遥感]