Dynamic Object Culling Based on Depth Camera and IMU in Dynamic Environment
WANG Xianfeng
LIU Zhaojie
ZHAO Tong
YANG Haowei
WANG Baitong
Abstract：At present,most excellent slam schemes are based on static environment,while the actual environment is often dy-namic,and the existence of dynamic goals has serious harm to the slam system.In order to improve the accuracy of pose estimation in dynamic environment,a dynamic point culling algorithm combining IMU rotation compensation and geometric constraints is pro-posed based on ORB-SAM2.In the tracking process,the rotation matrix obtained by IMU pre integration between two frames is used to compensate the previous frame,and the two images after compensation have the same rotation component.Then the motion vector of the corresponding feature points of the two frames is calculated,and the velocity vector of IMU between the two frames is used to distinguish the dynamic feature points from the environment.In order to further eliminate the dynamic feature points in the environment,the polar constraint is used to test and further filter the dynamic feature points.The experimental results in dynamic environment show that the proposed method can effectively remove dynamic feature points and improve the positioning accuracy in dynamic environment.
Keywords：IMURGBDdynamic environmentSLAM
Publication Date：2025-01-19
Online Publishing Date：2025-08-15(First online date of this platform, not the publication date of the document)
Pages：6( 90-95 )