中文摘要：

隧道作为高速公路上较为特殊的地段,其安全性受到更多重视。文中从高速公路隧道内行车安全管理的角度出发,分析了隧道内影响行车安全的各种异常状态,指出了其中车辆停驶是影响安全最不利的状态。在此基础上着重开展了针对隧道内停车异常的视频图像识别方法研究。为增强车辆识别能力,提出了将图像纹理特征、几何特征和边缘特征进行组合描述车辆特征的方法。以BP神经网络作为基分类器,通过Adaboost算法得到多个BP网络弱分类器组成的强分类器作为车辆识别模型,并对该识别模型的效果开展了实地实验验证。结果表明,本文所提出的模型对停车异常具有良好的识别效果。

英文摘要：

As a special location on the highway,the security of tunnel is paid more attention than the other area.In this paper,from the traffic safety management in highway tunnel we first analyzed the various abnormal states of the traffic in tunnels among all the states.The stopping car was the most harmful status to the safety in the tunnel.Research then focused on the video recognition method of the abnormal stopping car in highway tunnel.In order to enhance recognition capacity of vehicle,image texture,geometric,and edge features were extracted in video sequences and these features were combined into a new feature method.The combined feature method was input into BP neural network as the base classifier and the strong classifier was obtained by integrating BP neural networks under the Adaboost framework.A testing in a real tunnel was performed to verity the effectiveness of the method.Results show that the presented model has a satisfactory recognition effect for the abmormal state of the stopping car.