In conventional interactive multiple model (IMM) algorithm, the fixed transitionprobabilitymatrix results in slow model switching and low tracking accuracy. Existing adaptivetransitionprobability IMM algorithms st...
详细信息
Accurate tracking of maneuvering extended object remains a significant challenge in high-precision sensing and radar applications due to inaccurate state estimation resulting from the complex maneuverability of the ob...
详细信息
Accurate tracking of maneuvering extended object remains a significant challenge in high-precision sensing and radar applications due to inaccurate state estimation resulting from the complex maneuverability of the object, difficulties associated with model switching, and filter divergence. To address these issues, a maneuvering extended object tracking approach is proposed based on generalized conversion nonlinear filtering with an adaptive transition probability matrix and random weighted cubature rule sampling (MEOT-ARWGCF). Initially, algorithm accuracy is enhanced by incorporating random weighted factors in place of the original weighted factors, and the improved RWGCF filtering algorithm is utilized for maneuvering extended object tracking. Furthermore, to mitigate the issue wherein a fixed Markov matrix often fails to correspond with the object's current motion state, an adaptive transition probability matrix (ATPM) is introduced, effectively reducing tracking errors and alleviating the divergence problem caused by transition errors between models. Finally, the proposed algorithm is implemented for maneuvering star-convex extended object tracking, ensuring tracking algorithm stability. In comparison to MEOT-UKF and MEOT-CKF, the MEOT-ARWGCF method enhances RMSE position accuracy by 18.21% and 21.12%, velocity accuracy by 1.62% and 3.21%, and Hausdorff distance accuracy by 10.89% and 12.57%, respectively. These results indicate that the proposed method achieves superior precision in maneuvering extended object tracking.
暂无评论