Research On Detection Of Hazardous Material Using Chemometrics And Optical Methods

In recent years, drug smuggling is increasingly rampant all over the world. Among them, human body packing has turned to be one of the main routes for trafficking drugs. On the other hand, the terrorist activity using explosives has become a new kind of crime, and it become increasing intense, leading to a great loss. The explosives include two types:one is traditional explosive, such as TNT; the other is new type of explosive, such as liquid explosive. Nowadays, people and goods move very fast and frequently. In order to make balance between efficiency and safety, detecting drugs and explosives hidden in people and goods is closely related with the national security needs, and it is an important technological problem to be solved.At present, there are some well-developed technologies and products for on-site detection of drugs/explosives, such as X-ray transmission, ion mobility spectrometry, neutron analysis detection. At the same time, there are some booming technologies, such as X-ray cone beam three-dimensional imaging technology. These techniques achieve certain effects in practical applications, but there are still some problems as follows:they can only recognize the shape or density of the material, and can not obtain the intrinsic characteristics of the target; they are influenced by cover or package largely; their detection time is long; the instrument is huge and precious; the sample process will damage the target. Therefore, it is necessary to explore new techniques and instruments for fast and accurate on-site detection of drugs/explosives.Due to their advantages, such as non-destructive, high selection and high efficiency, energy dispersive X-ray diffraction (EDXRD) and energy dispersive X-ray scattering (EDXRS) become the focus in the field of hazardous material on-site detection. On the other hand, the optical waveguide technology can be highly sensitive for monitoring the reactions occurring on the interface because of evanescent field principle. At the same time, the CCD with the capability of time resolution helps optical waveguide technology study the process of absorption desorption kinetics on the interface. Therefore, it can not only detect hazardous material, but also research the physical and chemical property on the interface.In the on-site detection of drugs/explosives, due to the presence of cover or package and complex environment, the obtained signal often has poor signal-to-noise ratio, and the intrinsic characteristics of the target substance is often submerged in the signal of background. Because the target material and background material has the similar distribution in the frequency spectrum, the traditional signal processing methods, such as the discrete Fourier transform or wavelet transform is not suitable here. In this case, the finite impulse response filter or wavelet filter will not be able to effectively filter the noise. Therefore, there is an urgent need to use chemometric methods to extract features from the signal. These features can be seen as the "fingerprint" of the target. On the basis of these features, pattern classification can be applied to detect drugs/explosives in the complex environment.This thesis focuses on the detection of hazardous material, including drugs and explosives, especially for hazardous material hidden in the human body or complex environment. The detection technology based on the combination of optical methods and chemometrics is proposed to solve the problem of feature extraction and recognition in the case of complex background and high-dimensional variables, poor signal-to-noise ratio signal. The technology can fast and accurately detect hazardous material in the human body packing.The major work includes the following aspects:1. In order to study human tissue’s interference to hazardous material’s EDXRD signal, feature space of principal component analysis (PCA) is proposed to analyze the impact of fat and muscle on heroin identification. It is found that the impact of fat and muscle lies in moving the feature points between pure materials in the space of principal components, which shows that PCA can make the signal separation between heroin and skin.2. In order to study the signal model of hazardous material, partial least-squares (PLS) is proposed to set up the model of heroin covered by fat and muscle of different thicknesses, and a linear relationship is proved to be suitable in the case of setting fat and muscle as two separate factors.3. In order to solve the difficulty of physical interpretation of features extracted by PCA, nonnegative matrix factorization (NMF) is applied to extract features from hazardous materials’ EDXRD spectra. The extracted features are physically meaningful.4. Design identification software of hazardous material based on EDXRD according to the requirement of actual environment.5. Principal components regression (PCR) and partial least squares regression (PLSR) are applied for regression analysis for spectra of liquid hazardous material. It is found that the extracted features from samples of different concentrations follow linear model.6. In order to solve the difficulty of incomplete data showing using traditional method, color temperature is used to visualize the data of optical waveguide. PCA and NMF are applied to extract features from spectra of TNT. The extracted features efficiently express the main information of original data, and entirely show the absorption process of TNT on the surface of optical waveguide glass.