基于深度学习的驾驶员状态识别

提出了基于驾驶员脸部及周围信息的驾驶员状态检测方法。文章通过实车摄像头采集了驾驶员驾驶状态视频数据,利用Dlib和OpenCV库对采集的驾驶员图像进行脸部检测,基于驾驶员脸部数据建立了深度学习数据集,然后基于该数据集设计了一种卷积神经网络模型FaceNet,利用PyTorch深度学习框架在数据集上对模型进行训练,最终得到了有较高准确率的驾驶员状态检测模型,其可识别抽烟、睡觉、左手打电话和右手打电话四种驾驶员状态。

Driver State Recognition Based on Deep Learning

A driver state detection method based on driver's face and surrounding information is proposed.In this paper,the driver's driving status video data is collected by the real vehicle camera.The captured driver image is detected by Dlib and OpenCV library.The deep learning data set is built based on the driver's face data,and then based on the data set.A convolutional neural network model,FaceNet,uses the PyTorch deep learning framework to train the model on the data set,and finally obtains a driver state detection model with higher accuracy,which can identify smoking,sleeping,left-handed calling and right hand.Call four driver states.