Research of Visual SLAM Based on Semantic Information in Dynamic Scenes
Aiming at the problem of reduced positioning accuracy of visual simultaneous localization and mapping(SLAM)system in dynamic scenes,this paper proposes a visual SLAM system based on semantic informa-tion.Based on YOLO object detection algorithm,feature points of dynamic objects are eliminated and ellipsoid mod-eling of static objects is completed,and semantic maps are constructed.At the same time,the validity of the algo-rithm is verified by exposing data sets and actual scenarios.The results show that this algorithm can improve the po-sitioning accuracy of the system in dynamic environment.
dynamic scenessimultaneous localization and mappingsemantic information
万方数据
维普
