基于高光谱成像技术的山楂含水量快速无损分析

doi:10.12133/j.smartag.SA202505033

Smart Agriculture

• •

基于高光谱成像技术的山楂含水量快速无损分析

白瑞斌¹(), 王慧¹, 王宏鹏², 洪家顺³, 周骏辉¹, 杨健¹^,³()

^1. 中国中医科学院中药资源中心/道地药材品质保障与资源持续利用全国重点实验室，北京 100700，中国
^2. 浙江科技学院生物与化学工程学院，浙江杭州 310023，中国
^3. 江西省道地药材质量评价研究中心，江西赣江 330000，中国

收稿日期:2025-05-29 出版日期:2025-08-14
基金项目:
国家重点研发计划项目(2024YFC3506800); 中国中医科学院科技创新工程项目(CI2023E002); 中央本级重大增减支项目(2060302); 国家中医药管理局高水平中医药重点学科建设项目(ZYYZDXK-2023244); 财政部和农业农村部国家现代农业产业技术体系项目(CARS-21); 中国中医科学院基本科研业务费优秀青年科技人才培养专项(ZZ16-YQ-040); 中国中医科学院中药资源中心自主选题研究项目(ZZXT202312)
作者简介:
白瑞斌，博士，助理研究员，研究方向为中药资源品质评价。E-mail：bairuibin2022@163.com
通信作者:
杨健，博士，副研究员，研究方向为中药资源品质评价。E-mail：yangchem2012@163.com

Rapid and Non-Destructive Analysis of Hawthorn Moisture Content Based on Hyperspectral Imaging Technology

BAI Ruibin¹(), WANG Hui¹, WANG Hongpeng², HONG Jiashun³, ZHOU Junhui¹, YANG Jian¹^,³()

^1. State Key Laboratory for Quality Ensurance and Sustainable Use of Dao-di Herbs, National Resource Center for Chinese Materia Medica, China Academy of Chinese Medical Sciences, Beijing, 100700, China
^2. School of Biological and Chemical Engineering, Zhejiang University of Science and Technology, Hangzhou 310023, China
^3. Evaluation and Research Center of Daodi Herbs of Jiangxi Province, Nanchang 330000, China

Received:2025-05-29 Online:2025-08-14
Foundation items:National Key R&D Program of China(2024YFC3506800); Scientific and Technological Innovation Project of China Academy of Chinese Medical Sciences(CI2023E002); Major increase and decrease in expenditure at the central level(2060302); National Administration of Traditional Chinese Medicine High-level Key Discipline Construction Project of Traditional Chinese Medicine(ZYYZDXK-2023244); China Agricultural Research System of MOF and MARA(CARS-21); Excellent Young Scientists Cultivation Program of China Academy of Chinese Medical Sciences(ZZ16-YQ-040); Independent Research Project of National Resource Center for Chinese Materia Medica, China Academy of Chinese Medical Sciences(ZZXT202312)
About author:
BAI Ruibin, E-mail: bairuibin2022@163.com
Corresponding author:
YANG Jian, E-mail: yangchem2012@163.com

摘要/Abstract

摘要：

【目的/意义】 为实现山楂水分含量的快速无损检测，本研究探索了一种基于高光谱成像技术结合机器学习算法的检测方法。 【方法】 首先，收集458个来自不同产区不同品种的新鲜山楂样品，分别采集每个样品在可见-近红外波段（Visible to Near Infrared, VNIR）和短波红外（Short-Wave Infrared, SWIR）波段的高光谱数据，利用阈值分割算法确定每个山楂的感兴趣区域（Region of Interest, ROI），提取果实ROI的平均反射光谱作为原始数据。随后，采用卷积平滑、乘法散射校正、标准正态变换、一阶导数和二阶导数五种预处理方法，对原始光谱数据进行优化。在此基础上，结合偏最小二乘回归、支持向量回归（Support Vector Regression, SVR）、随机森林与多层感知机等机器学习方法，系统评估不同摆放方式（果柄朝侧面、朝上、朝下及三者融合）和光谱范围（VNIR、SWIR、VNIR+SWIR）对模型预测性能的影响。最后，采用连续投影算法、竞争自适应重加权采样算法、变量迭代空间收缩方法，以及离散小波变换-逐步回归（Discrete Wavelet Transform-Stepwise Regression, DWT-SR）四种方法对全波段数据进行降维处理，进一步减少数据冗余，提高模型效率。 【结果和讨论】 果柄朝下的摆放方式、SWIR波段范围（940~2 500 nm）及一阶导数预处理组合下，SVR模型表现最优，测试集的绝对系数（Coefficient of Determination, R²_p ）为0.860 5、平均绝对误差（Mean Absolute Error, MAE _p ）= 0.711 1、均方根误差（Root Mean Square Error, RMSE _p ）=0.914 2、相对分析误差（Ratio of Performance to Deviation, RPD）=2.677 6。在性能最优分析条件下，DWT-SR方法基于小波基函数“db6”在分解层级为1时，提取出17个关键特征波段，所建模型在降低数据维度的同时可以保持高水平预测性能（R²_p =0.857 1、MAE _p = 0.669 2、RMSE _p =0.925 2、RPD=2.645 7）。 【结论】 本研究证明了高光谱成像结合机器学习方法在山楂水分无损检测中的可行性，为果品水分在线监测及智能分选提供了理论依据与技术支撑。

关键词: 山楂, 高光谱成像, 小波变换, 支持向量机, 含水量

Abstract:

[Objective] This study aimed to develop a rapid and non-destructive method for determining the moisture content of hawthorn fruits using hyperspectral imaging (HSI) integrated with machine learning algorithms. By evaluating the effects of different fruit orientations and spectral ranges, the research provides theoretical insights and technical support for real-time moisture monitoring and intelligent fruit sorting. [Methods] A total of 458 fresh hawthorn samples, representing various regions and cultivars, were collected to ensure diversity and robustness. Hyperspectral images were acquired in two spectral ranges: visible–near-infrared (VNIR, 400–1 000 nm) and short-wave infrared (SWIR, 940－2 500 nm). A threshold segmentation algorithm was used to extract the region of interest (ROI) from each image, and the average reflectance spectrum of the ROI served as the raw input data. To enhance spectral quality and reduce noise, five preprocessing techniques were applied: Savitzky-Golay (SG) smoothing, multiplicative scatter correction (MSC), standard normal variate (SNV), first derivative (FD), and second derivative (SD). Four regression algorithms were then employed to build predictive models: partial least squares regression (PLSR), support vector regression (SVR), random forest (RF), and multilayer perceptron (MLP). The models were evaluated under varying fruit orientations (stem-side facing downward, upward, sideways, and a combined set of all three) and spectral ranges (VNIR, SWIR, and VNIR+SWIR). To further reduce the dimensionality of the hyperspectral data and minimize redundancy, four feature selection methods were applied: successive projections algorithm (SPA), competitive adaptive reweighted sampling (CARS), variable iterative space shrinkage approach (VISSA), and discrete wavelet transform combined with stepwise regression (DWT-SR). The DWT-SR method utilized the Daubechies 6 (db6) wavelet basis function at a decomposition level of 1. [Results and Discussions] Both fruit orientation and spectral range had a significant impact on model performance. The optimal prediction results were achieved when the stem-side of the fruit was facing downward, using the SWIR range (940–2 500 nm) and FD preprocessing. Under these conditions, the SVR model exhibited the highest predictive accuracy, with a coefficient of determination (R²ₚ) of 0.860 5, mean absolute error (MAEₚ) of 0.711 1, root mean square error (RMSEₚ) of 0.914 2, and residual prediction deviation (RPD) of 2.677 6. Further feature reduction using the DWT-SR method resulted in the selection of 17 key wavelengths. Despite the reduced input size, the SVR model based on these features maintained strong predictive capability, achieving R²ₚ = 0.857 1, MAEₚ = 0.669 2, RMSEₚ = 0.925 2, and RPD = 2.645 7. These findings confirm that the DWT-SR method effectively balances dimensionality reduction with model performance. The results demonstrate that the SWIR range contains more moisture-relevant spectral information than the VNIR range, and that first derivative preprocessing significantly improves the correlation between spectral features and moisture content. The SVR model proved particularly well-suited for handling nonlinear relationships in small datasets. Additionally, the DWT-SR method efficiently reduced data dimensionality while preserving key information, making it highly applicable for real-time industrial use. [Conclusions] In conclusion, hyperspectral imaging combined with appropriate preprocessing, feature selection, and machine learning techniques offers a promising and accurate approach for non-destructive moisture determination in hawthorn fruits. This method provides a valuable reference for quality control, moisture monitoring, and automated fruit sorting in the agricultural and food processing industries.

Key words: hawthorn, hyperspectral imaging, wavelet transform, support vector machine, moisture content

中图分类号:

白瑞斌, 王慧, 王宏鹏, 洪家顺, 周骏辉, 杨健. 基于高光谱成像技术的山楂含水量快速无损分析[J]. 智慧农业(中英文), doi: 10.12133/j.smartag.SA202505033.

BAI Ruibin, WANG Hui, WANG Hongpeng, HONG Jiashun, ZHOU Junhui, YANG Jian. Rapid and Non-Destructive Analysis of Hawthorn Moisture Content Based on Hyperspectral Imaging Technology[J]. Smart Agriculture, doi: 10.12133/j.smartag.SA202505033.

图/表 11

图1

图2

图3

表1

表2

在不同摆放位置和不同光谱范围时不同机器学习算法的最佳参数

摆放位置	光谱范围	回归方法	最佳参数
位置1	VNIR	PLSR	n_components= 28
		SVR	C= 10， gamma= 0.001
		RF	max_depth= 10， max_features= sqrt， min_samples_leaf= 2， min_samples_split= 10， n_estimators= 100
	SWIR	PLSR	n_components= 8
		SVR	C= 100， gamma= 0.000 1
		RF	max_depth= 15， max_features= sqrt， min_samples_leaf= 2， min_samples_split= 5， n_estimators= 300
	VNIR+SWIR	PLSR	n_components= 10
		SVR	C= 100， gamma= 0.000 1
		RF	max_depth= 10， max_features= sqrt， min_samples_leaf= 2， min_samples_split= 5， n_estimators= 300
位置2	VNIR	PLSR	n_components= 21
		SVR	C= 100， gamma= 0.001
		RF	max_depth= 10， max_features= sqrt， min_samples_leaf= 2， min_samples_split= 5， n_estimators= 300
	SWIR	PLSR	n_components= 7
		SVR	C= 100， gamma= 0.000 1
		RF	max_depth= 15， max_features= sqrt， min_samples_leaf= 2， min_samples_split= 5， n_estimators= 300
	VNIR+SWIR	PLSR	n_components= 13
		SVR	C= 10， gamma= 0.001
		RF	max_depth= 15， max_features= sqrt， min_samples_leaf= 2， min_samples_split= 5， n_estimators= 300
位置3	VNIR	PLSR	n_components= 24
		SVR	C= 10， gamma= 0.001
		RF	max_depth= 15， max_features= sqrt， min_samples_leaf= 2， min_samples_split= 10， n_estimators= 100
	SWIR	PLSR	n_components= 9
		SVR	C= 100， gamma= 0.000 1
		RF	max_depth= 10， max_features= sqrt， min_samples_leaf= 2， min_samples_split= 5， n_estimators= 300
	VNIR+SWIR	PLSR	n_components= 12
		SVR	C= 100， gamma= 0.000 1
		RF	max_depth= 15， max_features= sqrt， min_samples_leaf= 2， min_samples_split= 5， n_estimators= 100
综合位置	VNR	PLSR	n_components= 29
		SVR	C= 100， gamma= 0.001
		RF	max_depth= 15， max_features= sqrt， min_samples_leaf= 2， min_samples_split= 5， n_estimators= 300
	SWIR	PLSR	n_components= 28
		SVR	C= 10， gamma= 0.001
		RF	max_depth= 15， max_features= sqrt， min_samples_leaf= 2， min_samples_split= 5， n_estimators= 300
	VNIR+SWIR	PLSR	n_components= 25
		SVR	C= 100， gamma= 0.000 1
		RF	max_depth= 15， max_features= sqrt， min_samples_leaf= 2， min_samples_split= 5， n_estimators= 300

表2

表3

在不同摆放位置和不同光谱范围时不同机器学习算法对水分含量的预测性能

摆放位置	光谱范围	回归方法	校准集			验证集			测试集
摆放位置	光谱范围	回归方法	R ² _C	RMSE _C	MAE _C	R ² _V	RMSE _V	MAE _V	R ² _P	RMSE _P	MAE _p	RPD
位置1	VNIR	PLSR	0.784 9	1.0346	0.813 5	0.618 7	1.507 3	1.156 0	0.327 1	1.915 6	1.443 0	1.219 1
		SVR	0.687 1	1.2479	0.902 2	0.561 3	1.616 9	1.221 2	0.563 6	1.542 7	1.197 6	1.513 7
		MLP	0.835 3	0.9053	0.718 2	0.542 6	1.650 9	1.251 9	0.492 9	1.662 8	1.349 2	1.404 3
		RF	0.830 3	0.9191	0.715 0	0.475 9	1.767 3	1.314 9	0.518 6	1.620 2	1.222 6	1.441 3
	SWIR	PLSR	0.829 4	0.9749	0.693 9	0.632 7	1.347 4	1.001 0	0.719 4	1.093 5	0.877 3	1.887 9
		SVR	0.861 0	0.8799	0.493 6	0.644 7	1.325 2	0.951 8	0.781 4	0.965 3	0.751 6	2.138 8
		MLP	0.928 8	0.6295	0.459 5	0.660 8	1.294 7	0.940 5	0.753 7	1.024 5	0.785 6	2.015 0
		RF	0.896 3	0.7600	0.570 6	0.450 9	1.647 4	1.201 3	0.534 4	1.408 7	1.125 5	1.465 5
	VNIR+SWIR	PLSR	0.883 3	0.7759	0.572 2	0.752 0	1.075 6	0.817 9	0.711 4	1.361 8	1.018 3	1.861 4
		SVR	0.902 2	0.7100	0.368 1	0.752 5	1.074 6	0.786 9	0.739 6	1.293 5	0.906 6	1.959 7
		MLP	0.929 6	0.6021	0.470 6	0.708 4	1.166 2	0.939 5	0.683 9	1.425 0	1.121 0	1.778 7
		RF	0.905 6	0.6977	0.516 7	0.568 5	1.418 8	1.115 7	0.517 6	1.760 5	1.382 2	1.439 8
位置2	VNIR	PLSR	0.787 6	1.0068	0.784 7	0.727 5	1.380 5	1.057 0	0.660 5	1.497 2	1.187 2	1.716 2
		SVR	0.853 8	0.8352	0.506 1	0.686 6	1.480 5	1.132 9	0.682 8	1.447 0	1.077 7	1.775 6
		MLP	0.866 3	0.7985	0.624 9	0.630 3	1.607 7	1.224 0	0.679 2	1.455 2	1.131 2	1.765 5
		RF	0.890 3	0.7236	0.530 9	0.511 3	1.848 6	1.408 9	0.513 4	1.792 2	1.422 3	1.433 6
	SWIR	PLSR	0.787 0	1.0812	0.779 4	0.846 0	1.137 1	0.825 2	0.803 1	1.086 2	0.840 4	2.253 6
		SVR	0.841 4	0.9331	0.537 7	0.742 4	1.068 2	0.724 2	0.860 5	0.914 2	0.711 1	2.677 6
		MLP	0.890 1	0.7765	0.534 2	0.632 6	1.275 6	0.938 2	0.763 4	1.190 5	0.901 5	2.056 0
		RF	0.901 4	0.7358	0.556 4	0.452 1	1.557 9	1.163 6	0.569 0	1.607 1	1.222 1	1.523 2
	VNIR+SWIR	PLSR	0.871 6	0.7943	0.597 2	0.777 0	1.225 2	0.857 8	0.749 8	1.263 8	0.902 6	1.999 1
		SVR	0.925 6	0.6046	0.276 9	0.715 2	1.384 7	0.948 3	0.821 2	1.068 4	0.781 5	2.364 8
		MLP	0.917 6	0.6362	0.490 1	0.662 0	1.508 5	1.114 7	0.752 8	1.256 0	1.014 2	2.011 4
		RF	0.912 2	0.6570	0.478 3	0.539 6	1.760 6	1.351 4	0.574 5	1.648 0	1.268 3	1.533 1
位置3	VNIR	PLSR	0.784 5	1.0535	0.795 4	0.609 1	1.477 5	1.136 8	0.768 6	1.142 3	0.851 6	2.078 9
		SVR	0.711 8	1.2182	0.876 1	0.491 4	1.685 3	1.366 8	0.702 6	1.295 0	1.033 3	1.833 8
		MLP	0.840 2	0.9070	0.712 9	0.435 5	1.775 3	1.443 7	0.520 8	1.643 7	1.317 0	1.444 7
		RF	0.817 0	0.9708	0.739 2	0.381 3	1.858 7	1.466 3	0.593 4	1.514 2	1.168 4	1.568 3
	SWIR	PLSR	0.837 2	0.9515	0.695 9	0.683 9	1.276 1	0.912 3	0.632 2	1.193 0	0.845 5	1.648 8
		SVR	0.853 9	0.9012	0.528 2	0.705 6	1.231 6	0.896 1	0.725 4	1.030 7	0.764 6	1.908 5
		MLP	0.881 1	0.8129	0.613 5	0.694 5	1.254 3	0.920 4	0.670 7	1.128 6	0.835 2	1.742 8
		RF	0.885 9	0.7965	0.611 7	0.395 6	1.764 5	1.375 5	0.444 5	1.466 0	1.129 9	1.341 8
	VNIR+SWIR	PLSR	0.877 1	0.7875	0.583 4	0.696 3	1.399 1	1.072 8	0.815 9	0.956 8	0.718 2	2.330 5
		SVR	0.880 6	0.7760	0.424 2	0.730 1	1.319 0	0.959 0	0.849 3	0.865 6	0.669 0	2.576 0
		MLP	0.901 7	0.7040	0.521 5	0.688 3	1.417 3	1.043 9	0.697 4	1.226 3	0.917 1	1.818 1
		RF	0.903 3	0.6983	0.527 3	0.461 5	1.863 1	1.487 3	0.599 0	1.412 0	1.062 3	1.579 2
综合位置	VNR	PLSR	0.671 5	1.3146	1.027 2	0.583 6	1.482 8	1.135 9	0.592 4	1.533 8	1.191 5	1.566 3
		SVR	0.793 3	1.0428	0.686 8	0.633 0	1.392 2	1.042 1	0.663 8	1.3931	1.065 9	1.724 5
		MLP	0.815 6	0.9849	0.773 6	0.553 5	1.535 4	1.147 2	0.595 9	1.527 0	1.191 3	1.573 2
		RF	0.898 6	0.7304	0.549 8	0.477 1	1.661 8	1.259 3	0.447 8	1.785 2	1.374 8	1.345 7
	SWIR	PLSR	0.837 2	0.9489	0.712 8	0.680 0	1.253 4	0.976 0	0.707 0	1.181 8	0.918 7	1.847 4
		SVR	0.844 3	0.9281	0.550 1	0.736 0	1.138 4	0.846 2	0.758 3	1.073 5	0.788 0	2.033 9
		MLP	0.906 6	0.7185	0.551 7	0.701 1	1.211 0	0.904 1	0.699 7	1.196 4	0.910 6	1.824 8
		RF	0.898 2	0.7505	0.567 2	0.498 4	1.569 3	1.231 7	0.455 9	1.610 5	1.273 3	1.355 6
	VNIR+SWIR	PLSR	0.870 7	0.8382	0.626 6	0.670 2	1.234 3	0.918 8	0.743 4	1.248 1	0.874 6	1.974 0
		SVR	0.862 2	0.865 4	0.540 0	0.722 6	1.131 8	0.801 2	0.772 7	1.174 7	0.816 9	2.097 4
		MLP	0.874 3	0.826 3	0.636 9	0.670 8	1.232 9	0.936 4	0.750 3	1.231 0	0.885 3	2.001 4
		RF	0.913 7	0.684 7	0.513 9	0.558 5	1.428 0	1.086 0	0.548 5	1.655 6	1.278 7	1.488 2

表3

图4

表4

图5

表5

图6

参考文献 28

[1]	CUI M, CHENG L, ZHOU Z Y, et al. Traditional uses, phytochemistry, pharmacology, and safety concerns of hawthorn (Crataegus genus): A comprehensive review[J]. Journal of ethnopharmacology, 2024, 319: ID 117229.
[2]	ZHANG S Y, SUN X L, YANG X L, et al. Botany, traditional uses, phytochemistry and pharmacological activity of Crataegus pinnatifida (Chinese hawthorn): A reviewFree[J]. Journal of pharmacy and pharmacology, 2022, 74(11): 1507-1545.
[3]	张海芳, 纳日, 韩育梅, 等. 光谱无损检测技术在农产品产地溯源中的研究进展[J]. 食品工业科技, 2023, 44(8): 17-25.
	ZHANG H F, NA R, HAN Y M, et al. Research progress of spectral nondestructive testing technology in traceability of agricultural products[J]. Science and technology of food industry, 2023, 44(8): 17-25.
[4]	DI Y B, LUO H P, LIU H Y, et al. Quantitative detection of water content of winter jujubes based on spectral morphological features[J]. Agriculture, 2025, 15(5): ID 482.
[5]	YANG Y C, WIJEWARDANE N K, HARVEY L, et al. Sweetpotato moisture content and textural property estimation using hyperspectral imaging and machine learning[J]. Journal of food measurement and characterization, 2025, 19(4): 2700-2716.
[6]	CHEN Y Y, LI S P, ZHANG X B, et al. Prediction of apple moisture content based on hyperspectral imaging combined with neural network modeling[J]. Scientia horticulturae, 2024, 338: ID 113739.
[7]	MENG Q L, FENG S N, TAN T, et al. Visualisation of moisture content distribution maps and classification of freshness level of loquats[J]. Journal of food composition and analysis, 2024, 131: ID 106265.
[8]	RUNGPICHAYAPICHET P, CHAIYARATTANACHOTE N, KHUWIJITJARU P, et al. Comparison of near-infrared spectroscopy and hyperspectral imaging for internal quality determination of 'Nam Dok Mai' mango during ripening[J]. Journal of food measurement and characterization, 2023, 17(2): 1501-1514.
[9]	LI Y J, MA B X, LI C, et al. Accurate prediction of soluble solid content in dried Hami jujube using SWIR hyperspectral imaging with comparative analysis of models[J]. Computers and electronics in agriculture, 2022, 193: ID 106655.
[10]	中华人民共和国国家卫生和计划生育委员会. 食品安全国家标准食品中水分的测定: GB 5009.3—2016 [S]. 北京: 中国标准出版社, 2017.National Food Safety Standard Determination of Moisture in Food: GB 5009.3—2016 [S]. Beijing: Standards Press of China, 2017.
[11]	刘玲玲, 王游游, 杨健, 等. 基于高光谱技术的枸杞子化学成分含量快速检测技术研究[J]. 中国中药杂志, 2023, 48(16): 4328-4336.
	LIU L L, WANG Y Y, YANG J, et al. Rapid detection technology of chemical component content in Lycii Fructus based on hyperspectral technology[J]. China journal of Chinese materia Medica, 2023, 48(16): 4328-4336.
[12]	ZHOU X, SUN J, TIAN Y, et al. Detection of heavy metal lead in lettuce leaves based on fluorescence hyperspectral technology combined with deep learning algorithm[J]. Spectrochimica acta part A: Molecular and biomolecular spectroscopy, 2022, 266: ID 120460.
[13]	ISLAM ELMANAWY A, SUN D W, ABDALLA A, et al. HSI-PP: A flexible open-source software for hyperspectral imaging-based plant phenotyping[J]. Computers and electronics in agriculture, 2022, 200: ID 107248.
[14]	周聪, 王慧, 杨健, 等. 基于高光谱成像技术的中药栀子产地识别[J]. 中国中药杂志, 2022, 47(22): 6027-6033.
	ZHOU C, WANG H, YANG J, et al. Origin identification of Gardeniae Fructus based on hyperspectral imaging technology[J]. China journal of Chinese materia Medica, 2022, 47(22): 6027-6033.
[15]	李璇, 袁希平, 甘淑, 等. 多变分模态分解下的湿地植被高光谱识别特征波长优选与模型研究[J]. 光谱学与光谱分析, 2025, 45(3): 601-607.
	LI X, YUAN X P, GAN S, et al. Modelling wetland vegetation identification at multiple variational mode decomposition[J]. Spectroscopy and spectral analysis, 2025, 45(3): 601-607.
[16]	王飞. 基于变量迭代空间收缩法的土壤有机质含量高光谱快速检测[J]. 水利科技与经济, 2021, 27(11): 8-12.
	WANG F. Hyperspectral rapid detection of soil organic matter content based on variable iterative spatial shrinkage method[J]. Water conservancy science and technology and economy, 2021, 27(11): 8-12.
[17]	ZHOU X, SUN J, TIAN Y, et al. Development of deep learning method for lead content prediction of lettuce leaf using hyperspectral images[J]. International journal of remote sensing, 2020, 41(6): 2263-2276.
[18]	JI Y M, SUN L J, LI Y S, et al. Detection of bruised potatoes using hyperspectral imaging technique based on discrete wavelet transform[J]. Infrared physics & technology, 2019, 103: ID 103054.
[19]	ZHOU X, ZHAO C J, SUN J, et al. Nondestructive testing and visualization of compound heavy metals in lettuce leaves using fluorescence hyperspectral imaging[J]. Spectrochimica acta part A: Molecular and biomolecular spectroscopy, 2023, 291: ID 122337.
[20]	杨宝华, 高远, 王梦玄, 等. 基于光谱-空间特征的黄茶多酚含量估算模型[J]. 光谱学与光谱分析, 2021, 41(3): 936-942.
	YANG B H, GAO Y, WANG M X, et al. Estimation model of polyphenols content in yellow tea based on spectral-spatial features[J]. Spectroscopy and spectral analysis, 2021, 41(3): 936-942.
[21]	WANG J, CAI Z Y, JIN C, et al. Species classification and origin identification of Lonicerae japonicae Flos and Lonicerae Flos using hyperspectral imaging with support vector machine[J]. Journal of food composition and analysis, 2024, 132: ID 106356.
[22]	杨唯瀚, 郝经文, 黄和平, 等. 近红外漫反射光谱法快速测定蕨菜多糖含量的研究[J]. 中国现代应用药学, 2023, 40(5): 597-602.
	YANG W H, HAO J W, HUANG H P, et al. Rapid determination of polysaccharide in Pteridium aquilinum by near infrared diffuse reflectance spectroscopy[J]. Chinese journal of modern applied pharmacy, 2023, 40(5): 597-602.
[23]	ALLAM M, ZHANG L F, SUN X J, et al. Enhancing chlorophyll-a predictions using optimal machine learning models and field spectral reflectance[J]. Earth science informatics, 2025, 18(2): ID 384.
[24]	LI P, TANG S Q, CHEN S H, et al. Hyperspectral imaging combined with convolutional neural network for accurately detecting adulteration in Atlantic salmon[J]. Food control, 2023, 147: ID 109573.
[25]	ZHONG Q D, ZHANG H, TANG S Q, et al. Feasibility study of combining hyperspectral imaging with deep learning for chestnut-quality detection[J]. Foods, 2023, 12(10): ID 2089.
[26]	ZENG F Y, SHAO W D, KANG J M, et al. Detection of moisture content in salted sea cucumbers by hyperspectral and low field nuclear magnetic resonance based on deep learning network framework[J]. Food research international, 2022, 156: ID 111174.
[27]	GUO Z, ZHANG J, DONG H W, et al. Spatio-temporal distribution patterns and quantitative detection of aflatoxin B1 and total aflatoxin in peanut kernels explored by short-wave infrared hyperspectral imaging[J]. Food chemistry, 2023, 424: ID 136441.
[28]	MA T, TSUCHIKAWA S, INAGAKI T. Rapid and non-destructive seed viability prediction using near-infrared hyperspectral imaging coupled with a deep learning approach[J]. Computers and electronics in agriculture, 2020, 177: ID 105683.

预处理方法	潜变量	校准集			验证集			测试集
预处理方法	潜变量	R ² _C	RMSE _C	MAE _C	R ² _V	RMSE _V	MAE _V	R ² _P	RMSE _P	MAE _P	RPD
无	24	0.777 5	1.099 4	0.822 0	0.726 6	1.123 7	0.790 5	0.717 4	1.309 7	0.972 7	1.881 2
SG	29	0.772 5	1.111 8	0.832 3	0.718 0	1.141 2	0.822 8	0.726 5	1.288 5	0.969 4	1.912 2
MSC	10	0.605 2	1.464 6	1.136 2	0.330 1	1.221 8	1.391 9	0.547 6	1.657 2	1.287 3	1.486 7
SNV	20	0.681 5	1.315 5	0.994 2	0.647 9	1.275 3	0.970 6	0.594 7	1.568 5	1.151 5	1.570 8
FD	25	0.870 7	0.838 2	0.626 6	0.670 2	1.234 3	0.918 8	0.743 4	1.248 1	0.874 6	1.974 0
SD	29	0.872 7	0.831 6	0.621 8	0.631 4	1.304 9	0.957 1	0.731 5	1.276 8	0.898 7	1.929 7

数据处理方法	特征数	校准集			验证集			测试集
数据处理方法	特征数	R ² _C	RMSE _C	MAE _C	R ² _V	RMSE _V	MAE _V	R ² _P	RMSE _P	MAE _P	RPD
FD-SPA	35	0.584 3	1.510 6	1.043 0	0.524 9	1.450 7	1.096 9	0.604 4	1.539 7	1.169 2	1.589 8
FD-CARS	77	0.841 8	0.931 7	0.554 2	0.699 9	1.152 9	0.824 4	0.848 9	0.951 5	0.726 9	2.572 6
FD-VISSA	176	0.953 2	0.507 1	0.228 5	0.722 3	1.109 0	0.743 5	0.841 2	0.975 6	0.765 7	2.509 1

小波基函数	特征数	校准集			验证集			测试集
小波基函数	特征数	R ² _C	RMSE _C	MAE _C	R ² _V	RMSE _V	MAE _V	R ² _P	RMSE _P	MAE _P	RPD
db4	16	0.769 8	1.124 1	0.795 1	0.698 8	1.155 0	1.053 4	0.829 8	1.010 0	0.779 0	2.423 7
db6	17	0.785 8	1.084 4	0.653 0	0.715 6	1.122 3	1.060 5	0.857 1	0.925 2	0.669 2	2.645 7
sym5	22	0.795 9	1.058 3	0.679 5	0.721 7	1.110 3	1.036 2	0.829 0	1.012 2	0.784 5	2.418 3
coif3	19	0.765 5	1.134 6	0.726 6	0.635 6	1.270 5	1.156 2	0.826 0	1.021 2	0.750 1	2.397 1

基于高光谱成像技术的山楂含水量快速无损分析

Rapid and Non-Destructive Analysis of Hawthorn Moisture Content Based on Hyperspectral Imaging Technology

在线阅读

知网下载

本地下载

可视化

摘要/Abstract

引用本文

使用本文

图/表 11

参考文献 28

相关文章 14

编辑推荐

Metrics

本文评价

[1]	许瑞峰, 王瑶华, 丁文勇, 於俊琦, 闫茂仓, 陈琛. 基于改进YOLOv8和多元特征的对虾发病检测方法[J]. 智慧农业(中英文), 2024, 6(2): 62-71.
[2]	唐超礼, 李浩, 王儒敬, 王乐, 黄青, 王大朋, 张家宝, 陈翔宇. 非接触电导检测土壤养分离子的谱峰自动识别方法[J]. 智慧农业(中英文), 2024, 6(1): 36-45.
[3]	李嘉豪, 瞿宏俊, 高名喆, 仝德之, 郭亚. 基于PADC-PCNN与平稳小波变换多焦距绿色植株图像融合算法[J]. 智慧农业(中英文), 2023, 5(3): 121-131.
[4]	王敬湧, 张明珍, 凌华荣, 王梓廷, 盖倞尧. 干旱胁迫下玉米叶片叶绿素含量与含水量高光谱成像反演方法[J]. 智慧农业(中英文), 2023, 5(3): 142-153.
[5]	赖佳政, 李贝贝, 程翔, 孙丰, 陈炬廷, 王晶, 张芊, 叶协锋. 基于无人机高光谱遥感的烤烟叶片叶绿素含量估测[J]. 智慧农业(中英文), 2023, 5(2): 68-81.
[6]	魏永康, 杨天聪, 丁信尧, 高越之, 袁鑫茹, 贺利, 王永华, 段剑钊, 冯伟. 基于不同空间分辨率无人机多光谱遥感影像的小麦倒伏区域识别方法[J]. 智慧农业(中英文), 2023, 5(2): 56-67.
[7]	张文瀚, 杜克明, 孙彦坤, 刘布春, 孙忠富, 马浚诚, 郑飞翔. 基于探地雷达的田块尺度下不同深度土壤含水量监测[J]. 智慧农业(中英文), 2022, 4(1): 84-96.
[8]	赵晋陵, 杜世州, 黄林生. 联合多源多时相卫星影像和支持向量机的小麦白粉病监测方法[J]. 智慧农业(中英文), 2022, 4(1): 17-28.
[9]	肖仕杰, 王巧华, 李春芳, 赵利梅, 刘鑫雅, 卢士宇, 张淑君. 基于随机蛙跳和支持向量机的牛乳收购分级模型构建[J]. 智慧农业(中英文), 2021, 3(4): 77-85.
[10]	李东博, 黄铝文, 赵旭博. 基于介电特征的苹果霉心病检测方法[J]. 智慧农业(中英文), 2021, 3(4): 66-76.
[11]	FLORESPaulo, MATHEWJithin, JAHANNusrat, STENGERJohn. 利用图像和机器学习检测大豆作物幼苗期玉米杂苗[J]. 智慧农业(中英文), 2020, 2(3): 61-74.
[12]	刘恬琳, 朱西存, 白雪源, 彭玉凤, 李美炫, 田中宇, 姜远茂, 杨贵军. 土壤有机质含量高光谱估测模型构建及精度对比[J]. 智慧农业(中英文), 2020, 2(3): 129-138.
[13]	万亮, 岑海燕, 朱姜蓬, 张佳菲, 杜晓月, 何勇. 基于纹理特征与植被指数融合的水稻含水量无人机遥感监测[J]. 智慧农业(中英文), 2020, 2(1): 58-67.
[14]	周超, 徐大明, 吝凯, 陈澜, 张松, 孙传恒, 杨信廷. 基于近红外机器视觉的鱼类摄食强度评估方法研究[J]. 智慧农业(中英文), 2019, 1(1): 76-84.