Construction Of Heavy Truck Fatigue Driving Detection And Early Warning System

With the rapid development of the logistics industry,heavy-duty trucks play a very important role in the current transportation tasks.However,during the driving process,the driver may not be able to drive due to factors such as a lot of driving at night,long driving time,monotonous highway environment,and poor vehicle comfort.Operators are often more prone to fatigue,which can lead to major accidents.Therefore,establishing a real-time and accurate fatigue driving detection and early warning system can effectively reduce the fatigue driving of heavy truck drivers,reduce the occurrence of traffic accidents to a certain extent,and ensure the safety of drivers’ lives and property.Since the physiological feature detection in the detection method of the driver’s fatigue state is more intrusive to the driver,and the heavy goods vehicles are relatively concentrated in the transportation task at night,it will lead to the accurate detection of the facial feature detection method which is greatly affected by the environment and light.However,the detection method based on vehicle state information not only has high detection accuracy,but also is less affected by environmental changes and has low cost.Therefore,this paper chooses the vehicle-based state information more suitable for heavy-duty trucks to establish a fatigue driving detection model.In order to build a heavy-duty vehicle fatigue driving detection model with high accuracy and good real-time performance,this paper focuses on fatigue feature analysis and extraction,fatigue feature screening,fatigue feature fusion,The detection algorithm has been studied,the fatigue driving detection model has been established,and the models have been compared and analyzed.First,a fatigue driving simulation experiment was designed and carried out based on a simulated driver,and the vehicle state data of 10 experimental drivers driving in different fatigue states in the highway scene were obtained,including vehicle attitude angle,steering wheel angle,steering wheel angle rate,data such as speed and acceleration.On the basis of segmenting the vehicle state data,a fatigue state evaluation method based on the combination of subjective evaluation and objective detection is adopted to evaluate the collected sample data to construct a driver fatigue state sample database.Secondly,based on the selection of vehicle state data,the vehicle state data under three fatigue states are analyzed,and 12 fatigue characteristic parameters are extracted according to the hidden fatigue driving characteristics.In order to verify the significant difference between the extracted fatigue characteristic parameters and the fatigue state,the analysis of variance method was selected to screen the fatigue characteristic parameters.Then,in order to further optimize the fatigue characteristic parameters and improve the detection speed of the subsequent detection model,the unsupervised Principal Component Analysis(PCA)algorithm and the supervised Linear Discriminant Analysis(LDA)algorithm were used for fatigue characteristics.The parameters are fused with features,and the sample database after feature fusion is obtained.Then,based on the comparative analysis of several commonly used classification algorithms,the Support Vector Machine(SVM)and the Long Short-Term Memory(LSTM)neural network were selected to establish the fatigue driving detection model,and the selection of The data of 8 drivers is used as the training set,and the data of 2 drivers is used as the test set.The original extracted fatigue characteristic parameters,the PCA fusion fatigue characteristic parameters and the LDA fusion fatigue characteristic parameters are brought into the two models for training and testing,and compared and analyzed the accuracy and detection time of the model detection,so as to select a better fatigue driving detection model to build a fatigue driving early warning system.Finally,based on the respective advantages of the two selected models,a fatigue driving warning system is established by using the support vector machine fatigue driving detection model based on LDA data.At the same time,the detection results of the long-short-term memory neural network fatigue driving detection model based on LDA data are stored and remotely transmitted together with the results detected by the early warning system to provide technical data reference for transportation management departments to carry out transportation safety management.Then,according to the overall function and practical application of the early warning system,the hardware selection and software design of the early warning system are carried out respectively,and the feasibility of the early warning system is verified by Python programming.