Motion compensation for low-speed unmanned ground vehicles based on stereo camera and inertial measurement units

Inter-frame displacement caused by camera motion is a common challenge in autonomous driving, as it directly affects the accuracy of tasks in complex scenarios. This paper aims to reduce displacement caused by rapid camera motion, using a constant velocity motion compensation method based on inertial measurements. By integrating pose information obtained from IMU (Inertial Measurement Unit) with a rigid-body kinematic model, this study performs motion compensation on point cloud data generated from the KITTI dataset. The motion-compensated data is further utilized for dynamic-static object classification across various scenarios. Experimental results demonstrate that the method effectively counteracts the camera's motion and enhances classification performance under different scenarios.