期刊导航

论文摘要

一种基于视频序列单目视觉下的运动车辆轮廓重构算法

The contour reconstruction algorism of moving vehicle on monocular visionbased on video sequence

作者:赵树龙(武汉大学电子信息学院);王俊峰(四川大学计算机学院);潘 喆(苏州金螳螂怡和科技股份有限公司);王 芬(苏州金螳螂怡和科技股份有限公司);王新俊(苏州金螳螂怡和科技股份有限公司)

Author:ZHAO Shu Long(Electronic Information School,Wuhan University);WANG Jun Feng(College of Computer, Sichuan University);PAN Zhe(Suzhou Gold Mantis Yihe Technology Company Limited);WANG Fen(Suzhou Gold Mantis Yihe Technology Company Limited);WANG Xin Jun(Suzhou Gold Mantis Yihe Technology Company Limited)

收稿日期:2014-10-21          年卷(期)页码:2015,52(5):1007-1013

期刊名称:四川大学学报: 自然科学版

Journal Name:Journal of Sichuan University (Natural Science Edition)

关键字:单目视角;车辆轮廓重构;三维重建;车型识别

Key words:Monocular vision; Contour reconstruction of vehicle; 3D reconstruction; Vehicle recognition 1 引 言

基金项目:国家自然科学基金(91338107,11102124);教育部博士点博导基金(20130181110095);四川省科技支撑计划项目 (2013SZ0002, 2014SZ0109)

中文摘要

针对运动车辆在单目视觉下轮廓重构的问题,本文通过特征目标识别的方法解算出无约束运动车辆的运动参数,将无约束运动物体的轮廓重构问题转化成已知约束运动物体的轮廓重构问题,提出了一种基于视频序列的单目视觉下的运动车辆轮廓重构算法.实验结果证明,新算法能够较好地重构出不同大小车辆的3D轮廓,其车辆高度估算值的误差在4%以内

英文摘要

To solve the problem of contour reconstruction of moving vehicle on monocular vision, the authors propose a new algorism based on video sequence. In this paper, we obtain the motion parameters of unconstrained moving vehicle by solving the characteristic equations, and convert the contour reconstruction problem of unconstrained moving vehicle into the problem of constrained moving vehicle. Experiments show that the new algorism can reconstruct contour of different vehicle, and the error of vehicle height estimation is less than 4%.

关闭

Copyright © 2020四川大学期刊社 版权所有.

地址:成都市一环路南一段24号

邮编:610065