一种基于视觉测量障碍高度的车载摄像机标定新方法

A New On-board Camera Calibration Method Based on Vision Measurement the Height of Obstacle

从摄像机针孔成像几何模型入手,根据像点与物点之间的线性关系,进而分析车载摄像机景物点的图像坐标与实际坐标之间的关系,研究发现无法通过建立图像坐标与实际坐标之间的映射关系测量出景物的高度。基于此提出了一种新的摄像机标定方法,通过插值法和最小二乘法建立景物的图像高度(景物顶点的图像纵坐标与其在地面上投影点的图像纵坐标之差)与实际高度之间的映射关系,即车载摄像机标定的模型;并对该模型进行测试。测试结果最大偏差为0.078 75 cm,最小偏差为0.013 99 cm,平均偏差为0.034 79 cm,精度可达到0.11%。经验证该方法可精确地测量出障碍的高度,具有一定的实用性及可行性,适合车载摄像机标定,有助于汽车辅助驾驶系统的研究。

Starting from pinhole imaging geometric model of the camera,according to the linear relationship of the image point and the object point,the scenery spot of on-board camera of the relationship between the image coordinates and the actual coordinates was further analysed,found that cannot by establishing the mapping relationship between the image coordinates and the actual coordinates to measure the height of obstacles.So a new camera calibration method was presented,by interpolation method and least squares method to establish the mapping relationship between image height and actual height of the scenery,namely the calibration model of on-board camera,to test the model,and the test result of maximum deviation is 0.07875 cm,the minimum deviation is 0.01399 cm,the average deviation is 0.03479 cm,and the precision can be up to 0.11%.The method can accurately measure the height of obstacles,has a certain practicality and feasibility,suitable for the calibration of on-board camera,it will be helpful to the study of automotive driver assistance systems.