Safety Monitoring And Warning System For Vehicle Drivers Based On Android System

In recent years,with the increase of vehicles,traffic accidents have become more and more frequent,and traffic accidents caused by fatigue driving have become one of the main reasons.At present,the detection devices for driver fatigue driving are mostly contact type,which is complicated to install and costly,and is not suitable for popular use.Based on the analysis of existing driver fatigue detection devices,this paper designs a driver safety monitoring and early warning system based on HOG-LBP feature fusion.The system first uses the infrared high-definition camera to obtain the driver's face video,locates the face position by Seetaface Detection face detection method,and then uses the improved self-encoder network method to achieve 27-point face key point positioning.On this basis,the state recognition of the eye is performed according to the HOG-LBP feature fusion method,and finally the driver fatigue determination is performed according to the neural network method.Based on the fatigue algorithm,a driver safety monitoring and early warning system based on Android system is designed.The main contents of this paper are as follows:(1)Face detection.Two common face detection methods are compared and analyzed: Adaboost face detection method and Seetaface face detection method.The principle,advantages and disadvantages,and operation speed of the two algorithms are analyzed in detail,and the system is selected.The face detection algorithm used.(2)Positioning of key points of the face.Based on the cascading regression shape,depth learning and self-coded network based face key location method,a 27-point key point improved self-coding network face key point location method is proposed,which improves the positioning accuracy of the human eye and mouth.(3)Human eye state recognition and fatigue judgment.Based on the key point location,a human eye state recognition method based on HOG-LBP feature fusion is proposed..Firstly,the ability of HOG features and LBP features to describe the state of key points is analyzed separately.It is found that HOG features can describe the state of key points such as eyelids and pupils well.LBP features can describe the local information of key points very well.And the robustness to lighting is very strong.Therefore,combining the advantages of the two feature descriptors,this paper proposes a human eye state recognition method based on HOG-LBP feature fusion for the recognition of human eye state.On this basis,the driver fatigue state is obtained by neural network operation.(4)The implementation of driver safety monitoring and early warning algorithms.This paper focuses on the face detection,eye state discrimination and fatigue state discrimination,and finally implements these algorithms on the Android-based driver driving safety monitoring and early warning system.