Research On Fatigue Detection Of Agricultural Machinery Drivers Based On Deep Learning

At the present stage,the level of domestic agricultural mechanisation has been increasing,with the regional coverage rate of agricultural mechanisation reaching over 70%.With the increase in the usage rate of agricultural machinery,the accident rate of agricultural machinery is also on the rise,among which agricultural production accidents caused by fatigue driving account for a greater proportion.After a long period of development,fatigue driving detection technology has become an effective early warning means for safe driving,but less application and research in the field of agricultural machinery,and fatal agricultural machinery accidents occur from time to time.Therefore,it is necessary to conduct research on fatigue driving detection of agricultural machinery,reduce agricultural machinery production accidents caused by fatigue driving,improve the efficiency of agricultural machinery operations and protect the security and property of agricultural machinery drivers.The thesis addresses the characteristics of agricultural machinery operations,draws on fatigue driving detection technology in related fields,and conducts research on fatigue driving detection of agricultural machinery,with the main work as follows.(1)The research status,achievements and development trend of fatigue driving detection technology and fatigue assessment system at home and abroad are systematically analyzed,especially the research on fatigue driving detection methods in agricultural machinery field was summarized.The concept of fatigue and fatigue driving is explained in the paper,and the definition of fatigue driving in agricultural machinery is derived from it.This paper studies the main factors of fatigued drivers of agricultural machinery,and analyzes the fatigue characteristics of agricultural machinery drivers and the exercise table of agricultural machinery.The principles and structures of the Dlib open source framework,convolutional neural networks,and the LSTM and Bi LSTM networks in recurrent neural networks are introduced successively.(2)The main factors that cause fatigue in agricultural machine drivers are investigated,and the main representations of fatigue in agricultural machines are analysed to lay the foundation for the detection of fatigue in agricultural machines.The law of change of farm machinery fatigue driving representation with time is studied,the farm machinery fatigue driving questionnaire is designed and carried out,the weight of fatigue representation is calculated according to the statistical results,the fuzzy judgment matrix is designed,and the fuzzy hierarchical evaluation system of farm machinery driving fatigue is constructed.(3)The fatigue discrimination model for agricultural machinery drivers(AD-FDB)was constructed based on deep learning.The model is composed of two inputs: at one end is face recognition and key point localization based on Dlib;at the other end,a CNN-Bi LSTM structure is constructed based on convolutional neural network and bidirectional long and short-term memory network to extract the spatial and temporal features of the driver’s face.(4)The data set AMFD-DDS was constructed to verify the effect of the model.Experimental results show that the accuracy of the proposed model reaches98.85%,which is better than other detection methods in terms of accuracy and robustness.Furthermore,the accuracy of the fatigue quantization and grading evaluation model designed in this paper is illustrated.