调试通过可以使用，多目标跟踪的粒子滤波器，BP神经网络用于函数拟合与模式识别，是一种双隐层反向传播神经网络，利用自然梯度算法，基于分段非线性权重值的Pso算法。- Debugging can be used, Multi-target tracking particle filter, BP neural network function fitting and pattern recognition, Is a two hidden layer back propagation neural network, Use of natural gradient algorithm, Based on piecewise nonlinear weight value Pso algorithm.