基于D-S理论和BP网络的驾驶行为识别技术研究

Research on driving behavior recognition based on combination of D-S evidence theory with BP neural network

提出了BP神经网络与D-S证据理论相结合的驾驶行为识别、预测方法;将汽车行驶过程中人-车-路的信息作为BP神经网络的输入,利用BP神经网络对驾驶行为进行初步识别,并将BP神经网络输出的结果归一化处理后作为D-S证据理论的基本概率分布;利用证据距离理论对证据进行证据冲突处理,通过D-S证据组合理论对输入信息进行综合分析处理,决策识别出当前的驾驶行为;利用MATLAB语言编写了仿真测试程序,仿真结果表明该方法能够准确的识别出当前的驾驶行为。

An approach about driving behavior recognition and prediction is proposed based on combination of D-S evidence theory with BP Neural Networks.The human-vehicle-road information of car driving is used as the input of BP neural network to identify driving behavior preliminary.And the normalized output of BP Neural Networks is used as basic probability of D-S Evidence Theory.Then the theory of Evidence Distance is applied to deal with conflict evidences in order to recognize current driving behavior through analyzing and dealing with input information by combination rules of D-S evidence theory.Afterwards simulation and test program is compiled with MATLAB language,which result shows that the method can accurately identify the current driving behavior.