机器人平台上基于Kinect相机的地平面检测方法  

A Novel Ground Plane Detection Method with Kinect on a Mobile Robot

作　　者：刘俊[1] 夏杰[1] 陈雁秋[1] 

机构地区：[1]复旦大学计算机科学技术学院

出　　处：《微型电脑应用》2014年第5期10-13,共4页Microcomputer Applications

基　　金：国家自然科学基金(61175036)

摘　　要：地平面检测对于移动机器人进行3D导航和环境感知非常重要。描述一种基于Kinect相机的地平面检测的新算法。算法主要包括两个部分:首先,通过3D点云的多方位投影变换及2D投影面的L-p范数集中度度量获取地平面的法向量,然后,根据获得的法向量进行3D空间中地平面具体位置的确定。算法使用了安装在PR2机器人平台上的Kinect相机拍摄得到的数据集进行验证,实验结果表明该算法可以获得非常可靠的地平面检测结果。Ground plane detection is very important for 3D navigation and environment perception in robotics fields.A novel ground plane detection method using Kinect is introduced,which contains two main procedures.First,the normal vector of the ground plane is obtained by projecting the 3D point cloud to a series of directions and measuring the concentration levels of the 2D projected maps with L-p norm metric technology.Then the ground plane is further searched based on the calculated normal vector.The dataset captured with Kinect mounted on a PR2 robot is used for evaluation.Experiments show very promising results on this dataset.

关 键 词：地平面检测 机器人 KINECT 3D点云 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]