卡尔曼滤波是目标跟踪系统的基础。但由于目标动态模型和参数的不确定性以及传感器测量噪声等因素的影响,使得卡尔曼滤波器对目标状态的估值含有较大误差。本文将卡尔曼滤波器的估值能力与视经网络的学习能力相结合,在很大程度上提高了卡尔曼滤波器的跟踪精度,并提出了神经网络在多传感器数据融合中的应用。仿真结果表明,神经网络能大大改善卡尔曼滤波器的性能 Kalman filter is the basis of the target tracking system. However, due to the uncertainty of the target dynamic model and parameters as well as the influence of the sensor noise and other factors, the Kalman filter has a large error on the estimation of the target state. In this paper, we combine the estimation ability of Kalman filter with the learning ability of visualization network to improve the tracking accuracy of Kalman filter to a great extent and put forward the application of neural network in multi-sensor data fusion. Simulation results show that the neural network can greatly improve the performance of Kalman filter