基于单目视觉对地面特征点定位方法

为了解决机器人在室内导航定位的问题,本文提出基于单目视觉对地面特征点定位的新方法.为扩大摄像头的视野,本方法采用鱼眼摄像头,然后将鱼眼摄像头安装在小车底盘下面,实时校正鱼眼摄像头拍摄的图片.通过对摄像头进行标定得到其内外参数,进一步用新方法计算出摄像头安装在小车上的姿态角,建立平面成像模型得到世界坐标系与像素坐标系的对应关系.再进行RANSAC算法对地板的特征点和特征线进行提取,然后获得相关特征信息再经过逆透视变换算法变换到物理空间上,获取真实物理信息.实验结果表明利用该方法能够更精准获得摄像头与地板边缘线的相对位置信息,并且成本也低但定位精度高.

A Method of Locating Ground Feature Points Based on Monocular Vision

Aiming at the problem of indoor navigation and localization of robot,a new locating method is proposed in this paper based on monocular vision positioning ground feature points.In order to enlarge the view of the camera,the fisheye camera is adopt-ed,and then the fisheye camera is fixed under the chassis of the car to capture pictures in real time used for undistortion.Through the camera calibration to obtain intrinsic and extrinsic parameters of the camera,calculating the attitude angles of camera with a new method which mounted on the car,we set up the plane imaging model of relation between the world coordinate system and pixel co-ordinate system.Then,the RANSAC algorithm is used to extract the feature points and the feature lines of the floor,and the real physical location of the feature points and the feature lines is obtained through the inverse perspective transformation algorithm.The experimental results show that the method can achieve higher accuracy of distance between camera and edge of line of the floor. Furthermore,the cost is low but the positioning accuracy is high.