A nondestructive detection method for single maize seed germination rate based on photoacoustic spectrum deep scanning
-
摘要:目的
针对目前玉米种子发芽率快速无损检测方法易受种子表皮颜色影响的问题，拟采用光声光谱深度扫描技术提高玉米种子发芽率的检测精度。
方法选取3种不同颜色、6个品种的玉米样本，利用人工老化方法得到8种不同老化时间的玉米种子；通过调制光谱频率获得7种不同深度的光声光谱信息，并利用主成分分析分别得到最佳扫描频率和特征光谱，比较偏最小二乘法回归、BP神经网络、广义回归神经网络和支持向量回归等发芽率预测模型精度。
结果光声光谱最佳扫描频率为500 Hz，支持向量回归的预测模型精度最高，相关系数均超过0.980 0。6个品种玉米种子的发芽率预测相关系数分别为0.983 8，0.984 7，0.983 6，0.987 8，0.983 3和0.994 7，6个品种混合的玉米种子发芽率预测相关系数为0.942 1。
结论通过拓展光谱、声音和深度信息，光声光谱深度扫描技术适用于不同颜色的玉米发芽率高精度检测，具有较好的泛化能力。
Abstract:ObjectiveIn view of the problem that the existed rapid and non-destructive maize seed germination rate testing methods are easily affected by the color of seed skin, the photoacoustic spectroscopy deep scanning technology was proposed to improve the detection accuracy of maize seed germination rate.
MethodSix maize cultivar seeds with three different colors were selected and treated using artificial aging method to obtain eight kinds of maize seeds with different aging time. The photoacoustic spectrum information with seven different depths was obtained by modulating the spectral frequency. The best scanning frequency and characteristic spectrum were determined by principal component analysis method. Different modeling approaches including partial least squares regression, back propagation neural network, generalized regression neural network and support vector regression were applied for comparing the prediction accuracy to optimize maize seed germination rate model.
ResultThe best scanning frequency of photoacoustic spectrum was 500 Hz. The prediction model accuracy of support vector regression was the highest, and the correlation coefficients were all over 0.980 0. The prediction correlation coefficients of germination rates of six maize cultivar seeds were 0.983 8, 0.984 7, 0.983 6, 0.987 8, 0.983 3 and 0.994 7 respectively, while that of the mixed six cultivar maize seeds reached 0.942 1.
ConclusionThrough expanding the spectrum, sound and depth information, the photoacoustic spectrum depth scanning technology has a good generalization ability, and is suitable for high-precision germination rate detection of maize with different colors.
-
-
![]()
图 1 玉米种子光声光谱深度扫描原理
Figure 1. Photoacoustic spectroscopy deep scan schematic diagram of maize seed
![]()
图 2 光声光谱采集试验装置示意图
Figure 2. Schematic diagram of photoacoustic spectrum acquisition experimental device
![]()
图 3 ‘京黏1号’玉米的500 Hz (8 μm)原始图像和去噪图像
Figure 3. ‘Jingnian No.1’ maize original image and denoised image at 500 Hz (8 μm)
![]()
图 4 不同老化时间玉米种子的光声光谱
Figure 4. Photoacoustic spectroscopy of maize seeds with different aging time
![]()
图 5 ‘京黏1号’500 Hz (8 μm)特征光谱
Figure 5. Characteristic spectrum of ‘Jingnian No.1’ at 500 Hz (8 μm)
![]()
图 6 单一品种玉米和多种混合玉米的SVR建模结果
Figure 6. SVR modeling results of single cultivar mazie and mixed-cultivar maize
表 1 不同老化时间玉米种子的发芽率
Table 1 Germination rates of maize seeds with different aging times
% 老化时间/d
Aging time第1组
Group 1第2组
Group 2第3组
Group 3第4组
Group 4平均值
Average0 90.00 88.00 86.00 90.00 88.50 1 84.00 86.00 86.00 88.00 86.00 2 82.00 80.00 84.00 84.00 82.50 3 80.00 78.00 80.00 78.00 79.00 4 76.00 76.00 74.00 74.00 75.00 5 70.00 72.00 68.00 68.00 69.75 6 68.00 66.00 64.00 66.00 66.00 7 60.00 64.00 62.00 64.00 62.50 
下载: 导出CSV
表 2 不同深度‘京黏1号’光声光谱建模结果
Table 2 Photoacoustic spectroscopy modeling results of ‘Jingnian No.1’ with different depths
调制频率/Hz(深度/μm)
Modulation frequency
(Depth)建模方法
Modeling method训练集 Training set 预测集 Prediction set 扫描时间/min
Scan time相关系数
Correlation
coefficient均方根误差
Root mean
square error相关系数
Correlation
coefficient均方根误差
Root mean
square error1000(5.7) PLSR 0.976 2 1.923 7 0.841 7 5.270 0 5 BP 0.970 4 2.216 6 0.917 7 4.353 7 5 GRNN 0.972 6 2.792 9 0.932 5 3.853 8 5 SVR 0.994 7 1.031 4 0.976 9 1.720 7 5 800(6.4) PLSR 0.974 6 1.569 1 0.955 8 3.495 7 6 BP 0.984 9 1.544 7 0.961 3 3.789 1 6 GRNN 0.990 2 1.333 5 0.962 5 1.790 3 6 SVR 0.997 6 0.633 0 0.993 0 0.875 3 6 500(8.0) PLSR 0.992 6 1.100 5 0.967 5 1.100 3 8 BP 0.983 9 1.231 0 0.964 8 2.742 8 8 GRNN 0.989 4 1.336 2 0.961 7 2.373 5 8 SVR 0.993 6 0.971 9 0.983 6 1.010 3 8 400(9.0) PLSR 0.842 2 4.550 9 0.617 2 11.604 9 11 BP 0.990 0 1.505 2 0.806 7 3.811 1 11 GRNN 0.918 8 3.581 2 0.611 0 8.337 2 11 SVR 0.973 4 0.510 8 0.839 4 4.604 8 11 300(10.4) PLSR 0.989 1 1.271 0 0.976 5 2.115 6 15 BP 0.987 2 1.636 6 0.964 5 2.739 1 15 GRNN 0.971 6 2.336 9 0.940 8 2.657 1 15 SVR 0.994 1 0.975 1 0.980 7 1.761 4 15 200(12.0) PLSR 0.972 2 2.002 6 0.903 7 4.818 8 18 BP 0.990 9 1.353 7 0.908 6 0.812 0 18 GRNN 0.936 6 0.646 6 0.926 3 0.780 4 18 SVR 0.992 0 1.165 0 0.941 1 2.074 4 18 100(18.0) PLSR 0.908 2 3.793 4 0.848 3 4.697 7 20 BP 0.988 4 1.600 1 0.933 2 3.751 9 20 GRNN 0.927 0 3.581 1 0.913 1 3.797 6 20 SVR 0.998 7 0.952 1 0.952 1 2.885 5 20
下载: 导出CSV
表 3 单一品种玉米和混合玉米最优建模(SVR模型)结果
Table 3 Optimal modeling (SVR modle) results of single cultivar mazie and mixed-cultivar maize
玉米品种
Maize cultivar训练集 Training set 预测集 Prediction set 相关系数
Correlation coefficient均方根误差
Root mean square error相关系数
Correlation coefficient均方根误差
Root mean square error‘京糯808’ ‘Jingnuo 808’ 0.997 8 0.576 0 0.983 8 1.931 3 ‘银香’ ‘Yinxiang’ 0.994 2 0.947 5 0.984 7 1.833 8 ‘京黏1号’ ‘Jingnian No.1’ 0.993 6 0.971 9 0.983 6 1.010 3 ‘苏玉’ ‘Suyu’ 0.992 7 1.036 0 0.987 8 1.659 7 ‘黑糯4号’ ‘Heinuo No.4’ 0.994 2 0.992 4 0.983 3 1.347 2 ‘黑糯6号’ ‘Heinuo No.6’ 0.997 9 0.567 5 0.994 7 1.053 6 混合玉米 Mixed maize 0.994 5 0.974 8 0.942 1 1.387 8
下载: 导出CSV
-
[1] RAJJOU L, DUVAL M, GALLARDO K, et al. Seed germination and vigor[J]. Annu Rev Plant Biol, 2012, 63: 507-533. doi: 10.1146/annurev-arplant-042811-105550
[2] 沈宇峰, 王志安, 俞旭平, 等. 白术种子生活力测定方法及其与发芽率的相关性研究[J]. 中国中药杂志, 2008(3): 248-250. doi: 10.3321/j.issn:1001-5302.2008.03.006 [3] HIBBARD R P, MILLER E V. Biochemical studies on seed viability: I: Measurements of conductance and reduction[J]. Plant Physiol, 1928, 3(3): 335-352. doi: 10.1104/pp.3.3.335
[4] 闫彬, 杨福增, 郭文川. 基于机器视觉技术检测裂纹玉米种子[J]. 农机化研究, 2020, 42(5): 181-185. doi: 10.3969/j.issn.1003-188X.2020.05.031 [5] KUSUMANINGRUM D, LEE H, LOHUMI S, et al. Non-destructive technique for determining the viability of soybean (Glycine max) seeds using FT-NIR spectroscopy[J]. J Sci Food Agric, 2018, 98(5): 1734-1742. doi: 10.1002/jsfa.8646
[6] KIMULI D, WANG W, LAWRENCE K C, et al. Utilisation of visible/near-infrared hyperspectral images to classify aflatoxin B1 contaminated maize kernels[J]. Biosyst Eng, 2018, 166: 150-160. doi: 10.1016/j.biosystemseng.2017.11.018
[7] FENG L, ZHU S S, ZHANG C, et al. Identification of maize kernel vigor under different accelerated aging times using hyperspectral imaging[J/OL]. Molecules (Basel, Switzerland), 2018, 23(12): 3078. [2020-07-25]. https://doi.org/10.3390/molecules23123078.
[8] 赵欣欣, 刘继权, 王奇. 玉米种子对高温高湿老化的响应研究[J]. 吉林农业科学, 2015, 40(2): 11-15. [9] HERNÁNDEZ-AGUILAR C, DOMÍNGUEZ-PACHECO A, CRUZ-OREA A, et al. Depth profiles in maize (Zea mays L.) seeds studied by photoacoustic spectroscopy[J]. Int J Thermophys, 2015, 36(5/6): 891-899.
[10] 魏旭彤. 玉米种子活力影响因素研究[D]. 哈尔滨: 东北农业大学, 2018. [11] 中华人民共和国农业部, 全国农作物种子标准化技术委员会. 农作物种子检验规程 发芽试验: GB/T 3543.4—1995[S]. 北京: 国家技术监督局, 1995. [12] TIAN Z D, LI S J, WANG Y H, et al. A prediction approach using ensemble empirical mode decomposition-permutation entropy and regularized extreme learning machine for short-term wind speed[J]. Wind Energy, 2020, 23(2): 177-206. doi: 10.1002/we.2422
[13] 郭明军, 李伟光, 杨期江, 等. 基于SVD原理的PCA特征频率提取算法及其应用[J]. 华南理工大学学报(自然科学版), 2020, 48(1): 1-9. [14] SILVA G P, SALES J F, NASCIMENTO K J T, et al. Biochemical and physiological changes in Dipteryx alata Vog. seeds during germination and accelerated aging[J]. South African J Bot, 2020, 131: 84-92. doi: 10.1016/j.sajb.2020.02.007
[15] 鲜萱, 齐传东, 陈琦, 等. 人工老化处理对菠菜种子生理生化指标及萌发相关基因表达量的影响[J]. 中国农业大学学报, 2019, 24(3): 45-53. [16] 陆宇振, 杜昌文, 余常兵, 等. 红外光声光谱法测定油菜籽品质参数[J]. 分析化学, 2014, 42(2): 293-297. [17] BRYŚ A, BRYŚ J, MELLADO Á F, et al. Characterization of oil from roasted hemp seeds using the PDSC and FTIR techniques[J/OL]. J Therm Anal Calorim, 2019, 138(4): 2781-2786. [2020-07-25]. https://doi.org/10.1007/s10973-019-08640-8.
[18] LÜ G Q, DU C W, MA F, et al. In situ detection of rice leaf cuticle responses to nitrogen supplies by depth-profiling Fourier transform photoacoustic spectroscopy[J]. Spectrochim Acta Part A: Mol Biomol Spectrosc, 2020, 228: 117759. [2020-07-25]. https://doi.org/10.1016/j.saa.2019.117759.
计量
- 文章访问数: 1015
- HTML全文浏览量: 4
- PDF下载量: 902
