基于局部信息和卷积网络的分心行为识别

Distraction behavior recognition based on local information and convolutional network

针对车载设备采集驾驶员图像中包含大量冗余信息,导致驾驶员行为识别准确率低的问题,提出一种使用驾驶员局部定位信息来帮助卷积神经网络识别驾驶员分心驾驶的方法。利用驾驶员局部定位信息,可以减少冗余信息干扰,卷积神经网络能够更加有效地关注驾驶姿态。在VGG16、InceptionV3、Mobilenet分类模型的基础上,探讨将驾驶员局部信息作为卷积输入的识别率,实验结果表明,在相同的数据集,对上述三个模型均有不同程度的提升,最高识别率可提升至96.10%。

The driver image collected by on-board equipment contains a large amount of redundant information,resulting in low accuracy of driver behavior recognition,and for this kind of problem,a method for helping the convolutional neural network identify the driver’s distracted driving by using the local location information of the driver is proposed in this paper.With the local positioning information of the driver,the interference of redundant information can be reduced,and the convolutional neural network can pay more attention to the driving attitude.On the basis of VGG16,InceptionV3 and Mobilenet classification models,this paper discusses the recognition rate of taking the driver’s local information as the convolution input.The experimental results show that in the same data set,the above three models are improved to different degrees,with the highest recognition rate of 96.10%.