Research On Road Risk Assessment Technology Based On Deep Learning

With the rapid development of the road traffic system,the daily travel of Chinese residents is becoming increasingly convenient,but the incidence of traffic accidents is also increasing rapidly.It not only threatens the lives and property of residents but also seriously affects the stable development of the economy and the stability of the country.Therefore,it is urge to put more attention and research on this issue for the development of traffic system.Road risk assessment technology predicts the severity of traffic accidents according to the characteristics of the accidents,which can help to find the significant influencing factors of accidents and the potential law of accidents.It is beneficial to prevent traffic accidents and improve travel safety.Most of the researches on road risk assessment are based on traditional machine learning techniques which is high cost and hard to process tasks with a large amount of data.However,the road risk assessment method based on deep learning can overcome those limits.It can automatically obtain valuable information in the data through neural networks and it can handle tasks with a large amount of data.The model combining the attention mechanism and convolutional neural network based on deep learning-related theories is proposed in this paper and it is applied to the assessment of the severity of road traffic accidents.The performance of the model is proved according to the experiments.The main research work and innovations of this article are as follows:(1)Traditional road risk research often uses traffic accident data of a certain road segment or a local area as the research sample which has some limitations.The UK car accidents and the US accidents datasets used in this paper contains abundant data.It also has wide range of accident coverage and the characteristics of the accident are complex.It is more convincing to carry out related research based on these datasets.(2)The analysis of traffic incidents causation based on real traffic accident datasets is conducted to explore the potential laws of road traffic accidents.The relationship between relevant factors and road traffic accidents is discussed and it is important to the model which is proposed to predict the severity of traffic accidents.(3)A model combining the self-attention mechanism and convolutional neural network is proposed.The self-attention mechanism is used to recognize the key features of road traffic accident,which improves the accuracy of the model.The parameter optimization experiment is designed to determine the parameter settings of the model.At the same time,the comparing experiment is performed to verify the effectiveness of the model.The experimental results show that the performance of the model is better than the comparing experiments.(4)Based on the research of the self-attention mechanism,a model combining multi-head attention mechanism and convolutional neural network is proposed.The multi-head attention mechanism can focus on more key features,thereby further improve the accuracy of the model.The experiment is conducted to analyze the advantages and disadvantages of the two models.In addition,the feature focused by the two attention mechanisms is discussed and the conclusion that road factors and vehicle factors may have a greater impact on road accidents in the experimental settings of this paper is concluded.