利用油品紫外荧光特性的多光谱成像检测

Multispectral Imaging Detection Using The Ultraviolet Fluorescence Characteristics of Oil

利用石油及其产品具有的紫外荧光特性,搭建了一套紫外诱导多光谱成像系统。该系统主要由3个紫外诱导光源、8个滤波片和1个彩色CCD相机组成。采集了6种油品的多光谱图像,以有效光斑的24个颜色分量均值作为特征,提出了一种联合熵最大化的独立分量分析特征优化方法。K均值聚类和支持向量机识别结果表明,较改进前的ICA方法,该方法的特征优化性能得到了有效提高,油种识别率达到了92.3%。

Based on the UV fluorescence phenomena of oil and its products, a multispectral imaging system was constructed. This system was composed of 3 UV excitation light sources, 8 optics filters and a CCD camera. Using this system, multi-spectral images of 6 kinds of oil were collected. The mean of 24 color features of effective light spots was used as the feature set. Then, a novel method called maximize the joint entropy of independent component analysis ( ICA ) was proposed for K-mean cluster and SVM recognition. It is proved that this method is better than traditional ICA for feature optimized, and the identification rate is 92. 3%. This result has positive significance for oil detection.