基于近红外光谱透射法的汽车驾驶员血液酒精含量无损检测

发布时间：2019-03-12 18:55

【摘要】：在世界各国交通事故的法医学调查中,酒后驾驶是导致交通事故发生的重要原因,醉酒驾驶的检测与预防越来越受到世界各国的重视。目前,较为成熟的酒精检测方式为呼气式酒精检测和抽血检验,由于抽血检验属于有创伤的检测方法,呼气法检测又存在检测精度的问题,而近红外光谱法作为一种快速、准确、无损的检测方式,与现有的技术相比,有不可替代的优势。近红外光谱法已经被广泛的用来检测人体内的物质,如葡萄糖、血氧等含量的检测,可以说该技术已经进入到了成熟发展阶段,但是国内外近红外光谱法检测驾驶员血液中酒精含量尚处于研究阶段。本文的目的是实现驾驶员体内酒精含量的近红外光谱法无损检测,并与汽车驾驶系统结合,实现防醉酒驾驶。利用小波分析与偏最小二乘法对人体内酒精的近红外漫反射光谱进行了分析与研究,并在此基础上分析血液酒精在人体内随时间衰减以及吸收的规律,建立了偏最小二乘法检测人体组织内酒精含量的定量模型,并对所建立模型的预测精度及稳定性进行检验。人体内酒精近红外光谱信号含有较强的噪声,特别是直接在人体皮肤表面采集的信号,噪声干扰十分强烈,不能直接用于建立模型,本文采用采用小波变换以及数据平滑对酒精近红外光光谱进行预处理。选择硬阈值条件下,matlab缺省去噪方法,对所得光谱数据进行了去噪处理,并确定了酒精的透射特征光谱波长范围在1550nm-1800nm左右。根据小波分析等预处理之后的光谱数据,采用了偏最小二乘法建立定量校正模型,留一交叉验证法确定最佳主成分数。采用相关系数(R)、均方根误差(RMSEC)作为校正模型的评价参数,未知样本的预测结果采用预测均方差(RMSEP)及平均相对误差(MREP)进行评价,并对模型的重复性进行验证,实现了对未知酒精浓度的检测。最后,设计了防醉酒的驾驶预警系统。在硬件方面,使用S3C2410A作为主控节点的微控制器,选择MCP2515作为ARM主控节点的CAN控制器,选择PCA82C250作为CAN收发器,绘制了相关的电路,并编写了对应执行机构软件程序,实现了当驾驶员体内酒精含量超标时,完成发出报警提示音、点亮报警灯以及紧急制动等动作,从而有效的防止醉酒驾驶,避免不必要的损失。
[Abstract]:In the forensic investigation of traffic accidents all over the world, drunk driving is an important cause of traffic accidents. The detection and prevention of drunken driving has been paid more and more attention by many countries all over the world. At present, the more mature methods of alcohol detection are expiratory alcohol detection and blood sampling test. Because the blood sampling test belongs to the invasive detection method, the detection accuracy of breath method is also a problem, and the near infrared spectroscopy is a fast method. Accurate, non-destructive testing method, compared with the existing technology, has irreplaceable advantages. Near infrared spectroscopy (NIR) has been widely used to detect substances in human body, such as glucose, blood oxygen, etc. It can be said that the technology has entered the mature stage of development. However, near infrared spectroscopy (NIR) is still in the research stage to detect the alcohol content in the driver's blood at home and abroad. The purpose of this paper is to realize the nondestructive detection of alcohol content in the driver's body by near infrared spectroscopy (NIR), and combine it with the automobile driving system to realize the anti-drunken driving. The near infrared diffuse reflectance spectrum of alcohol in human body was analyzed and studied by wavelet analysis and partial least square method. On the basis of the analysis, the attenuation and absorption of blood alcohol in human body with time were analyzed. A quantitative model for the determination of alcohol content in human tissues by partial least square method was established and the prediction accuracy and stability of the model were tested. The near infrared spectrum signal of alcohol in human body contains strong noise, especially the signal collected directly on the surface of human skin. The noise interference is very strong and can not be used to establish the model directly. In this paper, wavelet transform and data smoothing are used to pre-process the near-infrared spectrum of alcohol. Under the condition of hard threshold, the matlab default de-noising method is selected, and the spectral data are de-noised, and the wavelength range of the transmission characteristic spectrum of alcohol is determined to be about 1550nm-1800nm. According to the pre-processed spectral data such as wavelet analysis, the partial least square method is used to establish the quantitative correction model, and a cross-check method is used to determine the optimal principal fraction. The correlation coefficient (R), root mean square error (RMSEC) is used as the evaluation parameter of the calibration model. The prediction results of the unknown samples are evaluated by the prediction mean variance (RMSEP) and the average relative error (MREP), and the repeatability of the model is verified. The detection of unknown alcohol concentration is realized. Finally, an anti-drunken driving warning system is designed. In the aspect of hardware, S3C2410A is used as the microcontroller of the main control node, MCP2515 is chosen as the CAN controller of the ARM master node, and PCA82C250 is chosen as the CAN transceiver. The relevant circuits are drawn, and the software program of the corresponding executing mechanism is programmed. When the alcohol content of the driver exceeds the standard, the actions such as issuing alarm tone, lighting the alarm light and emergency braking are completed, so as to effectively prevent drunken driving and avoid unnecessary loss.
【学位授予单位】：山东大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：U492.8;TN219

【参考文献】