Improved IMM Algorithm Based on Deep Learning for Maneuvering Target Tracking

To address the issues of model switching lag and insufficient accuracy in the Interacting Multiple Model (IMM) algorithm for tracking complex maneuvering targets, this paper proposes a deep learning-enhanced improvement method. By designing a deep learning model to generate prior information of model probabilities in real-time and constructing an adaptive fusion mechanism to integrate this information into the IMM framework, the proposed method overcomes the limitations of traditional approaches that rely on fixed transition probability matrices and process noise parameters. Theoretical analysis and application cases demonstrate that the proposed method significantly improves model switching speed and state estimation accuracy without significantly increasing computational complexity.