基于联合投票网络的交通场景天气分类方法

基于交通视频监控图像的天气识别已经成为智能交通系统中重要的研究课题. 虽然卷积神经网络（convolutional neural network，CNN）在图像识别技术获得了巨大的发展，但是针对复杂交通场景的天气识别问题，现有的模型在特征表达方面仍然面临着巨大的挑战. 为了提取丰富的语义特征，提出了基于联合投票机制的深度神经网络（deep neural network，DNN）模型. 所提出的模型包括两个核心模块:基于通道和空间注意力机制的二阶特征模块和基于复合特征结果联合投票机制的分类模块, 用以提取不同天气图像中的判别性信息，提高在复杂交通场景下的天气识别性能. 最后，在两个基准天气分类数据集上进行了验证试验，结果表明:对于复杂场景条件下的天气识别问题，所提出的基于联合投票机制的深度神经网络模型的识别正确率优于目前最好的天气识别方法的1.97%. 

Weather Classification in Traffic Scene Based on Joint Voting Network

Weather classification based on road monitoring images has become an important research topic in intelligent traffic system. With the application of convolutional neural network (CNN), image recognition has been greatly developed. However, the existing deep learning methods still face great challenges in weather recognition of complex traffic scenarios. A novel deep neural network (DNN) model based on joint voting framework is proposed to extract rich semantic features. The proposed model consists of two core modules: the second-order feature module based on channel and spatial attention mechanism and the joint voting classification module based on composite features, which can extract discriminant information from different weather images and improve the weather recognition performance in complex scenarios. Extensive experiments conducted on two benchmark weather classification datasets demonstrate that the proposed joint voting DNN outperforms the existing weather recognition method by 1.97%.