基于神经网络学习的公共汽车驾驶安全识别

Safety Identification of Bus Driving Safety Based on Neural Network Learning

文章针对目前公交车驾驶员身体突发疾病、乘客抢夺方向盘以及现阶段L2级别的智能驾驶中驾驶权转移的空白等行车安全问题,设计了表情识别和危险动作识别两款程序。其中表情识别程序是利用基于Python语言的Opencv计算机视觉库,以Qt5作上位机设计的一款在视频中实时采集的人脸表情识别程序。首先通过输入图像的处理提高有关信息的可监测性,进而提高监测质量,其次进行表情模型训练和数据采集等,最终实现识别功能。同时还介绍了基于LabVIEW的危险动作识别程序。通过摄像头传入的视频信息,采用帧间差分法以及背景学习等算法监测出运动物体,实现监测范围内可能对驾驶安全构成威胁的外来物体的实时捕捉。实验数据表明,两款程序能够准确、快速地捕捉相应表情和危险动作,在解决上述问题的同时还大大提高了公共汽车的行车安全性。

This paper designs two programs,expression recognition and dangerous action recognition,such as the sudden physical illness of bus drivers,passengers grabbing the steering wheel,and the blankness of driving right transfer in the current L2 level intelligent driving.Among them,the expression recognition program is a face expression recognition program designed with Qt5 as the host computer using the Opencv computer vision library based on the Python language.First of all,the monitorability of relevant information is improved through the processing of input images,and the monitoring quality is improved,and then the expression model training and data acquisition are carried out,and finally the recognition function is realized.A LabVIEW-based dangerous action recognition program is also introduced.Through the video information transmitted by the camera,algorithms such as the inter-frame difference method and background learning are used to detect moving objects,so as to realize the real-time capture of foreign objects that may pose a threat to driving safety within the monitoring range.Experimental data show that the two programs can accurately and quickly capture the corresponding expressions and dangerous movements,which greatly improves the driving safety of buses while solving the above problems.