视觉里程计中的相机姿态和高度实时测量方法  

Real-Time Measurement of Camera Attitude and Height in Visual Odometer

作　　者：曹毓[1] 张小虎[1] 冯莹[2] 

机构地区：[1]国防科学技术大学航天科学与工程学院,长沙410073 [2]国防科学技术大学光电科学与工程学院,长沙410073

出　　处：《传感技术学报》2015年第9期1354-1360,共7页Chinese Journal of Sensors and Actuators

基　　金：博士后基金项目(2014m562649)

摘　　要：在视觉里程计的应用中,实时准确的获得相机姿态和高度数据有助于提高视觉定位的精度。而现有解决方案要么成本过高,要么精度无法满足要求,为此提出了基于路面激光扫描的相机外参数实时测量方法。该方法将两台二维激光扫描仪相互正交安装且向下扫描,对获得的沿着两个方向的路面扫描线使用RANSAC算法估计出直线方程,根据两直线方程求得道路平面方程,并以该平面为参考获得相机相对路面的姿态和高度数据。室内实验结果表明:静态条件下对姿态的测量误差最大约0.1°,高度测量误差最大6 mm;室外动态实验结果表明:与传统的惯性测量方法不同,相机外参数测量结果不受车辆加减速运动的影响,且其动态姿态测量精度明显高于精度为1°的惯性测量系统。由于该方法获得的姿态和高度数据是以道路平面为参考基准,尤其适用于单目视觉里程计中以辅助提高定位精度。In the application of visual odometer, acquiring the high-precision attitude and height of camera in real time helps to improve the visual positioning accuracy. But existing solutions are either expensive or low in preci- sion, so we bring forward an real-time method to measure the camera＇ s external parameters based on the laser scan- ning of road surface. Two 2D laser scanner orthogonal installed and downward scan, then the linear equation was es- timated using the RANSAC algorithm. After that, road plane equation is obtained, which is the reference of the cam- era＇ s attitude and altitude data. the indoor experiment results showed that： in the static conditions, the measure- ment error of the attitude is about 0.1 degree, maximum height measurement error is about 6 mm; the outdoor dy- namic experimental results showed that ： different from the traditional method of inertial measurement, camera exter- nal parameters measurement results are not affected by the vehicle deceleration influence, and the attitude measure- ment precision is significantly higher than the inertial measurement system which has accuracy of 1 degree. Owing to obtaining the attitude and altitude data as the road plane for reference, the method is especially suitable for assist monocular visual odometry to improving the positioning accuracy.

关 键 词：计算机视觉 视觉里程计 激光扫描 姿态测量 高度测量 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]