FDA与BP神经网络在缅甸翡翠产地鉴别中的应用  

Application of FDA and BP Neural Network in Identifying the Origin of Burmese Jadeite

作　　者：雷蕾 滕亚君 刘汗青 闵红[3] 祖恩东[1] LEI Lei;TENG Ya-jun;LIU Han-qing;MIN Hong;ZU En-dong(Faculty of Material Science and Engineer,Kunming University of Science and Technology,Kunming 650093,China;Kunming Customs Technology Center,Kunming 650228,China;Industrial Products and Raw Materials Testing Technology Center of Shanghai Entry-exit Inspection and Quarantine Bureau,Shanghai 200135,China)

机构地区：[1]昆明理工大学材料科学与工程学院,云南昆明650093 [2]昆明海关技术中心,云南昆明650228 [3]上海出入境检验检疫局工业品与原材料检测技术中心,上海200135

出　　处：《软件导刊》2021年第3期57-61,共5页Software Guide

基　　金：国家重点研发计划项目(2018YFF0215400)。

摘　　要：传统翡翠鉴别分析方法繁琐,难以实现对大批量样品的快速鉴别,建立一种高效准确的鉴别方法对实现翡翠产地快速鉴别具有重要意义。以缅甸4个主要场口即隆肯、帕敢、达木坎、雷打翡翠为研究对象,将主成分分析结合Fisher判别分析法和BP神经网络分别建立翡翠产地判别分析模型。对鉴别效果进行对比,结果表明:PCA-FAD建立的判别模型综合鉴别准确率为56.75%,达木坎场口翡翠鉴别正确率为71.4%,PCA-BP神经网络模型的综合判别准确率为80.4%。主成分分析结合BP神经网络判别模型对不同场口翡翠的鉴别分类更准确、效果更好,PCA-BP神经网络判别方法具有快速高效可靠实用的鉴别效果。Traditional jadeite identification and analysis methods are cumbersome,and it’s difficult to quickly identify large quantities of samples,therefore,establishing an efficient and accurate identification method has great significance for rapid identification of jade⁃ite origin.Using PCA-FDA and PCA-BP neural network model build for the Discrimination of four main venues of Myanmar like Long⁃ken,Hpakant,Damukan,Leida,and compare the result of identification.The results shows that the correct discrimination rates of PCA-FDA model was 56.75%,Damukan was only 71.4%,the correct discrimination rates of PCA-BP neural network model was 80.4%.Principal component analysis combined with BP neural network discriminant model is more accurate and effective in distin⁃guishing and classifying jade from different fields.The PCA-BP neural network discriminant method has fast,efficient and reliable identification effect.

关 键 词：翡翠 场口 产地鉴别 主成分分析 费希尔判别分析 人工神经网络 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]