摘要
传统的视觉里程计算法对弱纹理区域和远距离区域的计算精度较低,且算法实时性较差。本文介绍了一种视觉里程计算的可行方法:通过特征提取、特征匹配和运动估计得出车体运动的距离和方向信息,并通过极线几何约束、多帧跟踪和变焦距等方法有效地提高了精度。采用Harris特征点提取和线性与非线性最小二乘结合的运动估计方法,在确保精度的同时提高了算法的实时性。实验表明效果良好。
Conventional visual odometries have low-precision in the low texture regions and long distance,and poor real-time performance.A method about visual odometry was proposed to estimate the motion information of the vehicle.It contains feature detection,feature matching,motion estimation and so on.To make the algorithm more precise,the epipolar geometry constraint,multi-frame tracking and alterable focus were applied.Moreover,to ensure the real-time performance and accuracy of the algorithm,Harris corner detection and combination of the linear and nonlinear least square method（LSM） were used.The experiments demonstrate that the method is valid.
出处
《兵工学报》
EI
CAS
CSCD
北大核心
2007年第12期1498-1502,共5页
Acta Armamentarii
基金
国家自然科学基金重点资助项目(NSFC60534070)
浙江省科技计划资助项目(2005C14008)
中国博士后科学基金资助项目(2006041036)
关键词
自动控制技术
陆地自主车
导航
视觉里程计
多帧跟踪
运动估计
automatic control technology
autonomous land vehicle（ALV）
navigation
visual odometry
multi-frame tracking
motion estimation
