To improve the recognition accuracy of roadside sensors and obtain high-quality vehicle trajectories, this paper proposes an adaptive Federal Kalman filtering fusion method for multi-source sensor acquisition data. This method aims to address the problems of insufficient data accuracy and inaccurate acquisition of vehicle position data when roadside multi-sensors are collecting data. First, the data collected by each sensor is unified in spatial and temporal dimensions and converted to Frenet coordinate system data. Next, the global nearest neighbor association algorithm is used to process redundant data from each sensor. Generate the trajectory set of each sensor for the same target trajectory. During the tracking process, the residual covariance is used to adaptively estimate observed noise, and the fluctuation correction coefficient is added to further optimize the Kalman filter gain by correcting observed noise. Finally, the optimal trajectory estimation for vehicle travel is obtained by optimally combining the target sequences using the Federal Kalman filter. Simulation experiments were conducted to verify the effectiveness of the proposed fusion model in this paper. The experimental results demonstrate a noticeable improvement in both position and velocity accuracy with the proposed method in this paper. Compared to the traditional federated Kalman algorithm, there is a 12.4% increase in position distance accuracy and a 62.5% improvement in velocity estimation accuracy. This improvement enables accurate tracking of target vehicles and provides high-quality data support for traffic safety research.
Roadside multi-sensor data fusion based on adaptive federal Kalman filtering
2023-08-04
1128212 byte
Aufsatz (Konferenz)
Elektronische Ressource
Englisch
Multi Sensor Fusion Based on Adaptive Kalman Filtering
Springer Verlag | 2017
|Fuzzy Adaptive Kalman Filtering for INS/GPS Data Fusion
British Library Conference Proceedings | 1999
|Fuzzy adaptive Kalman filtering for INS/GPS data fusion
AIAA | 1999
|British Library Conference Proceedings | 2004
|