摘要
目前研究的驾驶员异常行为预警方法,由于用户对于预警信号的理解能力较差,不能在短时间内迅速作出预警反应,导致安全风险较高。为了解决上述问题,提出基于标签集相关性学习的驾驶员异常行为预警方法。分别采用车辆颜色特征、纹理特征以及梯度特征作为车辆异常表征分析,通过梯度特征捕获到驾驶员的瞬时行为,研究颜色特征、纹理特征以及梯度特征之间存在的关联性,利用协方差描述因子对图像数据帧区域的信息进行获取,然后按照各个特征的特点,进行特征提取。应用标签集相关性学习方法,对驾驶员异常行为等级进行划分,设定三种异常行为等级,根据相关等级,给出不同的预警信号,采用语音、视觉以及感知三重叠加预警模式确保用户可以在短时间内发现异常行为。实验结果表明,基于标签集相关性学习的驾驶员异常行为预警方法能够帮助用户快速发现自己的异常行为,用户预警反应时间提高了50%以上,对于用户安全驾驶有重要意义。
At present,the early warning method of abnormal behavior of drivers is studied,but the user’s ability to understand the warning signal is poor,and the early warning response cannot be made quickly in a short time,resulting in high safety risks.In order to solve the above problems,an early warning method of driver’s abnormal behavior based on label set correlation learning is proposed.Vehicle color features,texture features and gradient features are used as vehicle anomaly characterization analysis,and the instantaneous behavior of the driver is captured by the gradient features,the correlation between color features,texture features and gradient features is studied,and the information of the image data frame area is obtained by using the covariance description factor,and then the features are extracted according to the characteristics of each feature.The label set correlation learning method is applied to divide the driver’s abnormal behavior level,set three abnormal behavior levels,give different warning signals according to the relevant level,and use voice,vision and perception three overlapping plus warning modes to ensure that users can find abnormal behavior in a short time.Experimental results show that the driver abnormal behavior early warning method based on tag set correlation learning can help users quickly discover their abnormal behavior,and the user warning response time is increased by more than 50%,which is of great significance for user safe driving.
作者
李光华
张洪涛
谢凤祥
韩名亮
LI Guanghua;ZHANG Hongtao;XIE Fengxiang;HAN Mingliang(State Energy Daduhe Big Data Service Co.,Ltd.,Chengdu 610041,China)
出处
《电子设计工程》
2023年第2期160-165,共6页
Electronic Design Engineering
关键词
标签集
相关性学习
驾驶员行为
异常行为
行为预警
tag set
correlation learning
driver’s behavior
abnormal behavior
behavior warning
