近红外光谱技术鉴别海面溢油

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56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。

Oil Spill Identification by Near-Infrared Spectroscopy

Petroleum oil spill happens occasionally at sea. It's important to differentiate the exact products in order to carry out following actions to decrease harmfulness. In the present study, a rapid oil spill identification method by near infrared spectroscopy coupled with pattern recognition techniques is proposed. 56 simulated spilled oils of gasoline, diesel fuel and lubricating oil in marine were chosen to develop the method. Organic reagent of CCl4 was used to extract the oil. Pattern recognition techniques were established by principal component analysis (PCA) coupled with Mahalanobis' distance with the multiplicative signal correction (MSC) and Norris first derivative pretreatment. The study shows that PCA technique is a useful method to extract the main characteristics, and Mahalanobis' distance is an ellipsoidal boundary that circumscribes a data cluster. And oil spill samples with concentration above 0.4 microL x mL(-1) can be successfully identified by the method. The developed technique could be further applied to the identification of spilled oil in marine.