摘要
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。
The goal of this study was to analyse natural Jade and fake Jade with the visible hyperspectral imaging, which could provide the basis of Jade identification and obtain the optimal predicted model. There were some problems of hyperspectral imaging because they were nonlinear, in small size and large spatial-spectral dimension. This study applied the principal component analysis （PCA） method to firstly reduce dimensions of the original spectral data. It found that there was a big gap between natural Jade and synthetic lake Jade through comparing the maps of the average spectrum distribution and variance, which proved the feasibility of the hyperspectral imaging technique in the area of Jade identification. Then fisher criterion model, neural network model of back prnpagation, and support vector machine criterion model, were used to correlate the visible reflectance spectra with the identification of Jade. The models performed well for predicting the category of Jade with the accuracy of 96. 37% and 82. 5%, respectively. The results clearly indicated that visible hyperspectral imaging technique in combination with the neural network model of back propagation has the potential as a fast and non-invasive method for identification of Jade.
出处
《分析试验室》
CAS
CSCD
北大核心
2015年第5期562-565,共4页
Chinese Journal of Analysis Laboratory
基金
质检公益性行业科研专项(201210094)资助
