基于LIDAR与深度相机的EKF数据融合研究  

Research on EKF Data Fusion Based on Lidar and Depth Camera

作　　者：尹辰柯 胡家辉 王靖宇 YIN Chenke;HU Jiahui;WANG Jingyu(School of Information Engineering,Southwest University of Science and Technology,Mianyang,Sichuan Province,621010 China)

机构地区：[1]西南科技大学信息工程学院,四川绵阳621010

出　　处：《科技资讯》2020年第29期33-35,共3页Science & Technology Information

基　　金：西南科技大学大学生创新基金精准资助专项资助《激光雷达及三维视觉融合的三维避障系统设计》(项目编号:JZ19-092)。

摘　　要：常用定位及避障传感器中,LIDAR不受光照影响、创建地图精度高,但数据量较少;Kinect深度相机成本低、数据量丰富,但受环境影响较大。在此基础上,提出一种将激光雷达与Kinect深度相机数据融合的方法。该方法预处理深度图像数据,转换为激光数据,从而投影到极坐标上,得到的非线性数据,在处理上引入了雅可比矩阵。接着通过EKF算法,将线性激光雷达信息和变换非线性深度相机信息融合。通过对比,EKF融合数据相较单独LIDAR数据或Kinect深度相机数据能够在大部分场景中更好地还原真实路径。Among the commonly used location and obstacle avoidance sensors,lidar is not affected by illumination and has high accuracy in creating maps,but the amount of data is small;Kinect depth camera has low cost and rich data,but it is greatly affected by the environment.On this basis,a method of data fusion between lidar and Kinect depth camera is proposed.In this method,the depth image data is preprocessed,converted into laser data,and then projected onto the polar coordinates.The Jacobian matrix is introduced in the processing of nonlinear data.Then,the linear lidar information and transform nonlinear depth camera information are fused by EKF algorithm.Compared with lidar data or Kinect depth camera data,EKF fusion data can restore the real path better in most scenes.

关 键 词：深度图像 极坐标 障碍物检测 路径规划 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]