Research On Vehicle Drunk Driving Detection System Based On Artificial Olfaction

At present,there is a huge stock of various cars,and traffic accidents occur from time to time,many of which are caused by drunk driving,posing a serious threat to the safety of people’s lives.The existing vehicle drunk driving detection system has problems such as high price,poor anti-interference,large error,large volume,and privacy concerns.Therefore,this thesis studies an artificial olfactory based vehicle drunk driving detection system.In this thesis,a step-by-step vehicle drunk driving detection method was designed with the drunk exhaled gas as the target gas and the alcohol volatile gas,perfume volatile gas,etc.as the interference gas.The first step is to determine whether anyone in the car has drunk alcohol by detecting whether there is gas exhaled after drinking in the cabin;The second step is to ask the driver to take the initiative to exhale to the artificial olfactory system to judge whether the driver is drunk when it is confirmed that someone is drunk.This method can reduce interference with drivers,improve driver acceptance,and improve the accuracy of detecting drunk driving.The work carried out in this thesis is as follows:(1)According to the gas composition to be detected and the experience of artificial olfactory bionic measurement,the original sensor array of the artificial olfactory system for vehicle drunk driving detection was constructed,including 21 sensors,and the power module,sampling circuit module and bionic detection chamber with small mouth and large cavity were designed.The hardware platform of the artificial olfactory system for vehicle drunk driving detection was independently constructed.(2)According to the step-by-step vehicle drunk driving detection method,a drunk exhalation detection experiment in the cabin and a direct exhalation drunk driving detection experiment were designed and implemented,and the original data were obtained.After noise reduction and feature extraction,the extracted features were classified using the K-nearest neighbor,support vector machine,and random forest classification algorithm.Among them,the mean value of the wavelet transform coefficient,maximum value,and integral value were good,the highest classification accuracy can reach 93.44%.The experimental results show that the artificial olfactory drunk driving detection system in this thesis can accurately detect whether the driver is drunk driving.(3)To avoid feature redundancy and reduce the operation time cost of the drunk driving detection system,the classification features were optimized by the integration of the min-redundancy and max-relevance algorithm,random forest algorithm,recursive feature elimination algorithm,and particle swarm optimization algorithm,and finally,eight integral value features were selected as the classification features.And based on this,eight sensors with the best detection effect for drunk driving were obtained,namely TGS2603,MP-9,MP-3B,MP402,WSP2110,WSP7110,MP-7,and TGS2600.By using these 8 sensors to construct a new sensor array,the accuracy of drunk driving detection can be improved,reaching 94.44%.The volume and cost of the detection system have also been reduced,which is conducive to its better promotion.Design and implement a real breath drunk driving detection experiment,and the experimental results verify the effectiveness of the vehicle drunk driving detection system based on artificial olfactory designed in this thesis.It can achieve the detection of their drinking status under the condition of reducing the interference to drivers,which can meet the needs of Carsharing service providers to prevent drivers from drunk driving,and also meet the needs of other car manufacturers or consumers who want to install vehicle drunk driving detection devices.