This paper proposes an adaptive interactive multiple models (IMM) tracking algorithm in order to improve tracking performance of strong maneuvering group targets in high measurement error. First of all, we introduce a strong tracking filtering algorithm with multiple suboptimal fading factors to improve the estimation accuracy of the group centroid in the strong maneuvering stage. Secondly, considering the influence of measurement accuracy on the extension state, the measurement error and extension state are formulated in a likelihood function, and then the error covariance of the measurement can be adaptively updated by using the innovation and the memory fading iterative process. Finally, we use the quasi-Bayesian approach to adaptively update the model transition probability. The model transition probability is modified by the measurement to suppress the non-matching model and amplify the matching model. By this way, the tracking model and targets state can be matched in real time. The simulation results show that, the proposed method is effective to improve the estimation accuracy of the centroid state and the extension state.