A New Approach for Object Detection Using YOLO v11 and Trajectory Estimation Using an IMM Filter
Summary
No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.
Summarise with VIRAT Free to try with your Science Catalogs account.Abstract
This paper presents a trajectory estimation framework for tracking a maneuvering ego vehicle using the Interacting Multiple Model (IMM) filter. Ground-truth data, including position, velocity, and acceleration, is obtained from the KITTI dataset’s onboard sensor measurements. To simulate realistic observations, Gaussian noise with a standard deviation of 5.0 meters is added to the ground-truth trajectory, and these noisy measurements are used as inputs to the IMM filter. The IMM framework combines Constant Velocity (CV) and Constant Acceleration (CA) motion models to adaptively estimate the ego vehicle’s state over time. Root Mean Squared Error (RMSE) is used to evaluate tracking accuracy, and 50 Monte Carlo simulations are conducted to assess robustness under varying noise conditions. The initial design assumes both motion models are equally likely. For the baseline configuration, the process noise standard deviations are set to σ_CV = 0.1 and σ_CA = 0.1. In addition, in this work we implement early sensor fusion by projecting LiDAR point clouds onto camera images and performing object detection using YOLOv11. Although these detections are not yet used within the IMM framework, they establish a foundation for future work involving integrated tracking and fusion across modalities.
Fields of Study & Subjects
Target Tracking and Data Fusion in Sensor Networks
Autonomous Vehicle Technology and Safety
Inertial Sensor and Navigation
Publication details
- Journal / source
- International Conference on Consumer Electronics
- Pages
- 1–6
- Publication date
- 3 February 2026
- Language
- EN