论文部分内容阅读
为了给玉石鉴定提供依据以及得到优化预测模型,分别对天然玉石和假玉石的可见光高光谱图像进行分析。针对高光谱图像数据的非线性、小样本以及空间光谱维数大等问题,本研究首先对原始光谱数据进行主成分分析(PCA),使高维光谱数据降维,通过对比分析其平均光谱图和方差贡献率图,发现天然玉石与假玉石的谱线之间存在很大的差距,证明了高光谱成像技术在玉石鉴定领域的可行性。然后分别采用费希尔(Fisher)判别法、反向传输(BP)神经网络以及支持向量机(SVM)判别法建立的三种数学模型对玉石进行分类模式判别,结果显示,用Fisher判别法能直接得到预测的类别归属,用BP神经网络以及SVM判别法得到的类别鉴定准确率分别为96.37%,82.5%。研究结果表明,高光谱技术结合BP人工神经网络预测建模方法可以作为快速和非破坏性预测玉石真假的有效手段。
In order to provide the basis for identification of jade and obtain the optimal prediction model, the visible light hyperspectral images of natural jade and fake jade were analyzed respectively. In order to solve the problems of non-linearity, small sample size and large spatial dimension of hyperspectral image data, principal component analysis (PCA) of the original spectral data is carried out in this study. The dimensionality of high-dimension spectral data is reduced. And variance contribution rate map, found that there is a big gap between natural jade and fake jade spectrum line, proved the feasibility of hyperspectral imaging in the field of jade identification. Then, the classification models of jade were identified by using Fisher discriminant, back propagation (BP) neural network and support vector machine (SVM) discriminant respectively. The results show that Fisher discriminant method The classification accuracy of direct predictive categories was 96.37% and 82.5% with BP neural network and SVM discriminant respectively. The results show that hyperspectral technology combined with BP artificial neural network prediction modeling method can be used as an effective and rapid non-destructive prediction of jade artifacts.