微型机与应用Issue(5)：50-53,4.DOI:10.19358/j.issn.1674-7720.2017.05.016
基于ORB-SLAM的移动机器人嵌入式实现与优化
Embedded implementation and optimization of mobile robot based on ORB-SLAM
摘要
Abstract
The source code of ORB-SLAM (Simultaneous localization and mapping system based on the ORiented Brief feature recognition) can't be run on the embedded development board,and construction of the point cloud is too sparse to meet the requirements of mobile robot path planning.To solve this problem,the ORB-SLAM was improved and optimized,and then transplanted to the SLAM process in the embedded development board.Firsty,the trajectory,point cloud,and some dependency database in original Linux system under the PC were deleted.Then the most of the C++ code in src and include folder were retained and improved.Subsequently,the improved C++ code were called by JNI method in the embedded platform,and OpenCV,g2o,DBoW2,Eigen,and other dependent libraries were increased.Finally,according to the key frame connected to the grid map,the real-time track display and map building was completed.The results showed that the SLAM process,hardware configuration requirements and cost of the hardware configuration were greatly reduced by transplanting the embedded development board,and the construction of the grid map of the stock was smaller and more intuitive to reflect the actual layout of the environment.The construction chart error was controlled in the high precision of ±0.5 m similarly,which greatly improved the performance of SLAM.关键词
ORB-SLAM/嵌入式/移植/栅格地图Key words
ORB-SLAM/embedded/transplantation/grid map分类
信息技术与安全科学引用本文复制引用
林连秀,叶芸,姚剑敏,郭太良..基于ORB-SLAM的移动机器人嵌入式实现与优化[J].微型机与应用,2017,(5):50-53,4.基金项目
福建省自然科学基金(2014J01236) （2014J01236）
福建省科技重大专项(2014HZ0003-1) （2014HZ0003-1）
福建省资助省属高校专项课题(JK2014003). （JK2014003）
