基于多传感融合目标检测的动态物剔除SLAM算法OA

To address the issue that the positioning accuracy and mapping quality of Simultaneous Localization And Map-ping(SLAM)algorithms degrade due to interference from dynamic goose flocks on feed delivery mobile robots in modern goose farming scenarios,a dynamic SLAM algorithm based on multi-sensor fusion target detection is proposed.This al-gorithm is built on the LIO-SAM framework,integrating LiDAR and inertial measurement unit(IMU)to construct the SLAM system,and adopts a front-end and back-end architecture to optimize positioning and mapping performance.It uses the Hungarian algorithm to track the movement state of goose flocks in real time,and combines with multi-sensor fusion target detection algorithm to accurately identify and eliminate feature points generated by dynamic goose flocks,ef-fectively reducing positioning and mapping errors.Tested on public datasets such as KITTI and UrbanNav as well as ac-tual farming scenario data,in the KITTI07 sequence,the root mean square error(RMSE)is reduced by 33.18％com-pared with classic algorithms like LeGO-LOAM,LIO-SAM and LVI-SAM.In actual goose farming environments,it can quickly filter out dynamic goose flock interference,improving mapping quality and navigation reliability.This re-search provides a new technical solution for intelligent feed delivery in goose farming and promotes the development of animal husbandry automation.