Evaluation of Kalman and Unscented Kalman Filters for Multiple Object Tracking in ADAS Applications using the KITTI Dataset

Multiple Object Tracking (MOT) is a crucial component of Advanced Driver Assistance Systems (ADAS), enabling accurate perception and prediction of dynamic objects in complex environments. This paper evaluates the Kalman Filter (KF) and Unscented Kalman Filter (UKF) for MOT using sensor fusion of LiDAR and camera data. The study examines measurement and process noise tuning, data association strategies, and their impact on tracking accuracy. Experiments conducted on the KITTI dataset demonstrate that UKF outperforms KF in nonlinear scenarios, offering improved tracking precision and robustness. Insights from this study provide valuable guidelines for real-time ADAS implementations leveraging LiDAR-camera data fusion.