Research On Recognition Algorithm Of Drivers And Occupants' Abnormal Behavior Based On Deep Learning

The rapid development of online car-hailing has brought a good service experience to the society and the people,but it has also caused a large number of personal safety accidents for drivers and passengers.The abnormal behavior of the drivers and passengers in the online car-hailing is the main cause of these accidents.Existing identification methods are not comprehensive enough to identify behaviors.In most cases,they are based on post-event report analysis,which has low identification efficiency and poor applicability.The recognition method based on deep learning can analyze the advanced features in the image through the multi-layer neural network autonomous learning,and provides a new intelligent processing method for the recognition of abnormal behaviors of drivers and passengers.Therefore,it has certain practical application value to carry out research on the recognition algorithm of abnormal behavior of drivers and passengers based on deep learning.Based on the analysis of the behavior characteristics of the drivers and passengers in the car,this paper separately studied the use of depth learning methods to identify the abnormal driving behavior of the driver and the abnormal behavior between the driver and the passenger in the image;A convolutional neural network model with simple structure and fast convergence;in order to enhance the robustness of the network model to lighting,the data set is expanded by using data enhancement technology,and the key points of the human skeleton are introduced to improve the network structure and further improve the driving behavior Recognition accuracy: Aiming at the identification of abnormal behaviors between drivers and passengers,Alpha Pose is used to extract the key points of the human skeleton,and a sitting posture parameter model of the driver and occupants is constructed.According to the coordinate information in the sitting posture parameter model,a key point based on the human body is proposed.Probabilistic neural network recognition algorithm.Experiments were carried out on the public driving behavior data set and the self-built behavior data set between drivers and passengers,and the results proved that the algorithm proposed in this paper can effectively detect the abnormal behavior of drivers and passengers,and achieved the expected purpose.