Motion Estimation Using Region-Level Segmentation and Extended Kalman Filter for Autonomous Driving

Motion estimation is crucial to predict where other traffic participants will be at a certain period of time, and accordingly plan the route ego-vehicle. This paper presents novel approach estimate motion state by using region-level instance segmentation extended Kalman filter (EKF). involves three stages object detection, tracking parameter estimate. We first use accurately locate region for latter two stages. The combines color, temporal (optical flow), spatial (depth) information as basis super-pixels Conditional Random Field. optical flow then employed track feature points within area. In stage estimate, we develop relative model ego-vehicle object, establish an EKF point integrates ego-motion, flow, disparity generate optimized parameters. During apply edge constraint consistency eliminate outliers so that used are ensured body estimates refined inner points. Experiments have been conducted on KITTI dataset, results demonstrate our method excellent performance outperforms state-of-the-art methods either in