中文摘要：

针对交互多模型（interacting multiple model，IMM）在多机动目标跟踪算法中存在的缺陷以及目标跟踪精度问题，提出了基于变结构多模型（variable structure multiple model，VSMM）的高斯混合基数概率假设密度（Gaussian mixture cardinalized probability hypothesis density，GMCPHD）滤波算法。该算法利用了VSMM具有自适应性、时变性的特点，达到了在某一时刻能够选取与目标运动模式相匹配的模型集合的目的，相比于IMM考虑的仅是固定的模式集合具有很强的优越性。此外，GMCPHD滤波算法不仅避免了数据关联问题，而且通过高斯分布递推PHD函数的同时递推基数分布。最后，利用雷达作为传感器，对跟踪机动目标进行仿真，证明VSMM相比于IMM对于多机动目标跟踪更具有优越性，同时验证了VSMM—GMCPHD滤波算法具有提高机动目标跟踪精度，减小跟踪误差的作用。

英文摘要：

To deal with the defects and the target tracking precision problem in the interacting multiple model （IMM） algorithm for multiple maneuvering targets tracking, a Gaussian mixture cardinalized probability hypothesis density （GMCPHD） filter algorithm based on variable structure multiple model （VSMM） is pro- posed. Compared with the IMM algorithm which only considers the fixed model collection, the GMCPHD filter algorithm is superior. Utilizing the adaptive and time-varying which both are the characteristics of VSMM algo- rithm, this approach reaches the goal that the model collection matching the target motion model can be selected in certain time. In addition, the GMCPHD filter algorithm not only avoids the data association problem, but al- so propagates the radix distribution while propagates PHD function by using Gaussian distribution. Finally, a radar is chosen as the sensor and some simulation experiments on tracking a variety of maneuvering target are done. The simulation results prove that the VSMM algorithm is superior to IMM algorithm for multiple maneu- vering targets tracking and illustrate that the VSMM-GMCPHD filter algorithm can improve the maneuvering target tracking precision and reduce the tracking error.