本文選題：高光譜技術(shù) + 陳皮��； 參考：《光譜學(xué)與光譜分析》2017年06期

【摘要】：市場上陳皮以次充好現(xiàn)象時有發(fā)生,而年份是衡量陳皮品質(zhì)的重要指標。研究用高光譜技術(shù)結(jié)合化學(xué)計量學(xué)算法,在380~1 023及874~1 734nm兩波段對不同放置方式的陳皮進行年份鑒別。為了尋找更合適的波段和模擬實際生產(chǎn)檢測中陳皮放置的隨機性,采集了四個年份共180個樣本在380~1 023及874~1 734nm的正、反面高光譜圖像(720幅)。用主成分分析法(principal component analysis,PCA)對陳皮光譜信息進行定性分析,發(fā)現(xiàn)不同年份陳皮基于正反面光譜有明顯的聚類;而后以回歸系數(shù)法(regression coefficient,RC)選取陳皮年份相關(guān)的特征波段以減少變量;用偏最小二乘判別分析(partial least squares discriminant analysis,PLS-DA)基于全波段、特征波段對三種放置方式(正、反、正反混合)的樣本建立模型,最后對特征波段建立線性PLS-DA模型和非線性ELM模型并進行比較分析。研究表明:在380~1 023nm的預(yù)測效果大多高于874~1 734nm,基于非線性ELM的判別結(jié)果均高于線性PLS-DA模型,準確率最高可達到建模集100.00%,預(yù)測集98.33%,陳皮正、反、正反混合三種放置方式預(yù)測準確率多數(shù)可高于85%,故采用高光譜技術(shù)可實現(xiàn)對不同放置方式的陳皮年份進行無損鑒別,為進一步開發(fā)便攜儀器或在線生產(chǎn)設(shè)備提供方法和理論依據(jù)。
[Abstract]:In the market, the phenomenon of surrogate skin often occurs, and the year is an important index to measure the quality of the skin. In this paper, hyperspectral technique combined with chemometrics algorithm is used to identify the years of different placement patterns in the two bands of 380 ~ (10 23) and 874 ~ (-1) 734nm. In order to find a more suitable band and to simulate the randomness of skin placement in practical production detection, a total of 180 samples from four years were collected in the positive and negative hyperspectral images of 380 / 1023 and 874 / 1 734nm. The principal component analysis method (principal component analysis) was used to qualitatively analyze the spectral information of pericarp, and it was found that there were obvious clusters based on the positive and negative spectra in different years, and then the characteristic bands were selected by regression coefficient method to reduce the variables. The partial least squares discriminant Analysis (PLS-DA) is used to establish a model based on the samples of three different placement modes (positive, negative, positive and negative), which are based on the full band and the characteristic band. Finally, linear PLS-DA model and nonlinear ELM model are established and compared with each other. The results show that most of the prediction results of 3801 023nm are higher than 874Nm, the discriminant results based on nonlinear ELM are higher than that of linear PLS-DA model, and the highest accuracy can reach 100.0000g of modeling set, 98.33of prediction set, 98.33of Chen Pei Zheng and reverse. The accuracy of prediction of the three placement methods of positive and negative mixing can be higher than that of 85, so using hyperspectral technology can realize the nondestructive identification of the aged skin years with different placement methods. It provides method and theoretical basis for further development of portable instruments or on-line production equipment.
【作者單位】： 浙江大學(xué)生物系統(tǒng)工程與食品科學(xué)學(xué)院;
【基金】：國家自然科學(xué)基金項目(31471417) 國家高技術(shù)研究發(fā)展計劃項目(2013AA10030401)資助
【分類號】：O657.3;R284

【相似文獻】

相關(guān)碩士學(xué)位論文 前1條

1 金靈;基于高光譜技術(shù)的寧夏灘羊肉內(nèi)部品質(zhì)檢測方法研究[D];寧夏大學(xué);2014年

，

本文編號：1882352