近红外光谱对贮藏期猕猴桃不同深度果肉色泽的变化研究

引用本文

余克强, 孟浩, 曹晓峰, 赵艳茹. 近红外光谱对贮藏期猕猴桃不同深度果肉色泽的变化研究[J]. 光谱学与光谱分析, 2020,40(7): 2240-2245.
YU Ke-qiang, MENG Hao, CAO Xiao-feng, ZHAO Yan-ru. Near-Infrared Spectroscopy for Analyzing Changes of Pulp Color of Kiwifruit in Different Depths[J]. Spectroscopy and Spectral Analysis, 2020,40(7): 2240-2245.
Doi:10.3964/j.issn.1000-0593(2020)07-2240-06 复制到剪切板

Permissions

《光谱学与光谱分析》期刊社所有

近红外光谱对贮藏期猕猴桃不同深度果肉色泽的变化研究

余克强^1,^2,³, 孟浩¹, 曹晓峰¹, 赵艳茹^1,^2,³

1. 西北农林科技大学机械与电子工程学院, 陕西杨凌 712100

2. 农业农村部农业物联网重点实验室, 陕西杨凌 712100

3. 陕西省农业信息感知与智能服务重点实验室, 陕西杨凌 712100

作者简介: 余克强, 1986年生, 西北农林科技大学机械与电子工程学院讲师 e-mail: yuke406336022@163.com

收稿日期: 2019-06-26

基金: 国家自然科学基金项目(61705188), 中国博士后科学基金项目(2017M613218), 陕西省博士后科学基金项目(2017BSHYDZZ61)和中央高校基本科研业务费专项资金项目(2452017125)资助

摘要

猕猴桃是我国发展势头和经济效益比较突出的水果之一, 其果肉色泽是评价猕猴桃果实品质的重要指标。利用近红外光谱技术对贮藏期猕猴桃不同深度果肉色泽的变化进行研究。以贮藏期“哑特”猕猴桃果皮下0, 5和10 mm处果肉色泽(L^*, a^*和b^*)为研究对象, 用近红外光谱(830~2 500 nm)结合化学计量学方法对猕猴桃果肉色泽特征进行预测分析。通过建立基于全波段的偏最小二乘回归(PLSR)模型, 发现猕猴挑果皮下5 mm处色泽特征($L^{*}_{5}$,$a^{*}_{5}$,$b^{*}_{5}$)所建立的校正预测模型效果好, 说明该处的色泽数据和近红外光谱信息的相关度较高。运用竞争性自适应重加权采样法(CARS)和无信息变量消除法(UVE)两种算法从高维近红外光谱全波段信息中选取与颜色特征相关的特征波长信息, 并与猕猴桃果皮下5 mm处的色泽($L^{*}_{5}$,$a^{*}_{5}$,$b^{*}_{5}$)分别建立PLSR和多元线性回归(MLR)预测模型。其中对果肉色泽$L^{*}_{5}$所建立的模型中, CARS-PLSR模型的校正和预测效果均为最好, R_C达到0.942 7, RMSEC为1.699 7, R_P达到0.885 0, RMSEP为1.642 4; 对猕猴桃果肉色泽$a^{*}_{5}$所建立的模型中, UVE-MLR模型的校正和预测效果最好, R_C达到0.946 3, RMSEC为0.342 4, R_P达到0.854 9, RMSEP为0.629 6; 对猕猴桃果肉色泽$b^{*}_{5}$所建立的模型中, CARS-MLR模型的效果最好, R_C达到0.944 3, RMSEC为1.010 1, R_P达到0.839 8, RMSEP为1.354 3。研究表明近红外光谱分析技术检测猕猴桃果皮下5 mm处色泽($L^{*}_{5}$,$a^{*}_{5}$和$b^{*}_{5}$)具有良好的准确度, 为猕猴桃品质评价提供技术支撑。

关键词: 猕猴桃; 近红外光谱; 贮藏期; 不同深度; 果肉色泽

中图分类号:O657.3 文献标志码:A

Near-Infrared Spectroscopy for Analyzing Changes of Pulp Color of Kiwifruit in Different Depths

YU Ke-qiang^1,^2,³, MENG Hao¹, CAO Xiao-feng¹, ZHAO Yan-ru^1,^2,³

1. College of Mechanical and Electronic Engineering, Northwest A&F University, Yangling 712100, China

2. Key Laboratory of Agricultural Internet of Things, Ministry of Agriculture and Rural Affairs, Yangling 712100, China

3. Shaanxi Key Laboratory of Agricultural Information Perception and Intelligent Service, Yangling 712100, China

Abstract

Kiwifruit is one of the fruits with strong development momentum and economic benefits in China, its pulp color has become an important indicator for evaluating the quality of kiwifruit. Here, near-infrared spectroscopy was employed to study the changes in pulp color in different depths of kiwifruit during different storage periods. In this study, the “Mute” kiwifruit’s pulp color features ( L^*, a^*, b^*) in depths of 0, 5, and 10 mm under the skin wereviewed as the research object, the near-infrared spectroscopy (830~2 500 nm) was used as a technical tool, and chemometric methods were combined to analyze the pulp color features of kiwifruit. By establishing a partial least-square regression (PLSR) model based on the full-wavelengths, it found that the established model offered good results by using color features ($L^{*}_{5}$,$a^{*}_{5}$,$b^{*}_{5}$) at a depth of 5 mm, which indicated that the pulp color features and the spectrum data had a relatively high correlation. Then, the competitive adaptive reweighted sampling (CARS) and uninformative variable elimination (UVE) algorithms were used to select the characteristic wavelengths related to color features from the high-dimensional full-wavelengths. And the PLSR and multiple linear regression (MLR) prediction models were respectively established based on the color features ($L^{*}_{5}$,$a^{*}_{5}$,$b^{*}_{5}$) and spectra at characteristic wavelengths. Results revealed that the CARS-PLSR model with the R_C=0.942 7, RMSEC=1.699 7, R_P=0.885 0, and RMSEP=1.642 4 has the best predictive effect for the pulp color feature $L^{*}_{5}$; the UVE-MLR model with the R_C=0.946 3, RMSEC=0.342 4, R_P=0.854 9, and RMSEP of 1.354 3 exhibited the best predictive results for pulp color feature $a^{*}_{5}$, the CARS-MLR model with the R_C=0.944 3, RMSEC=1.010 1, R_P=0.839 8, and RMSEP=1.354 3 performed best predictive results for pulp color feature $b^{*}_{5}$ The results demonstrated that the near-infrared spectroscopy technique would be employed to detect the color features at different depths of kiwifruit, which provided technical support for the quality evaluation of kiwifruit.

Keyword: Kiwifruit; Near-infrared spectroscopy; Storage periods; Different depths; Pulp color features

文章图片

引言

猕猴桃, 又名奇异果, 是我国主要水果产物之一。因猕猴桃果肉富含维生素等营养物质, 拥有“ 水果之王” 、 “ VC之王” 的美称, 受到很多消费者的青睐。然而, 猕猴桃的外部特征(颜色、大小、形状、表面缺陷等)和内部品质(糖度、酸度、果肉色泽、 VC等)是新品种培育和果实分级销售的重要参考指标。其中, 果肉色泽在猕猴桃不同品种、成熟和贮藏等过程中都存在很大差异, 果肉色泽已成为评价猕猴桃品质的重要指标^{[1, 2]}。

通常猕猴桃果肉色泽测量的方法是有损检测方法, 在实际生产应用中无法实现全样本测量, 且费时费力, 无法满足新品种培育和消费市场等对色泽高通量检测需求。因此寻求一种准确、快速检测猕猴桃果肉色泽方法对猕猴桃育种、栽培和采后贮藏销售等具有重要意义、近红外光谱技术是一种通过获取待测物漫反射、透射和透漫反射信息反应样本组成成分和分子结构等的无损检测技术, 凭借其高效、成本低以及多组分检测等的特点, 广泛应用于食品、制药、化工、农业等领域。在水果品质检测方面, Kusumiyati等^[3]运用手持式的近红外光谱仪结合偏最小二乘回归(partial least square regression, PLSR)对植株上生长和采后的番茄硬度、颜色参数(L^*, a^*, b^*, 色彩角h° 和色度C^*)和番茄红素进行检测, 结果较好。 Sun等^[4]开发手持式近红外光谱测试仪来评估南丰蜜桔品质, 运用支持向量机(support vector machine, SVM)、向后传播神经网络(back propagation neural network, BPNN)和PLSR预测果皮颜色、糖分、滴定酸和VC含量。 Kamal-Eldin等^[5]借助多元化学计量学方法对不同品种枣的颜色(L^*, a^*, b^*)、外形(重量, 长度, 体积, 直径和密度)、化学(糖和有机酸)、质构特征(硬度, 附着力, 回弹性, 脆性和弹性)进行分析, 研究结果对水果品质评价和分级有重要意义。

以不同贮藏期猕猴桃为研究对象, 分析不同深度(0, 5和10 mm)果肉色泽与近红外光谱反射率之间的变化关系, 并结合化学计量学方法对猕猴桃果肉色泽特征进行预测分析, 研究结果为猕猴桃品质评价提供技术支撑。

1 实验部分

1.1 猕猴桃样本

猕猴桃品种为“ 哑特” , 于2018年10月18日采摘于陕西武功县某猕猴桃果园, 选择大小近似的猕猴桃样本共300个, 采摘后置于4 ℃冰箱中贮藏。后期实验中, 每次取出60个猕猴桃进行近红外光谱数据和色泽参数的测量, 5次实验共得到300个猕猴桃样本的光谱和色泽数据。

1.2 仪器

采用傅里叶近红外光谱仪(MPA, RUKER, Germany)对猕猴桃样本光谱数据采集, 其光谱范围在833~2 500 nm, 分辨率为0.3~0.5 nm, 一次测量扫描64次(约30 s)。采用内置反射背景作为参考, 固体光纤探头进行测量。

猕猴桃颜色参数测试采用色差计(CR-10 plus, Konica Minolta, Japan)来完成。其测量范围: L^*为1~100, a^*为-60~+60, b^*为-60~+60。光源为脉冲氙灯, 测量时间约1 s, 使用端口为USB2.0, 10° 观察角。温湿度范围0~40 ℃; 相对湿度: 85%以下。采用的标准白板L^*为95, a^*为-3, b^*为-4。

1.3 猕猴桃光谱数据获取与色泽特征测量

利用MPA近红外光谱仪固体光纤探头测量猕猴桃样本赤道位置的反射光谱, 间隔90° 均匀采集每个样本的四个位置的光谱数据, 并取四个位置反射光谱平均值作为该样本的光谱数据。光谱数据采集后, 用管壁厚度为0.15 mm、直径为15 mm的不锈钢管从近红外光谱测量点[图1(a)]的位置刺入, 取出高度约15~25 mm的猕猴果肉圆柱体[图1(b)], 用薄刀片去除果肉圆柱体的果皮得到果皮下0 mm果肉的横切面, 并依次切出果肉圆柱体果皮下5和10 mm的果肉横切面, 用色差计依次测量果皮下0, 5和10 mm处的果肉色差数据(L^*, a^*, b^*), 分别如图1(c), (d)和(e)所示, 并对四个位置的色差数据求取平均值作为该样本的色差数据^[6]。

	Figure Option View Download New Window
	图1 猕猴桃色泽数据测量示意图Fig.1 Schematic diagram of color data measurement for kiwifruit samples

1.4 光谱数据处理方法介绍

由于获取的猕猴桃光谱数据中存在噪声信息, 采用S-G平滑法^[7]对猕猴桃原始光谱数据进行预处理。后期采用SPXY算法^[8]划分为校正集和预测集。光谱数据的压缩和特征波段的选择, 选用竞争性自适应重加权采样法(competitive adaptive reweighted sampling, CARS)^[8]和无信息变量消除法(uninformative variable elimination, UVE)^[9]来实现。光谱数据建模主要采用偏最小二乘回归(partial least-squares regression, PLSR)^{[8, 10]}和多元线性回归(multiple linear regression, MLR)^[11]来完成。模型评价主要是利用相关系数(correlation coefficient, R)和均方根误差(root mean square error, RMSE)^{[8, 10]}。光谱数据处理软件主要有The Unscrambler X 10.1(CAMO, Norway)和MATLAB 2016a(The MathWorks, USA)。

2 结果与讨论

2.1 贮藏期猕猴桃样本近红外反射光谱特征分析

图2是采集的猕猴桃样本在830~2 500 nm的光谱反射曲线。图2(a)是利用S-G平滑算法对300个猕猴桃样本原始光谱进行预处理后的曲线, 图2(b)是5个不同贮藏时间(1, 5, 8, 11和14 d)下猕猴桃样本的平均光谱曲线。从图2可得: 不同贮藏期猕猴桃样本反射光谱曲线趋势是大致相同, 只是在反射率方面有所差异; 而不同贮藏期猕猴桃平均光谱曲线存在交叉, 没有一定的规律, 可能是因为猕猴桃在贮藏过程中淀粉分解成葡萄糖、果糖(与C— H, C— O键的倍频吸收有关)等物质含量变化造成的^[12]。从图中可得, 光谱曲线的吸收峰位置主要在960, 1 180, 1 450, 1 850和2 250 nm等处, 其中980, 1 180和1 450 nm处的吸收峰与O— H键的倍频有关, 反应猕猴桃水分信息, 1 650和2 250 nm处的吸收峰则与C— H和C— O键有关^{[13, 14, 15]}。

	Figure Option View Download New Window
	图2 猕猴桃样本的近红外反射光谱图 (a): S-G平滑处理后; (b): 不同贮藏时期的平均光谱Fig.2 Near-infrared reflectance spectra of kiwifruit samples (a): S-G smoothing preprocessing; (b): Mean spectral curves during different storage periods

2.2 贮藏期猕猴桃不同深度果肉色泽分析

猕猴桃5个不同贮藏期(1, 5, 8, 11和14 d)猕猴桃样本剖面图, 如表1所示。

表1 不同贮藏时期猕猴桃样本剖面图 Table 1 Sectional views of kiwifruit samples during different storage periods

表1中可得, 随着贮藏时间增长, 猕猴桃果肉颜色逐渐加深, 主要是因为猕猴桃果肉中葡萄糖、果糖等物质含量变化造成。贮藏期猕猴桃不同深度下果肉色泽(L^*, a^*, b^*)最小值、最大值、均值以及标准差统计结果, 如表2所示。

表2 贮藏期猕猴桃不同深度果肉色泽测量结果统计 Table 2 Color features measurement results of kiwifruit at different depths during storage

从表2中得出, 随着猕猴桃果肉深度(0, 5和10 mm)变大, 色泽参数L^*均值绝对值变小, 果肉色泽越亮; 色泽参数a^*均值的绝对值变小, 果肉色泽由深绿色转变为浅绿色; 而色泽参数b^*的均值变小, 颜色由高亮度黄色转变为低亮度黄色。从色泽特征的标准差来看, 猕猴桃5 mm深度的果肉色泽变化比其他0和10 mm两个位置对应色泽特征的标准差相比要小, 表明猕猴桃5 mm深度的果肉色泽比较稳定。

2.3 猕猴桃样本集数据划分结果分析

将300个贮藏期猕猴桃样本光谱信息以及对应的果肉色泽(L^*, a^*, b^*), 采用SPXY算法按照3: 1划分为校正集(225个样本)与预测集(75个样本)。猕猴桃样本的果皮下0, 5和10 mm处果肉色泽最小值、最大值、平均值以及标准差的统计结果, 分别见表3所示。

表3 猕猴桃0, 5和10 mm处果肉色泽在建模集和预测集的统计结果 Table 3 Statistical results of pulp color at 0, 5 and 10 mm depths in calibration and prediction sets

从表3中可得样本集中不同深度色泽数据变化相对稳定, 样本不同部位预测集的色泽数据在对应校正集色泽数据变化范围之内, 说明本研究SPXY方法划分样本是合理的。

2.4 基于全波段光谱的猕猴桃色泽PLSR预测模型

以猕猴桃样本全波段光谱数据为自变量, 以对应的猕猴桃样本果皮下0, 5和10 mm处的果肉色泽指标L^*, a^*, b^*为因变量, 分别建立近红外光谱与色泽的PLSR模型, 并且以因变量将9个模型分别命名为 $L_{0}^{*}$ , $a_{0}^{*}$ , $b_{0}^{*}$ ; $L_{5}^{*}$ , $a_{5}^{*}$ , $b_{5}^{*}$ ; $L_{10}^{*}$ , $a_{10}^{*}$ , $b_{10}^{*}$ 。以此建立的PLSR模型的结果如表4所示。

表4 基于PLSR的猕猴桃不同果肉色泽预测模型的结果 Table 4 Results of different PLSR predictive models for kiwifruit color with different depths

Depths /mm	Models	Calibration		Cross-validation		Prediction
Depths /mm	Models	R_C	RMSEC	R_CV	RMSECV	R_P	RMSEP
	$L_{0}^{*}$	0.755 1	3.617 6	0.740 7	3.707 1	0.732 7	3.528 9
0	$a_{0}^{*}$	0.765 0	0.886 3	0.706 0	0.974 8	0.709 7	0.619 9
	$b_{0}^{*}$	0.821 2	2.373 8	0.842 9	2.580 5	0.767 0	2.448 6
	$L_{5}^{*}$	0.903 6	2.153 2	0.910 1	2.110 1	0.844 2	1.828 7
5	$a_{5}^{*}$	0.950 5	0.328 9	0.803 9	0.630 4	0.702 5	0.524 8
	$b_{5}^{*}$	0.864 7	1.542 3	0.849 8	1.618 8	0.736 0	1.623 5
	$L_{10}^{*}$	0.899 6	3.233 5	0.892 8	3.335 7	0.871 2	2.927 5
10	$a_{10}^{*}$	0.737 9	0.686 6	0.734 3	0.690 7	0.753 1	0.613 6
	$b_{10}^{*}$	0.787 1	2.511 0	0.752 7	2.680 2	0.719 7	2.303 9

表4 基于PLSR的猕猴桃不同果肉色泽预测模型的结果 Table 4 Results of different PLSR predictive models for kiwifruit color with different depths

从表4可得: 9个校正模型的预测精度较好, 其中猕猴桃样本果皮下5 mm处 $L_{5}^{*}$ , $a_{5}^{*}$ , $b_{5}^{*}$ 模型的R_C值比样本果皮下0和10 mm处对应的L^*, a^*, b^*模型的R_C值大, R_P的值差距较小, RMSEC比其他深度对应的RMSEC值小。说明在果皮下5 mm选取的色泽参数与近红外光谱信息所建立的模型效果比在样本果皮下0和10 mm处色差指标所建立的模型效果好, 即猕猴桃样本近红外光谱与果肉5 mm处色泽相关度较高, 所以对猕猴桃样本果皮下5 mm处色差 $L_{5}^{*}$ , $a_{5}^{*}$ , $b_{5}^{*}$ 模型进一步研究。

2.5 猕猴桃果肉色泽特征波长选择

为了加快运算速率, 减少冗余光谱信息对模型的影响, 采用CARS算法和UVE算法从近红外全波长光谱提取特征波长, 用来优化猕猴桃近红外光谱与果肉色泽定量模型。根据CARS算法, 在猕猴桃样本色泽 $L_{5}^{*}$ 特征波长变量筛选过程中, 当RMSECV最低为2.01时, 筛选了猕猴桃色泽 $L_{5}^{*}$ 特征波长有848, 871, 883, 1 184, 1 327, 1 819和2 134 nm等共计33个。根据CARS提取猕猴桃色泽 $L_{5}^{*}$ 特征波长的情况, 猕猴桃果肉色泽 $L_{5}^{*}$ 的特征波长主要分布在850~980, 1 150~1 280, 1 350~1 480和1 980~2 160 nm四个范围内。利用同样的过程, 对猕猴桃样本色泽特征 $a_{5}^{*}$ 和 $b_{5}^{*}$ 特征波长也进行筛选, 分别得到42和35个。

然后运用UVE算法, 选取猕猴桃果肉色泽 $L_{5}^{*}$ 特征波长, 筛选的波长有842, 943, 1 027, 1 378, 1 695, 2 027和2 225 nm等共计25个, 特征波长主要分布在850~1 080, 1 180~1 340, 1 220~1 440和1 890~2 140 nm这四个范围内。同样利用UVE算法, 对猕猴桃样本色泽特征 $a_{5}^{*}$ 和 $b_{5}^{*}$ 特征波长也进行筛选, 分别得到30和34个。

2.6 猕猴桃果肉色泽预测模型的建立和分析

猕猴桃果皮下5 mm果肉色泽特征( $L_{5}^{*}$ , $a_{5}^{*}$ , $b_{5}^{*}$ )结合CARS和UVE两种算法选取的特征波长分别建立的PLSR与MLR预测模型结果, 如表5所示。

表5 猕猴桃果肉5 mm深度处色泽不同预测模型的结果统计 Table 5 The results of different models for predicting kiwifruit pulp color features at the depth of 5 mm

Color features	Models	Calibration		Crossvalidation		Prediction
Color features	Models	R_C	RMSEC	R_CV	RMSECV	R_P	RMSEP
$L_{5}^{*}$	CARS-PLSR	0.942 7	1.699 7	0.927 5	1.792 0	0.885 0	1.642 4
	CARS-MLR	0.948 2	1.618 1	-	-	0.875 6	1.692 0
	UVE-PLSR	0.929 9	1.872 6	0.919 3	1.887 8	0.840 0	1.957 2
	UVE-MLR	0.956 5	1.485 4	-	-	0.814 4	2.189 4
$a_{5}^{*}$	CARS-PLSR	0.914 7	0.428 7	0.882 8	0.469 7	0.819 9	0.465 8
	CARS-MLR	0.926 9	0.397 6	-	-	0.806 1	0.478 6
	UVE-PLSR	0.922 3	0.409 3	0.840 9	0.541 2	0.759 2	0.554 6
	UVE-MLR	0.946 3	0.342 4	-	-	0.854 9	0.629 6
$b_{5}^{*}$	CARS-PLSR	0.923 0	1.181 3	0.897 9	1.284 2	0.814 6	1.436 6
	CARS-MLR	0.944 3	1.010 1	-	-	0.839 8	1.354 3
	UVE-PLSR	0.901 2	1.331 2	0.863 7	1.470 4	0.771 4	1.571 3
	UVE-MLR	0.909 4	1.277 3	-	-	0.782 1	1.570 8

表5 猕猴桃果肉5 mm深度处色泽不同预测模型的结果统计 Table 5 The results of different models for predicting kiwifruit pulp color features at the depth of 5 mm

对比表5和表4发现: 经过提取特征波长后建立的猕猴桃果肉色泽(L^*, a^*, b^*)模型的精确度有了较大的提高, 基于特征波长所建模型的R_C均达到0.9以上。对果肉色泽参数 $L_{5}^{*}$ 所建立的模型中, CARS-PLSR模型的校正效果和预测效果均为最好[图3(a)]; 果肉色泽 $a_{5}^{*}$ 所建立的模型中, UVE-MLR模型的校正和预测效果最好[图3(b)]; 而果肉色泽 $b_{5}^{*}$ 所建立的模型中, CARS-MLR模型的效果最好(图3(c))。

	Figure Option View Download New Window
	图3 不同模型下参考和预测的色泽参数L^(a), a^(b), b^(c)在校正集和预测集的回归线Fig.3* Regression plots of reference vs. predicted color features L^* (a), a^* (b), b^* (c) in calibration and prediction sets by different models

3 结论

利用近红外光谱, 探讨了不同贮藏时期猕猴桃不同深度果肉色泽的变化情况。探讨了猕猴桃样本果皮下0, 5和10 mm处果肉色泽特征(L^*, a^*, b^*)的变化情况。基于PLSR建立猕猴桃果皮下0, 5和10 mm处色泽(L^*, a^*, b^*)和830~2 500 nm范围内的近红外光谱信息的校正预测模型, 得出深度5 mm处色泽( $L_{5}^{*}$ , $a_{5}^{*}$ , $b_{5}^{*}$ )的预测效果好, 并基于5 mm处色泽数据和光谱反射率, 利用CARS和UVE两种算法分别选择相应的特征波长, 并建立基于PLSR与MLR的校正预测模型, 模型的R_C均达到0.9以上, 对果肉色泽 $L_{5}^{*}$ 所建立的模型中, CARS-PLSR模型的效果最好; 对色泽 $a_{5}^{*}$ 所建立的模型中, UVE-MLR模型的效果最好; 对色泽 $b_{5}^{*}$ 所建立的模型中, CARS-MLR模型的效果最好。今后准备选择其他颜色果肉的猕猴桃样本, 以及不同品种和不同生长环境的猕猴桃作为研究对象, 对模型适用性进一步研究。

参考文献

文献列表

[1]	Li H, Pidakala P, Billing D, et al. Postharvest Biology and Technology, 2016, 120: 127. [本文引用:1]
[2]	Li B, Cobo-Medina M, Lecourt J, et al. Postharvest Biology and Technology, 2018, 141: 8. [本文引用:1]
[3]	Kusumiyati, Akinaga T, Tanaka M, et al. Journal of Food, Agriculture & Environment, 2008, 6(2): 327. [本文引用:1]
[4]	Sun X D, Zhang H L, Liu Y D. International Journal of Agricultural & Biological Engineering, 2009, 2(1): 65. [本文引用:1]
[5]	Kamal-Eldin A, Ghnimi S. Journal of Food Measurement & Characterization, 2018, 12: 1020. [本文引用:1]
[6]	Nicola B M, Defraeye T, De Ketelaere B, et al. Annual Review of Food Science & Technology, 2014, 5(1): 285. [本文引用:1]
[7]	ZHANG Chu, LIU Fei, KONG Wen-wen, et al(张初, 刘飞, 孔汶汶, 等). Transactions of the Chinese Society of Agricultural Engineering(农业工程学报), 2013, 29(20): 270. [本文引用:1]
[8]	Yu K Q, Zhao Y R, Li X L, et al. Plos One, 2014, 9(12): e116205. [本文引用:4]
[9]	Tan C, Wang J, Wu T, et al. Spectrochimica Acta Part A, 2010, 77(5): 960. [本文引用:1]
[10]	YU Ke-qiang, ZHAO Yan-ru, LIU Fei, et al(余克强, 赵艳茹, 刘飞, 等). Transactions of the Chinese Society of Agricultural Engineering(农业工程学报), 2016, 32(15): 197. [本文引用:2]
[11]	Antonio M, Maggio R M. Journal of Pharmaceutical Biomedical Analysis, 2018, 149(2): 603. [本文引用:1]
[12]	Guo W, Zhao F. Food Analytical Methods, 2016, 9(1): 38. [本文引用:1]
[13]	Li M, Pullanagari R, Pranamornkith T, et al. Journal of Food Engineering, 2017, 202: 46. [本文引用:1]
[14]	Yu K, Zhao Y, Li X, et al. Computers & Electronics in Agriculture, 2014, 103(103): 1. [本文引用:1]
[15]	Yu K Q, Zhao Y R, Liu Z Y, et al. Food & Bioprocess Technology, 2014, 7(11): 3077. [本文引用:1]
[16]	Alguero-Chaparro L, Baeten V, Juan A, et al. Food Chemistry, 2013, 139(1-4): 1121. [本文引用:1]

2016

0.0

... 其中, 果肉色泽在猕猴桃不同品种、成熟和贮藏等过程中都存在很大差异, 果肉色泽已成为评价猕猴桃品质的重要指标^[1,2] ...

2018

0.0

... 其中, 果肉色泽在猕猴桃不同品种、成熟和贮藏等过程中都存在很大差异, 果肉色泽已成为评价猕猴桃品质的重要指标^[1,2] ...

2008

0.0

... 在水果品质检测方面, Kusumiyati等^[3]运用手持式的近红外光谱仪结合偏最小二乘回归(partial least square regression, PLSR)对植株上生长和采后的番茄硬度、颜色参数(L^{#cod#x0002A ...}

2009

0.0

... Sun等^[4]开发手持式近红外光谱测试仪来评估南丰蜜桔品质, 运用支持向量机(support vector machine, SVM)、向后传播神经网络(back propagation neural network, BPNN)和PLSR预测果皮颜色、糖分、滴定酸和VC含量 ...

2018

0.0

... Kamal-Eldin等^[5]借助多元化学计量学方法对不同品种枣的颜色(L^{#cod#x0002A ...}

2014

0.0

... ), 分别如图1(c), (d)和(e)所示, 并对四个位置的色差数据求取平均值作为该样本的色差数据^[6] ...

2013

0.0

ZHANG

Chu

, LIU

Fei

, KONG

Wen-wen

, et al(张初, 刘飞, 孔汶汶, 等). Transactions of the Chinese Society of Agricultural Engineering(农业工程学报), 2013, 29(20): 270.

Abstract: Watermelon seed variety selection plays a vital role in watermelon planting, and the variety of watermelon seeds directly affect the yield and quality of watermelons. In this study, we aimed to identify the cultivars of watermelon seeds by using a novel, rapid, non-invasive, and low cost technique named hyperspectral imaging. 121 samples of four different cultivars of watermelon seeds were investigated, and a near-infrared hyperspectral imaging system (874-1734 nm with 256 bands) was established to acquire the hyperspectral images of the samples. A region of interest (ROI) with 15×15 pixels of the hyperspectral image of each sample was defined, and the average reflectance spectrum of the ROI were extracted. To remove the absolute noises of the spectra, only the spectral range 1 042-1 646 nm was used for analysis, and to reduce the noises existed in spectral range 1 042-1 646 nm, the extracted 121 reflectance spectra were preprocessed by Savitzky-Golay smoothing (SG), Empirical Mode Decomposition (EMD), and Wavelet Transform (WT) methods. The preprocessed spectra were then used to select sensitive wavelengths by Successive Projections Algorithm (SPA) and Genetic Algorithm-partial least squares (GA-PLS) methods. Different numbers of sensitive wavelengths were selected by different variable selection methods with different preprocessing methods. 24, 16, and 15 sensitive wavelengths were selected by SPA with spectra preprocessed by SG, EMD, and WT, respectively. Moreover, 38, 33. and 32 sensitive wavelengths were selected by GA-PLS with spectra preprocessed by SG, EMD. and WT, respectively. Partial least squares - discriminant analysis (PLS-DA) was used to build discriminant models with the full spectra, and back-propagation neural network (BPNN) and extreme learning machine (ELM) were applied to build discriminant models with the selected wavelength variables. A PLS-DA model with spectra preprocessed by EMD obtained the best identification rate among all PLS-DA models, with an identification rate of 91.57% in the calibration set and 78.95% in the prediction set. SPA-BPNN models showed relatively worse results than GA-PLS-BPNN models with the same spectral preprocessing methods. The SG-GA-PLS-BPNN model obtained the best performance among all BPNN models, with an identification rate of 92.77% in the calibration set and 86.84% in the prediction set. Compared with the PLS-DA models and the BPNN models, ELM models obtained the best results. All ELM models obtained an identification rate over 90% in the calibration set and the prediction set, and the SG-SPA-ELM model, SG-GA-PLS-ELM model, and WT-SPA-ELM model obtained the identification rate of 100% of calibration and prediction. The overall results showed that BPNN and ELM models performed better than PLS-DA models, and the ELM models with the selected wavelengths based on SG preprocessed spectra obtained the best results, with 100% classification accuracy for both the calibration set and the prediction set. The SG preprocessing method showed the best performance in all PLS-DA, BPNN, and ELM models. The results indicated that it was feasible to use near-infrared hyperspectral imaging to identify the watermelon seed varieties, and near-infrared hyperspectral imaging provided an alternate way of rapid identification of watermelon seed variety. ELM, as a single hidden layer feed-forward network, was an effective classification method in watermelon seed cultivar identification. Moreover, the results in this paper showed the great potential of hyperspectral imaging in the seed industry for on-line identification of seed cultivars and detection of the seed quality parameters.

为了研究采用近红外高光谱图像技术对西瓜种子品种快速无损鉴别的可行性，该文采用近红外高光谱图像技术，通过提取西瓜种子的光谱反射率，结合Savitzky-Golay (SG)平滑算法，经验模态分解算法（empirical mode decomposition，EMD）和小波分析（wavelet transform，WT）对提取出的光谱数据进行去除噪声处理，采用连续投影算法（successive projections algorithm，SPA）和遗传-偏最小二乘法（genetic algorithm-partial least squares，GA-PLS）进行特征波长选择。基于全波段光谱建立了偏最小二乘判别分析（partial least squares-discriminant analysis，PLS-DA），基于特征波长建立了反向传播神经网络（back-propagation neural network，BP NN) 判别模型和极限学习机（extreme learning machine，ELM）判别模型。试验结果表明，基于特征波长的BPNN模型和ELM模型的结果优于基于全部波长的PLS-DA模型，基于SG预处理光谱提取的特征波长建立的ELM模型取得最优的判别效果，建模集和预测集的判别正确率均为100%。结果表明应用近红外高光谱成像技术对西瓜种子品种鉴别是可行的，为西瓜种子的品种快速鉴别提供了一种新方法。

... 4 光谱数据处理方法介绍由于获取的猕猴桃光谱数据中存在噪声信息, 采用S-G平滑法^[7]对猕猴桃原始光谱数据进行预处理 ...

2014

0.0

... 后期采用SPXY算法^[8]划分为校正集和预测集 ...

... 光谱数据的压缩和特征波段的选择, 选用竞争性自适应重加权采样法(competitive adaptive reweighted sampling, CARS)^[8]和无信息变量消除法(uninformative variable elimination, UVE)^[9]来实现 ...

... 光谱数据建模主要采用偏最小二乘回归(partial least-squares regression, PLSR)^{[8, 10]}和多元线性回归(multiple linear regression, MLR)^[11]来完成 ...

... 模型评价主要是利用相关系数(correlation coefficient, R)和均方根误差(root mean square error, RMSE)^{[8, 10]} ...

2010

0.0

2016

0.0

Ke-qiang

, ZHAO

Yan-ru

, LIU

Fei

, et al(余克强, 赵艳茹, 刘飞, 等). Transactions of the Chinese Society of Agricultural Engineering(农业工程学报), 2016, 32(15): 197.

为了满足土壤重金属快速准确检测的需求,同时为土壤的重金属污染防治和农业的可持续发展提供理论指导,该研究利用激光诱导击穿光谱(laser-induced breakdown spectroscopy,LIBS)技术结合定标曲线法和化学计量学方法对土壤中重金属铅(Pb)和镉(Cd)元素进行定量分析。在获取LIBS数据之后,结合土壤LIBS发射谱线中Pb和Cd谱峰信息以及美国国家标准与技术研究院(national institute of standards and technology,NIST)的标准原子光谱数据库,选取了Pb和Cd的特征谱线分别为Pb I 405.78和Cd I 361.05 nm。首先对谱线信息进行预处理后,根据谱峰信息和元素的含量,分别建立了基于谱线峰强度、归一化后洛伦兹拟合强度、谱峰积分强度与对应元素浓度之间的关系模型和定标曲线。对于Pb元素的3种定标方法得到的线性关系的决定系数(R2)分别为0.983 85、0.970 97、0.993 21,且模型反演的结果与实际值的相对误差较小；而Cd元素的3种定标方法没有得到明显线性关系。然后运用偏最小二乘回归(partial least-squares regression,PLSR)建立了土壤Pb和Cd元素的定量分析模型,Pb元素PLSR模型的结果与定标曲线法的结果类似,其预测的相关系数(RP)为0.948 5,预测均方根误差(RMSEP)为2.044 mg/g；而Cd元素的PLSR模型的结果比起定标曲线法有较大提升,其预测的相关系数(RP)为0.994 9,预测均方根误差(RMSEP)为97.05 μg/g,结果说明PLSR方法在光谱化学分析领域中比定标曲线法进行定量分析有更好的适用性。研究表明,LIBS技术能够实现对土壤重金属Pb和Cd含量的定量检测,为开发实时便携式LIBS土壤重金属检测仪提供了理论基础。

... 光谱数据建模主要采用偏最小二乘回归(partial least-squares regression, PLSR)^{[8, 10]}和多元线性回归(multiple linear regression, MLR)^[11]来完成 ...

... 模型评价主要是利用相关系数(correlation coefficient, R)和均方根误差(root mean square error, RMSE)^{[8, 10]} ...

2018

0.0

... 光谱数据建模主要采用偏最小二乘回归(partial least-squares regression, PLSR)^{[8, 10]}和多元线性回归(multiple linear regression, MLR)^[11]来完成 ...

2016

0.0

... O键的倍频吸收有关)等物质含量变化造成的^[12] ...

2017

0.0

... O键有关^[13,14,15] ...

2014

0.0

... O键有关^[13,14,15] ...

2014

0.0

... O键有关^[13,14,15] ...

2013

0.0