铁路运行环境下的车载相机姿态估计

Attitude estimation of the train-borne camera from the railway-environment

基于对铁路运行环境中所拍摄的图像特征的分析,提出了一种修正的线段检测子(MLSD),并利用该线段检测子结合最小二乘拟合及交叉迭代优化方法对消失点进行快速检测,进而利用消失点坐标实现了相机姿态的估计,为后续调整机器视觉算法进行铁路运行环境的场景重建提供重要依据。运用该方法,基于车载相机所拍摄的图像对车载相机姿态进行估计,实验所得的估计值满足了车载检测系统的精度要求,表明了此方法对于估计相机姿态有效的。

The attitude of the train-bore camera is a set of important external parameters. It is very helpful to analyze the images using the machine vision methods. And adjusting the machine vision algorithms and understanding the scene can be guided by estimating the attitude of the camera. By analyzing the features of images taken from the railway envi-ronment, a method of estimating the attitude of the train-bore camera is proposed, i.e. using the detecting algorithm of vanishing points based on the proposed modified line segment detector （MLSD）, least-square technique and cross iteration methods to obtain the attitude of the camera. The experimental results show that the validation and accuracy of method can meet the requirements of the precision of vehicle detection system.