This article focuses on the cooperative localization of maneuvering targets on the ground from unmanned aerial vehicles (UAVs) with bearing-only measurements. A maneuvering target tracking algorithm is proposed based on the interactive multiple model (IMM) framework. In order to fully utilize the localization redundancies and improve the accuracy of target localization in multi-UAVs system, a belief factor for state fusion is proposed considering the positioning and measurement capabilities of each UAV. Moreover, a trajectory smoothing algorithm is proposed by the quasi-uniform B-spline (QUBS) method, which is utilized for reducing the localization error caused by measurement noise. By calculating the discrete curvature of the maneuvering target trajectory, an adaptive identification method of turn rate is derived to alleviate the delay of the model switch. Besides, a maneuvering target tracking model is established and a unified framework based on the IMM method is proposed to improve the applicability of the maneuvering target tracking. Simulation results demonstrate that the proposed turn rate identification based on IMM (TRIIMM) algorithm outperforms the other algorithms in terms of tracking accuracy and takes the least time cost.