在以Ｋａｌｍａｎ滤波为基础的各种被动式跟踪算法中，一直存在初值敏感性难题，由于实际系统中初值往往难以较精确给出，导致滤波器的发散，本文给出了修正的极坐标系下的推广卡尔曼滤波算法，ＭｏｎｔｅＣａｒｌｏ仿真表明该算法对初值选取不敏感，具有推广应用价值。 In the passive tracking algorithm based on Kalman filtering, there is always a problem of initial value sensitivity. Because the initial value of the actual system is often difficult to be given more accurately, resulting in the divergence of the filter, a modified polar coordinate system Under the promotion of Kalman filter algorithm, MonteCarlo simulation shows that the algorithm is not sensitive to the initial selection, with the promotion of application value.