[1]任亚举,王瑞敏,薛冬,等.基于深度学习结合高光谱技术的大豆种子活力检测方法[J].江苏农业学报,2025,(05):927-936.[doi:doi:10.3969/j.issn.1000-4440.2025.05.011]
　REN Yaju,WANG Ruimin,XUE Dong,et al.Soybean seed vigor detection based on deep learning combined with hyperspectral technology[J].,2025,(05):927-936.[doi:doi:10.3969/j.issn.1000-4440.2025.05.011]
点击复制

基于深度学习结合高光谱技术的大豆种子活力检测方法()
分享到：

江苏农业学报[ISSN:1006-6977/CN:61-1281/TN]

卷:
期数:
2025年05期
页码:
927-936
栏目:
农业信息工程
出版日期:
2025-05-31

文章信息/Info

Title:
Soybean seed vigor detection based on deep learning combined with hyperspectral technology
作者:
任亚举1; 2; 王瑞敏2; 薛冬2; 周琰琰2; 陈新2; 袁星星2; 闫强2; 罗楚平1
(1.淮阴工学院生命科学与食品工程学院,江苏淮安223003；2.江苏省农业科学院经济作物研究所,江苏南京210014)
Author(s):
REN Yaju1; 2; WANG Ruimin2; XUE Dong2; ZHOU Yanyan2; CHEN Xin2; YUAN Xingxing2; YAN Qiang2; LUO Chuping1
(1.School of Life Science and Food Engineering, Huaiyin Institute of Technology, Huai’an 223003, China；2.Institute of Industrial Crops, Jiangsu Academy of Agricultural Sciences, Nanjing 210014, China)
关键词:
大豆; 种子活力; 检测; 高光谱; 深度学习; 注意力机制
Keywords:
soybean; seed vigor; detection; hyperspectral; deep learning; attention mechanism
分类号:
S127
DOI:
doi:10.3969/j.issn.1000-4440.2025.05.011
文献标志码:
A
摘要:
为实现大豆种子活力的高效精准无损识别,本研究以大豆品种Williams82种子为试验材料,通过不同程度的人工老化处理构建不同活力的大豆种子库,然后采集其成像高光谱图像和RGB图像,生成3个图像集(RGB数据集、SIQ数据集、ENVI数据集),利用4个深度学习模型(Vgg16Net、GoogLeNet、MobileV3Net、ResNet-34)对种子活力进行检测,筛选出较优模型和数据集。并进一步在较优模型中添加坐标注意力机制(Coordinate attention,CA)和标签平滑损失函数提高模型的检测性能及鲁棒性。结果表明,基于SIQ数据集,ResNet-34模型的训练集和验证集识别准确率分别达到97.6%和96.8%,检测性能优于其他模型和数据集组合。在ResNet-34模型中添加坐标注意力机制和标签平滑损失函数构建的CA-ResNet-34模型,基于SIQ数据集对大豆种子活力检测的准确率可达到98.5%,比原始模型ResNet-34提升1.7个百分点。本研究结果为大豆种子活力准确、无损、高效检测提供新的方法。
Abstract:
To achieve efficient, accurate, and non-destructive identification of soybean seed vigor, this study used seeds of the soybean variety Williams 82 as experimental materials. A library of soybean seeds with different levels of vigor was constructed through artificial aging treatments. Hyperspectral images and RGB images of the seeds were then collected to generate three image datasets (RGB dataset, SIQ dataset, and ENVI dataset). Four deep learning models (Vgg16Net, GoogLeNet, MobileV3Net, and ResNet-34) were employed to detect seed vigor, and the optimal models and datasets were selected. Furthermore, the coordinate attention (CA) mechanism and label smoothing loss function were incorporated into the optimal models to enhance their detection performance and robustness. The results demonstrated that using the SIQ dataset and ResNet-34 model, the recognition accuracy reached 97.6% and 96.8% on the training set and validation set, respectively. The detection performance was superior to other combinations of models and datasets. The CA-ResNet-34 model, which incorporated the CA mechanism and label smoothing loss function into the ResNet-34 model, achieved a detection accuracy of 98.5% for soybean seed vigor based on the SIQ dataset. This represented an improvement of 1.7 percentage points in accuracy compared to the original ResNet-34 model. The results of this study can provide a new method for the accurate, non-destructive, and efficient detection of soybean seed vigor.

参考文献/References:

[1]高琴梅,卢新雄,朱凌燕,等. 大豆种子老化MDA和4-HNE的含量变化相关性研究[J]. 种子,2019,38(4):1-9.
[2]田桂霞. 种子老化的生理化与分子机理研究进展[J]. 农业与技术,2018,38(10):42.
[3]ZHANG G S, XU T Y, TIAN Y W. Hyperspectral imaging-based classification of rice leaf blast severity over multiple growth stages[J]. Plant Methods,2022,18(1):123.
[4]YANG M, KANG X Y, QIU X F, et al. Method for early diagnosis of Verticillium wilt in cotton based on chlorophyll fluorescence and hyperspectral technology[J]. Computers and Electronics in Agriculture,2024,216:108497.
[5]GUO A T, HUANG W J, DONG Y Y, et al. Wheat yellow rust detection using UAV-based hyperspectral technology[J]. Remote Sensing,2021,13(1):123.
[6]GUI J S, XU H R, FEI J Y. Non-destructive detection of soybean pest based on hyperspectral image and attention-ResNet meta-learning model[J]. Sensors,2023,23(2):678.
[7]WANG Z L, FAN S X, AN T, et al. Detection of insect-damaged maize seed using hyperspectral imaging and hybrid 1D-CNN-BiLSTM model[J]. Infrared Physics & Technology,2024,137:105208.
[8]NGUYEN H D, NANSEN C. Hyperspectral remote sensing to detect leafminer-induced stress in Bok choy and spinach according to fertilizer regime and timing[J]. Pest Management Science,2020,76(6):2208-2216.
[9]张伏,禹煌,熊瑛,等. 热损伤玉米种子的高光谱无损检测[J]. 光谱学与光谱分析,2024,44(4):1165-1170.
[10]LIU F S, FU J, ZHAO R Q. Pixel-wise mechanical damage detection of waxy maize using spectral-spatial feature extraction and hyperspectral image[J]. Computers and Electronics in Agriculture,2023,209:107853.
[11]ZHANG L, SUN H, RAO Z H, et al. Hyperspectral imaging technology combined with deep forest model to identify frost-damaged rice seeds[J]. Spectrochimica Acta Part A:Molecular and Biomolecular Spectroscopy,2020,229:117973.
[12]彭彦昆,赵芳,李龙,等. 利用近红外光谱与PCA-SVM识别热损伤番茄种子[J]. 农业工程学报,2018,34(5):159-165.
[13]PENG D D, JIN C, WANG J, et al. Defects recognition of pine nuts using hyperspectral imaging and deep learning approaches[J]. Microchemical Journal,2024,201:110521.
[14]AL-AMERY M, GENEVE R L, SANCHES M F, et al. Near-infrared spectroscopy used to predict soybean seed germination and vigour[J]. Seed Science Research,2018,28(3):245-252.
[15]ZHANG T T, LU L, YANG N, et al. Integration of hyperspectral imaging,non-targeted metabolomics and machine learning for vigour prediction of naturally and accelerated aged sweetcorn seeds[J]. Food Control,2023,153:109930.
[16]AMBROSE A, KANDPAL L M, KIM M S, et al. High speed measurement of corn seed viability using hyperspectral imaging[J]. Infrared Physics & Technology,2016,75:173-179.
[17]XU P, ZHANG Y P, TAN Q, et al. Vigor identification of maize seeds by using hyperspectral imaging combined with multivariate data analysis[J]. Infrared Physics & Technology,2022,126:104361.
[18]丁子予,岳学军,曾凡国,等. 基于机器学习和深度学习的玉米种子活力光谱检测[J]. 华中农业大学学报,2023,42(3):230-240.
[19]QI H N, HUANG Z H, JIN B C, et al. SAM-GAN:an improved DCGAN for rice seed viability determination using near-infrared hyperspectral imaging[J]. Computers and Electronics in Agriculture,2024,216:108473.
[20]HONG S J, PARK S, LEE A, et al. Nondestructive prediction of pepper seed viability using single and fusion information of hyperspectral and X-ray images[J]. Sensors and Actuators A:Physical,2023,350:114151.
[21]杨波,段明磊,杨童. 基于高光谱成像技术的西瓜种子活力等级分类方法研究[J]. 河南农业科学,2022,51(9):151-158.
[22]周翔,兴旺,杨军,等. 甜菜种子活力的测定:基于近红外高光谱技术[J]. 中国农学通报,2021,37(29):7-12.
[23]黄瑶,乔爱民,孙敏. 热水加速老化黄瓜种子的试验研究[J]. 种子,2004,23(8):11-12,26.
[24]董永梅,谢宗铭,王志军,等. 热水老化对棉花种子活力及遗传完整性的影响[J]. 种子,2018,37(3):19-22,27.
[25]HUANG Y Q, DONG C, LUO X S, et al. Facial expression recognition algorithm based on improved VGG16 network[C]//IEEE. 2021 6th international symposium on computer and information processing technology. Changsha,China: IEEE,2021:480-485.
[26]SZEGEDY C, LIU W, JIA Y Q, et al. Going deeper with convolutions[C]//IEEE. 2015 IEEE conference on computer vision and pattern recognition. Boston,MA,USA: IEEE,2015:1-9.
[27]YI P Y, GU W X, ZHANG K, et al. Improvement and application of MobileNetV3-small model in the field of smart grid[C]//IEEE. 2022 2nd international conference on electrical engineering and control science. Nanjing,China:IEEE,2022:1144-1148.
[28]PATIL O S, LE D M, GRIFFIS E J, et al. Deep residual neural network (ResNet)-based adaptive control:a Lyapunov-based approach[C/OL]//IEEE. 2022 IEEE 61st conference on decision and control (CDC). Cancun,Mexico:IEEE,2022. https://arxiv. org/pdf/2404. 07385v2.
[29]WEI Y X, LIU Q X, ZHANG G X, et al. Label smoothing technique for ordinal classification in cloud assessment[C/OL]//IEEE. IGARSS 2020-2020 IEEE international geoscience and remote sensing symposium. Waikoloa,HI,USA:IEEE,2020. https://ieeexplore. ieee. org/document/9323714.
[30]ATKINSON G, LI X M, METSIS V. Conditional diffusion with label smoothing for data synthesis from examples with noisy labels[C]//IEEE. 2023 31st european signal processing conference (EUSIPCO). Helsinki,Finland:IEEE,2023:1300-1304.
[31]LIU R J, CUI B G, FANG X, et al. Super-resolution of GF-1 multispectral wide field of view images via a very deep residual coordinate attention network[J]. IEEE Geoscience and Remote Sensing Letters,2022,19:5513305.
[32]WANG E Y, SU R Z, HUANG B Y, et al. Enhancing YOLOv7-based fatigue driving detection through the integration of coordinate attention mechanism[C]//IEEE. 2023 IEEE international conference on image processing and computer applications. Changchun,China: IEEE,2023:725-729.
[33]HOU Q B, ZHOU D Q, FENG J S. Coordinate attention for efficient mobile network design[C]//IEEE. 2021 IEEE/CVF conference on computer vision and pattern recognition. Nashville,TN,USA:IEEE,2021. https://arxiv. org/abs/2103. 02907.
[34] KRSTINIC D, SERIC L, SLAPNICAR I. Comments on ‘MLCM:Multi-Label confusion matrix’[J]. IEEE Access,2023,11:40692-40697.
[35]LARIOS G, NICOLODELLI G, RIBEIRO M, et al. Soybean seed vigor discrimination by using infrared spectroscopy and machine learning algorithms[J]. Analytical Methods,2020,12(35):4303-4309.

相似文献/References:

[1]刘朝茂,李成云.玉米与大豆间作对玉米叶片衰老的影响[J].江苏农业学报,2017,(02):322.[doi:doi:10.3969/j.issn.1000-4440.2017.02.013]
　LIU Chao-mao,LI Cheng-yun.Effects of maize/soybean intercropping on maize leaf senescence[J].,2017,(05):322.[doi:doi:10.3969/j.issn.1000-4440.2017.02.013]
[2]张令瑄,谢婷婷,王瑾,等.大田条件下ＵＶ-Ｂ 辐射增强对大豆根际土壤相关指标的影响[J].江苏农业学报,2016,(01):118.[doi:10.3969/j.issn.1000-4440.2016.01.018]
　ZHANG Ling-xuan,XIE Ting-ting,WANG Jin,et al.Soybean rhizosphere soil parameters in response to enhanced UV-B radiation under field condition[J].,2016,(05):118.[doi:10.3969/j.issn.1000-4440.2016.01.018]
[3]宁丽华,何晓兰,张大勇.大豆耐盐相关基因GmNcl1功能标记的开发及验证[J].江苏农业学报,2017,(06):1227.[doi:doi:10.3969/j.issn.1000-4440.2017.06.005]
　NING Li-hua,HE Xiao-lan,ZHANG Da-yong.Development and validation of the function marker of soybean salt tolerance gene GmNcl1[J].,2017,(05):1227.[doi:doi:10.3969/j.issn.1000-4440.2017.06.005]
[4]杨艳丽,杨勇,李大红,等.转桃PpCuZnSOD基因大豆的耐旱性[J].江苏农业学报,2018,(05):978.[doi:doi:10.3969/j.issn.1000-4440.2018.05.003]
　YANG Yan-li,YANG Yong,LI Da-hong,et al.Drought tolerance of transgenic soybean with PpCuZnSOD gene[J].,2018,(05):978.[doi:doi:10.3969/j.issn.1000-4440.2018.05.003]
[5]孙彦坤,陈睿,李静,等.不同降雨年型下反枝苋和大豆光合特征的比较[J].江苏农业学报,2019,(03):554.[doi:doi:10.3969/j.issn.1000-4440.2019.03.008]
　SUN Yan-kun,CHEN Rui,LI Jing,et al.Comparison of photosynthetic characteristics between Amaranthus retroexus and Glycine max under different annual rainfall pattern[J].,2019,(05):554.[doi:doi:10.3969/j.issn.1000-4440.2019.03.008]
[6]曹媛媛,陈春,郭婷婷,等.亲和性促生菌DW12-L的定殖及其对大豆生长的影响[J].江苏农业学报,2019,(04):776.[doi:doi:10.3969/j.issn.1000-4440.2019.04.004]
　CAO Yuan yuan,CHEN Chun,GUO Ting ting,et al.Colonization of soybean affinity rhizobacteria strain DW12-L and its effects on soybean growth[J].,2019,(05):776.[doi:doi:10.3969/j.issn.1000-4440.2019.04.004]
[7]丁俊男,于少鹏,李鑫,等.生物炭对大豆生理指标和农艺性状的影响[J].江苏农业学报,2019,(04):784.[doi:doi:10.3969/j.issn.1000-4440.2019.04.005]
　DING Jun nan,YU Shao peng,LI Xin,et al.Effects of biochar application on soybean physiological indices and agronomic traits[J].,2019,(05):784.[doi:doi:10.3969/j.issn.1000-4440.2019.04.005]
[8]王新忠,卢青,张晓东,等.基于高光谱图像的黄瓜种子活力无损检测[J].江苏农业学报,2019,(05):1197.[doi:doi:10.3969/j.issn.1000-4440.2019.05.028]
　WANG Xin-zhong,LU Qing,ZHANG Xiao-dong,et al.Non-destructive detection of cucumber seeds vigor based on hyperspectral imaging[J].,2019,(05):1197.[doi:doi:10.3969/j.issn.1000-4440.2019.05.028]
[9]曹帅,杜仲阳,刘鹏,等.碱胁迫对大豆光合特性及内源激素含量的影响[J].江苏农业学报,2020,(02):284.[doi:doi:10.3969/j.issn.1000-4440.2020.02.005]
　CAO Shuai,DU Zhong-yang,LIU Peng,et al.Effects of alkaline stress on photosynthetic characteristics and endogenous hormone contents of soybean[J].,2020,(05):284.[doi:doi:10.3969/j.issn.1000-4440.2020.02.005]
[10]潘威,杨晓东,乔雨,等.抗冷测定法评价烟草种子活力[J].江苏农业学报,2020,(02):312.[doi:doi:10.3969/j.issn.1000-4440.2020.02.009]
　PAN Wei,YANG Xiao-dong,QIAO Yu,et al.Research on tobacco seeds vigor by cold resistance measurement[J].,2020,(05):312.[doi:doi:10.3969/j.issn.1000-4440.2020.02.009]

备注/Memo

备注/Memo:
收稿日期：2024-08-06基金项目：生物育种钟山实验室项目（ZSBBL-KY2023-03）；种质资源精准鉴定评价项目(005012691230229)作者简介：任亚举(1997-),男,河南新乡人,硕士研究生,主要从事农作物图像识别。(E-mail)13419891193@163.com通讯作者：闫强,yanqiang@jaas.ac.cn；罗楚平,(E-mail)luochuping@163.com
更新日期/Last Update: 2025-06-24