Enhancing Maneuvering Target Tracking with a Proposed Dual Parameterized Extended Kalman Filter in the IMM Framework

Tracking multiple targets in dynamic settings poses significant challenges in fields like robotics and surveillance. This process requires detecting and estimating object trajectories while addressing motion variability, measurement noise, and interactions. Traditional methods, such as the Kalman Filter, excel in single-object tracking but face difficulties in multi-target situations due to proximity ambiguities and rapid movement changes. The Interacting Multiple Model (IMM) framework enhances tracking by considering various motion hypotheses, yet traditional IMM implementations often use predefined probabilities, which can limit adaptability in highly dynamic environments. This paper introduces a novel tracking framework integrating three models using Dual Parameterized Extended Kalman Filters (DPEKFs). The first model assesses the Extended Kalman Filter (EKF) outputs based on likelihood estimates using the Gaussian likelihood function. The second adapts transition probabilities according to target accelerations, improving handling of abrupt motion changes. The last incorporates a modified Kalman gain to improve the fusion of estimates from the two EKFs within the IMM framework. The models performance is assessed using mean square error (MSE), highlighting their tracking accuracy and adaptability in diverse scenarios.