基于BP和Adaboost-BP神经网络的羊肉新鲜度高光谱定性分析
|
摘要
【目的】实现对羊肉新鲜度的快速准确鉴别。【方法】研究通过对460~1 000 nm的羊肉高光谱图像纯肌肉部分提取光谱数据,以挥发性盐基氮(TVB-N)值对新鲜度等级进行划分,对预处理后的光谱数据分别采用连续投影算法(SPA)、主成分分析(PCA)两种压缩降维方法和反向传播(BP)神经网络、自适应提升BP(Adaboost-BP)神经网络两种建模方法开展羊肉新鲜度的分类比较。【结果】其中采用SPA、PCA建立的BP模型校正集与预测集准确率均为100%、83.33%,建立的Adaboost-BP模型校正集与预测集准确率均为100%、94.44%,两种压缩降维方法下Adaboost-BP模型效果均优于BP模型。【结论】利用高光谱图像技术结合Adaboost-BP方法对羊肉新鲜度等级进行分类判别是可行的。
【Objective】To realize the rapid and accurate identification of mutton freshness. 【Method】The spectral data of the pure muscle was extracted from the spectral range of 460-1,000 nm of the mutton hyperspectral image,and the freshness grades were divided according to the volatile base nitrogen( TVB-N)value. After spectral pretreatment,a study on classification comparison of mutton freshness was carried out by using two dimensionality compression method of successive projections algorithm( SPA) and principal component analysis( PCA),two modeling methods of back propagation( BP) neural network and adaptive boosting BP( Adaboost-BP) neural network.【Result】Based on SPA and PCA methods,the accuracy of calibration and prediction sets of the established BP models was 100% and 83. 33%,and that of the Adaboost-BP models was 100% and 94. 44%,respectively. The results of Adaboost-BP model were better than BP model under these two dimensionality compression methods. 【Conclusion 】The study shows that hyperspectral image technology combined with Adaboost-BP method can achieve the accurate qualitative discrimination of freshness of mutton.
【Objective】To realize the rapid and accurate identification of mutton freshness. 【Method】The spectral data of the pure muscle was extracted from the spectral range of 460-1,000 nm of the mutton hyperspectral image,and the freshness grades were divided according to the volatile base nitrogen( TVB-N)value. After spectral pretreatment,a study on classification comparison of mutton freshness was carried out by using two dimensionality compression method of successive projections algorithm( SPA) and principal component analysis( PCA),two modeling methods of back propagation( BP) neural network and adaptive boosting BP( Adaboost-BP) neural network.【Result】Based on SPA and PCA methods,the accuracy of calibration and prediction sets of the established BP models was 100% and 83. 33%,and that of the Adaboost-BP models was 100% and 94. 44%,respectively. The results of Adaboost-BP model were better than BP model under these two dimensionality compression methods. 【Conclusion 】The study shows that hyperspectral image technology combined with Adaboost-BP method can achieve the accurate qualitative discrimination of freshness of mutton.
引文
[1]杨东,陆安祥,王纪华.高光谱成像技术定量可视化检测熟牛肉中挥发性盐基氮含量[J].现代食品科技,2017,33(9):1-7.YANG Dong,LU An-xiang,WANG Ji-hua,et al.(2017).Quantification and Visualization of the Total Volatile Basic Nitrogen of Cooked Beef Based on Hyperspectral Imaging Technique[J].Modern Food Science and Technology,33(9):1-7.(in Chinese)
[2]郭培源,林岩,付妍,等.基于近红外光谱技术的猪肉新鲜度等级研究[J].激光与光电子学进展,2013,50(3):183-189.GUO Pei-yuan,LIN Yan,FU Yan,et al.(2013).Research on Freshness Level of Meat Based on Near-Infrared Spectroscopic Technique[J].Laser and Optoelectronics Progress,50(3):183-189.(in Chinese)
[3]成军虎.基于高光谱成像鱼肉新鲜度无损快速检测方法研究[D].广州:华南理工大学博士学位论文,2016.CHENG Jun-hu.(2016).Non-destructive and Rapid Detection of Fish Freshness using Hyperspectral Imaging Technique[D].Ph D Thesis.South China University of Technology,Guangzhou.(in Chinese)
[4]Crichton,S.O.J.,Kirchner,S.M.,Porley,V.,Retz,S.,Gersdorff,G.V.,&Hensel,O.,et al.(2017).High ph thresholding of beef with vnir hyperspectral imaging.Meat Science,(134):14-17.
[5]朱荣光,姚雪东,段宏伟,等.羊肉挥发性盐基氮的高光谱图像快速检测研究[J].光谱学与光谱分析,2016,36(3):806-810.ZHU Rong-guang,YAO Xue-dong,DUAN Hong-wei,et al.(2016).Study on the Rapid Evaluation of Total Volatile Basic Nitrogen(TVB-N)of Mutton by Hyperspectral Imaging Technique[J].Spectroscopy and Spectral Analysis,36(3):806-810.(in Chinese)
[6]樊阳阳,裘正军,陈俭,等.基于近红外高光谱成像技术的干制红枣品种鉴别[J].光谱学与光谱分析,2017,37(3):836-840.FAN Yang-yang,QIU Zheng-jun,CHEN Jian,et al.(2017).Identification of Varieties of Dried Red Jujubes with Near-Infrared Hyperspectral Imaging[J].Spectroscopy and Spectral Analysis,37(3):836-840.(in Chinese)
[7]Wang,W.,Li,C.,Tollner,E.W.,Gitaitis,R.D.,&Rains,G.C.(2012).Shortwave infrared hyperspectral imaging for detecting sour skin(burkholderia cepacia)-infected onions.Journal of Food Engineering,109(1):38-48.
[8]GB/T 5009.44-2003.肉与肉制品卫生标准的分析方法[S].GB/T 5009.44-2003.Method for Analysis of Hygienic Standard of Meat and Meat Products[S].(in Chinese)
[9]田卫新,何丹丹,杨东,等.一种基于高光谱图像的熟牛肉TVB-N含量预测方法[J].食品与机械,2016,32(12):70-74.TIAN Wei-xin,HE Dan-dan,YANG Dong,et al.(2016).A method for predicting TVB-N content of cooked beef based on hyperspectral image[J].Food and Machinery,32(12):70-74.(in Chinese)
[10]罗微,杜焱喆,章海亮.PCA和SPA的近红外光谱识别白菜种子品种研究[J].光谱学与光谱分析,2016,36(11):3 536-3 541.LUO Wei,DU Yan-zhe,ZHANG Hai-liang.(2016).Discrimination of Varieties of Cabbage with Near Infrared Spectra Based on Principal Component Analysis and Successive Projections Algorithm[J].Spectroscopy and Spectral Analysis,36(11):3,536-3,541.(in Chinese)
[11]刘思伽,田有文,张芳,等.采用二次连续投影法和BP人工神经网络的寒富苹果病害高光谱图像无损检测[J].食品科学,2017,38(8):277-282.LIU Si-jia,TIAN You-wen,ZHANG Fang,et al.(2017).Hyperspectral imaging for nondestructive detection of Hanfu apple diseases using successive projections algorithm and BP neural network[J].Food Science,38(8):277-282.(in Chinese)
[12]梁栋,张凤琴,陈大武,等.一种基于决策树和遗传算法-BP神经网络的组合预测模型[J].中国科技论文,2015,10(2):169-174.LIANG Dong,ZHANG Feng-qin,CHEN Da-wu,et al.(2015).A composite prediction model based on decision tree and GA-BPNN[J].China Sciencepaper,10(2):169-174.(in Chinese)
[13]王军,费凯,程勇.基于改进的Adaboost-BP模型在降水中的预测[J].计算机应用,2017,37(9):2 689-2 693.WANG Jun,FEI Kai,CHENG Yong.(2017).Prediction of rainfall based on improved Adaboost-BP model[J].Journal of Computer Applications,37(9):2,689-2,693.(in Chinese)
[14]段宏伟,朱荣光,王龙,等.感兴趣区域对羊肉p H高光谱检测模型的影响研究[J].光谱学与光谱分析,2016,36(4):1 145-1 149.DUAN Hong-wei,ZHU Rong-guang,WANG Long,et al.(2016).Effects of regions of interest(ROIs)on Detection Models of Mutton p H Based on Hyperspectral Imaing[J].Spectroscopy and Spectral Analysis,36(4):1,145-1,149.(in Chinese)
[2]郭培源,林岩,付妍,等.基于近红外光谱技术的猪肉新鲜度等级研究[J].激光与光电子学进展,2013,50(3):183-189.GUO Pei-yuan,LIN Yan,FU Yan,et al.(2013).Research on Freshness Level of Meat Based on Near-Infrared Spectroscopic Technique[J].Laser and Optoelectronics Progress,50(3):183-189.(in Chinese)
[3]成军虎.基于高光谱成像鱼肉新鲜度无损快速检测方法研究[D].广州:华南理工大学博士学位论文,2016.CHENG Jun-hu.(2016).Non-destructive and Rapid Detection of Fish Freshness using Hyperspectral Imaging Technique[D].Ph D Thesis.South China University of Technology,Guangzhou.(in Chinese)
[4]Crichton,S.O.J.,Kirchner,S.M.,Porley,V.,Retz,S.,Gersdorff,G.V.,&Hensel,O.,et al.(2017).High ph thresholding of beef with vnir hyperspectral imaging.Meat Science,(134):14-17.
[5]朱荣光,姚雪东,段宏伟,等.羊肉挥发性盐基氮的高光谱图像快速检测研究[J].光谱学与光谱分析,2016,36(3):806-810.ZHU Rong-guang,YAO Xue-dong,DUAN Hong-wei,et al.(2016).Study on the Rapid Evaluation of Total Volatile Basic Nitrogen(TVB-N)of Mutton by Hyperspectral Imaging Technique[J].Spectroscopy and Spectral Analysis,36(3):806-810.(in Chinese)
[6]樊阳阳,裘正军,陈俭,等.基于近红外高光谱成像技术的干制红枣品种鉴别[J].光谱学与光谱分析,2017,37(3):836-840.FAN Yang-yang,QIU Zheng-jun,CHEN Jian,et al.(2017).Identification of Varieties of Dried Red Jujubes with Near-Infrared Hyperspectral Imaging[J].Spectroscopy and Spectral Analysis,37(3):836-840.(in Chinese)
[7]Wang,W.,Li,C.,Tollner,E.W.,Gitaitis,R.D.,&Rains,G.C.(2012).Shortwave infrared hyperspectral imaging for detecting sour skin(burkholderia cepacia)-infected onions.Journal of Food Engineering,109(1):38-48.
[8]GB/T 5009.44-2003.肉与肉制品卫生标准的分析方法[S].GB/T 5009.44-2003.Method for Analysis of Hygienic Standard of Meat and Meat Products[S].(in Chinese)
[9]田卫新,何丹丹,杨东,等.一种基于高光谱图像的熟牛肉TVB-N含量预测方法[J].食品与机械,2016,32(12):70-74.TIAN Wei-xin,HE Dan-dan,YANG Dong,et al.(2016).A method for predicting TVB-N content of cooked beef based on hyperspectral image[J].Food and Machinery,32(12):70-74.(in Chinese)
[10]罗微,杜焱喆,章海亮.PCA和SPA的近红外光谱识别白菜种子品种研究[J].光谱学与光谱分析,2016,36(11):3 536-3 541.LUO Wei,DU Yan-zhe,ZHANG Hai-liang.(2016).Discrimination of Varieties of Cabbage with Near Infrared Spectra Based on Principal Component Analysis and Successive Projections Algorithm[J].Spectroscopy and Spectral Analysis,36(11):3,536-3,541.(in Chinese)
[11]刘思伽,田有文,张芳,等.采用二次连续投影法和BP人工神经网络的寒富苹果病害高光谱图像无损检测[J].食品科学,2017,38(8):277-282.LIU Si-jia,TIAN You-wen,ZHANG Fang,et al.(2017).Hyperspectral imaging for nondestructive detection of Hanfu apple diseases using successive projections algorithm and BP neural network[J].Food Science,38(8):277-282.(in Chinese)
[12]梁栋,张凤琴,陈大武,等.一种基于决策树和遗传算法-BP神经网络的组合预测模型[J].中国科技论文,2015,10(2):169-174.LIANG Dong,ZHANG Feng-qin,CHEN Da-wu,et al.(2015).A composite prediction model based on decision tree and GA-BPNN[J].China Sciencepaper,10(2):169-174.(in Chinese)
[13]王军,费凯,程勇.基于改进的Adaboost-BP模型在降水中的预测[J].计算机应用,2017,37(9):2 689-2 693.WANG Jun,FEI Kai,CHENG Yong.(2017).Prediction of rainfall based on improved Adaboost-BP model[J].Journal of Computer Applications,37(9):2,689-2,693.(in Chinese)
[14]段宏伟,朱荣光,王龙,等.感兴趣区域对羊肉p H高光谱检测模型的影响研究[J].光谱学与光谱分析,2016,36(4):1 145-1 149.DUAN Hong-wei,ZHU Rong-guang,WANG Long,et al.(2016).Effects of regions of interest(ROIs)on Detection Models of Mutton p H Based on Hyperspectral Imaing[J].Spectroscopy and Spectral Analysis,36(4):1,145-1,149.(in Chinese)