1. 南京农业大学工学院，江苏 南京 210031
2. United States Department of Agriculture Agricultural Research Service (USDA/ARS), Michigan State University, East Lansing, MI 48824, USA
3. Department of Biosystems and Agricultural Engineering, Michigan State University, East Lansing, MI 48824, USA
Tomato Maturity Classification Based on Spatially Resolved Spectra
1. College of Engineering, Nanjing Agricultural University, Nanjing 210031, China
2. United States Department of Agriculture Agricultural Research Service (USDA/ARS), Michigan State University, East Lansing, MI 48824, USA
3. Department of Biosystems and Agricultural Engineering, Michigan State University, East Lansing, MI 48824, USA
摘要： 对基于空间可分辨光谱的番茄成熟度分类判别方法进行了试验研究。首先根据番茄的内部颜色，将600个番茄分为6个不同成熟度 (green, breaker, turning, pink, light red and red) ，然后用自行开发的多通道高光谱成像探头采集番茄的空间可分辨(SR)光谱，建立基于空间可分辨光谱的番茄成熟度偏最小二乘判别(PLSDA) 模型和支持向量机判别(SVMDA)模型。结果显示，对于PLSDA模型，SR光谱15为最佳分类光谱，分类正确率达到81.3%；对于SVMDA模型，SR光谱10为最佳预测分类光谱，分类正确率为86.3%。对六个成熟度等级番茄的判别分类，SVMDA模型要明显优于PLSDA模型。此外，相对于较小的光源-检测器距离SR光谱，较大的光源-检测器距离SR光谱可以获得更好的判别效果，显示出空间可分辨光谱在果蔬品质检测方面的应用潜力。
关键词：空间可分辨；成熟度；番茄；多通道高光谱成像
Abstract：A multichannel hyperspectral imaging probe with 30 optic fibers covering the wavelength range of 550～1 650 nm and the light source-detector distances of 1.5～36 mm was recently developed for property and quality assessment of food products. Spatially resolved spectra were acquired using the new developed multichannel probe for 600 “Sun Bright” tomato fruit, which were grouped into six maturity grades (i.e., green, breaker, turning, pink, light red, red), based on their internal color distributions. Partial least squares discriminant analysis (PLSDA) and support vector machine discriminant analysis (SVMDA) models for the 15 spatially resolved spectra were developed and compared to determine the optimal models for classification of tomato maturity. The results showed that for PLSDA models, SR spectra 15 gave the best classification results with the accuracy of 81.3%, while for SVMDA models, SR spectra 10 had the best accuracy of 86.3%. Overall, SVMDA models will provide better performance than PLSDA models for tomato maturity classification. Moreover, spatially resolved spectra with larger source-detector distances could offer better classification results, which suggests that spatially resolved spectra has potential for measuring fruit and vegetables quality.
[1] Zhu Q, He C, Lu R, et al. Postharvest Biology and Technology, 2015，103: 27.
[2] Xie L, Ying Y, Ying T, et al. Analytica Chimica Acta, 2017, 584(2): 379.
[3] Dorais M, Ehret D L，Papadopoulos A P. Phytochemistry Reviews, 2008, 7(2): 231.
[4] Qin J, Lu R. Postharvest Biology and Technology, 2008, 49(3): 355.
[5] Arefi A, Motlagh A M, Mollazade K，et al. Australian Journal of Crop Science, 2011, 5(10): 1144.
[6] Tiwari G, Slaughter D C，Cantwell M. Postharvest Biology and Technology, 2013, 86: 221.
[7] Sirisomboon P, Tanaka M, Kojima T, et al. Journal of Food Engineering, 2012, 112(3): 218.
[8] Clément A, Dorais M，Vernon M. Journal of Agricultural and Food Chemistry, 2018, 56(5): 1538.
[9] Gómez A H, Hu G, Wang J, et al. Computers and Electronics in Agriculture, 2006, 54(1): 44.
[10] Yibin Z, Jun W. Transactions of The Chinese Society of Agricultural Engineering, 2015, 4: 25.
[11] Zhang L，McCarthy M J. Postharvest Biology and Technology, 2012，67: 37.
[12] Qin J, Chao K, Kim M S. Postharvest Biology and Technology, 2012，71: 21.
[13] Polder G, van der Heijden G W，Young I. Transactions of the ASAE, 2002, 45(4): 1155.
[14] Clément A, Dorais M, Vernon M. Journal of Agricultural and Food Chemistry, 2008, 56(21): 9813.
[15] Peiris K, Dull G, Leffler R，et al. Journal of the American Society for Horticultural Science, 1998, 123(6): 1089.
[16] Saad A G, Jaiswal P，Jha S N. International Journal of Advanced Research, 2014, 2(12): 632.
[17] USDA. U.S. Standards for Grades of Fresh Tomatoes. United States Department of Agriculture, Agricultural Marketing Service, Washington DC, 1991.
[18] Huang Y, Lu R，Chen K. Postharvest Biology and Technology, 2017，133: 88.
