论文部分内容阅读
为快速了解和掌握海面溢油的种类 ,以便采取应急措施 ,提出了近红外光谱技术结合模式识别鉴别海面溢油的方法。自行配制了 5 6个汽油、柴油、润滑油的模拟海水样品 ,用有机溶剂萃取出海水中的溢油后记录其近红外光谱 ,将原始光谱进行多元散射校正 (MSC)和Norris一阶导数平滑预处理后 ,在主成分分析 (PCA)提取不同种类溢油样品特征的基础上引入马氏距离建立溢油样品的识别模型。研究了光谱预处理对溢油鉴别的影响 ;探讨了马氏距离阈值的确定。结果表明 ,主成分分析可将原始数据压缩而马氏距离判别可给出离群点的阈值 ,本文建立的校正模型能正确判别浓度在 0 4 μL·mL-1以上的溢油类别 ,为近红外光谱结合化学计量学方法建立校正模型进行海面实际溢油样品的分类提供了思路。
In order to quickly understand and grasp the types of oil spills on the sea so as to take emergency measures, a method of near-infrared spectroscopy combined with pattern recognition to identify oil spills in the sea surface is proposed. Fifty-six simulated seawater samples of gasoline, diesel oil and lubricating oil were prepared by themselves. The oil in seawater was extracted with organic solvent and its near-infrared spectra were recorded. The original spectra were smoothed by multivariate scatter correction (MSC) and Norris first derivative After the pretreatment, the identification model of oil spill samples was established by introducing the Mahalanobis distance based on the principal components analysis (PCA) to extract the characteristics of different types of spilled oil samples. The effect of spectral pretreatment on oil spill identification was studied. The determination of Mahalanobis distance threshold was also discussed. The results show that principal component analysis can compress the original data and Mahalanobis distance can give out the threshold of outliers. The calibration model established in this paper can correctly identify the oil spill categories with a concentration above 0 4 μL · mL-1 Infrared spectroscopy combined with chemometric methods to establish a calibration model for the actual classification of oil spill samples provided the idea.