Non-destructive Test Blood Alcohol Content Of Driver Based On Near-infrared Spectrum

In the forensic pathology investigation of the traffic accidents all over the world, drunken driving is an important reason leading to traffic accidents. Detection and prevention of drunken driving draw more and more attentions in the world. At present, the mature methods of alcohol detection are breath alcohol detection and blood test. The blood test is invasive detection method and the breath alcohol detection is inaccuracy. On the contrary, near-infrared spectroscopy (NIRS) as a rapid, accurate, non-destructive detection method, compared with existing technology, has superior advantages. Near-infrared spectroscopy has been widely used to detect the human body substances, such as the detection of glucose, oxygen content, we can say that the technology has entered a mature stage of development, however, near-infrared spectroscopy to detect drivers blood alcohol content is still in the research stage.The purpose of this paper is to realize the noninvasive detection of the blood alcohol content of the drivers based on the near-infrared spectroscopy. The paper studies the near-infrared diffuse reflectance spectroscopy of blood alcohol content with the methods of wavelet analysis and partial least squares method and studies the law of the establishment and absorption of blood alcohol in the human body. A quantitative model based on the partial least squares method to detect the alcohol content is established and the stability and the precision are tested.The near-infrared spectroscopy signals in vivo have many noises, especially when the test is collected directly on the surface of the skin, which can not be directly used to establish the model. The pretreatment of the near-infrared spectroscopy we use in this paper is wavelet analysis. The paper compares the differences of the default de-noising method in matlab. The results are evaluated according to noise ratio and the root mean square error of different methods. From the results of de-noising, we determine the characteristics of the spectral wavelength of the alcohol ranges from1550nm to1800nm.With the spectral data based on wavelet analysis, partial least squares (PLS) method was utilized to establish the best mathematic calibration model that for describing the relationship between the near-infrared information and each component. The optimal principal components is determined by leave-one-out cross validation. The paper adopts the correlation coefficient (R), root mean square error (RMSEC) as the evaluation parameters of the calibration model, the prediction mean square error (RMSEP) and the average relative error (MREP) to evaluate the results of the prediction of unknown samples. And we evaluate the repeatability of the model, which turns out to be a good result. In other words, the model can be used to verify the alcohol concentration detection.At the end, the anti-drunk driving warning system is designed.On the hardware,use the S3C2410A as microcontroller as the master node,select MCP2515as the controller of ARM, select PCA82C250as a CAN transceiver;Draw the associated circuitry,and the preparation of the software program.When the driver realized the alcohol content exceeded, the alarm beeps, lights and emergency braking was completed,so as to effectively prevent drunk driving, to avoid unnecessary losses.