| With the wave of deep learning,artificial intelligence has been pushed to the top of history again,and technologies such as unmanned driving have naturally become one of the hot spots of the entire industry in recent years.The unmanned driving system is a very large and complex comprehensive system composed of various advanced modules.While satisfying the basic driving functions,it is more important as an intelligent system that needs to interact with humans and analyze human driving intentions.Human inten-tions are incorporated into driving strategies.At the same time,in order to be able to more safely ensure the safety of the driving process and the correct execution of driving intentions,the intelligent driving system needs to be able to sense and analyze the status of the driver/passenger,so as to ensure the correct delivery of commands and the relative safety of the driving process,herefore,this article starts from two aspects to establish a driving decision analysis system based on deep learning.On the one hand,it is based on the driver’s external state detection data,and on the other hand,based on the driver’s own intention information,combining these two aspects to construct a credible driving intention analysis system:1.Decision analysis method based on driver’s external data(a)A lightweight driving state analysis system based on deep learning,which can analyze the driver’s state and fatigue in the cockpit in real time,and give an alarm in time when an abnormal state occurs,so as to avoid unpleasant events..Based on deep learning technology,this paper develops a driving state classification algorithm based on MobileNet,and proposes a weighted pooling method to improve the overall classification accuracy,and integrates fatigue detection algorithms based on visual key point capture to ensure the safety of driving state.Based on the data obtained by these drivers’ external detectors,passive analysis and decision-making are carried out to ensure driving safety.(b)This paper investigates and analyzes some of the current domestic and foreign practical fatigue monitoring algorithms based on vision in detail,this part is mainly used as a supplementary part of the driving state analysis system.Under the condition of ensuring accuracy and practicability,the key points of the face are extracted based on the computer vision technology ensemble feature engineering,and then the key point information of the face is used to organize a set of reliable algorithms to define fatigue State,and designed a detailed experimental process to prove the practicability of the method.2.Decision analysis based on the driver’s own intentionA driving intention analysis and understanding system based on deep learning.The system is based on a deep learning knowledge system,combined with computer vision technology and natural language processing technology to build a scene understanding model of multi-modal information fusion.The system will be based on the instructions given by humans in the scene.To find the relative goal.In order to achieve high-precision and efficient driving intention analysis and understanding algorithms,this paper proposes extractors for different modal features and attention mechanisms under different modal-ities,combined with deep learning-based speech recognition technology and target de-tection technology to build The entire analysis and understanding system.Based on the driver’s active interaction demands,conduct driving intention analysis to guide unmanned driving,making unmanned driving more convenient,intelligent and practical.The two core contents of this paper are designed to serve for future-oriented driverless driving.Based on intent analysis to achieve human-machine-environment tripartite inter-action,people can intervene in the process of driverless driving,thereby trusting driver-less driving and solving the problem of unmanned driving.The purpose of driving au-tonomously.And use the driving state analysis system to ensure that the driver is in a reliable state,and can deal with abnormal situations that occur,ensuring the safety and reliability of the overall driving process.Therefore,the realization of the driving decision analysis system proposed in this paper integrates the driver’s intention analysis technol-ogy and the driver’s state monitoring technology,so as to build a reliable,reliable and interactive future unmanned driving technology. |
