-
开放科学(资源服务)标志码(OSID):
-
枇杷(Eriobotrya japonica Lindl.)属于蔷薇科(Rosaceae)枇杷属常绿多年生植物,原产于中国南方及与云南省交界的老挝、越南等东南亚国家[1]. 枇杷果实含有丰富的矿物质、糖类和有机酸等营养成分,根据果肉颜色可分为红肉枇杷和白肉枇杷两种类型[2]. 枇杷属于呼吸非跃变型果实,成熟度低的枇杷糖分低、含酸量高,鲜食口感偏差; 八成熟以上果实含糖量持续增加,酸度降低[3-4]. 当前,枇杷采收后的品质分级主要通过肉眼可观测和可触摸的外观指标(颜色、大小、硬度、质量等)来进行[5],但人的主观判断容易受很多因素的影响,效率和准确性都较差. 采用理化实验的方法可以对枇杷果实的糖度、酸度、硬度等进行较高精度的检测,但这些检测方法均为破坏性实验,需要耗费大量的化学试剂且费时费力. 而采用光学无损检测,建立数学模型后,仅通过机器扫描就能定量分析果实中的营养指标含量,这为我国未来水果成熟度的准确预测以及果品质量分级制度的建立带来了可能性.
高光谱成像技术(hyperspectral imaging,HSI)利用很多窄的电磁波波段的电磁光谱以成像的形式获取物体特性的有关数据,把传统的二维成像技术和光谱技术有机地结合在一起,能同时分析样品的光谱信息和相应的空间信息[6-7],具有“图谱合一”的特点[8]. 高光谱成像获取的原始图像是三维的,是一系列光波波长处的光学图像,图像像素的横坐标轴和纵坐标轴分别用x和y表示,光谱的波长信息以λ(z轴)表示[9]. 当前,一些学者研究了高光谱技术对香蕉[10-11]、苹果[12]、桃子[13]等水果品质及成熟度的无损检测,取得了较好的效果,但关于枇杷果实品质与成熟度的高光谱成像检测研究鲜有报道.
本文以枇杷为研究对象,在可见-近红外(363~1 026 nm)波长区域内获得高光谱信息,建立果实光谱信息与可溶性固形物(soluble solid content,SSC)、硬度、成熟度回归模型,旨在对HSI用于枇杷果实品质无损检测和成熟度的预测潜力进行评估.
Detection of Loquat Fruit Quality Based on Hyperspectral Imaging
-
摘要: 采用高光谱成像技术(HSI)在可见/近红外(363~1 026 nm)区域检测枇杷果实的可溶性固形物(SSC)和硬度,并判断其成熟度,以实现枇杷果实品质的无损检测和分级分选. 利用蒙特卡洛法(MC)剔除异常样本,基于联合X-Y距离(SPXY)进行建模集和预测集样本的划分,再采用竞争性自适应权重采样算法(CARS)和连续投影算法(SPA)选取特征波长,与全波段光谱(FS)比较,分别建立偏最小二乘回归(PLSR)模型. 结果显示,CARS-PLSR模型更优,CARS提取的SSC特征波长和硬度特征波长分别占总波长的8.52%和5.36%,枇杷果实中SSC和硬度的建模集相关系数Rc分别为0.981 7,0.970 7,预测集相关系数Rp分别为0.918 5,0.742 3,说明CARS能有效地对光谱进行降维,简化了数据处理过程. 枇杷果实SSC和硬度的变化与果实成熟度显著相关,建立判别偏最小二乘法(DPLS)成熟度预测模型,预测集总识别准确率为89.29%. 由此说明,高光谱成像技术可对枇杷品质进行有效检测,为枇杷果实的无损检测和分级分选提供了理论依据.Abstract: Hyperspectral imaging technology was used to detect the soluble solid content (SSC), firmness and maturity of loquat fruits in visible and near infrared (363~1 026 nm) region, in order to achieve nondestructive testing and sorting for loquat fruit. Abnormal samples were eliminated by using Monte Carlo (MC) method and the remaining samples were divided into calibration set and prediction set based on joint X-Y distances (SPXY). Then the competitive adaptive reweighted sampling (CARS) and successive projections algorithm (SPA) were used to select characteristic wavelengths, and compared with full spectrum (FS), the partial least squares regression (PLSR) models were established, respectively. The results showed that the CARS-PLSR models were better. The characteristic wavelengths of SSC and firmness extracted by CARS accounted for 8.52% and 5.36% of the total wavelength, respectively. The correlation coefficients Rc of calibration set for SSC and firmness of loquat fruit were 0.981 7 and 0.970 7, and the correlation coefficients Rp of prediction set were 0.918 5 and 0.742 3, respectively, which indicated that CARS effectively reduced the dimension of spectrum and simplified the data processing. There was a significant correlation between the changes of SSC and firmness of loquat fruit with fruit maturity. The discriminant partial least squares (DPLS) model of maturity was established and the total recognition accuracy of prediction set was 89.29%. Therefore, hyperspectral imaging technology can effectively detect the quality and maturity of loquat fruit, also can provide a theoretical basis for nondestructive rapid detection, sorting and grading of loquat fruit.
-
Key words:
- loquat /
- hyperspectral imaging /
- quality detection /
- PLSR /
- DPLS .
-
表 1 枇杷品质指标理化测定结果
品质指标 建模集(样本数85) 预测集(样本数28) 最小值 最大值 平均值 标准偏差 最小值 最大值 平均值 标准偏差 SSC/°Brix 5.90 13.57 8.85 1.53 6.77 11.87 8.40 1.07 硬度/(kg·cm-2) 4.07 7.17 5.90 0.47 5.53 6.63 5.99 0.26 表 2 不同特征波长下SSC的PLSR模型
品质指标 特征波长提取方法 波长数 建模集 预测集 Rc RMSEC Rp RMSEP SSC/°Brix FS(全光谱) 1 232 0.905 0 0.647 5 0.780 9 0.617 2 CARS 105 0.981 7 0.294 2 0.918 5 0.373 8 SPA 12 0.689 5 1.102 6 0.504 0 0.903 4 表 3 不同特征波长下硬度值的PLSR模型
品质指标 特征波长提取方法 波长数 建模集 预测集 Rc RMSEC Rp RMSEP 硬度/(kg·cm-2) FS(全光谱) 1 232 0.805 1 0.278 6 0.728 8 0.202 8 CARS 66 0.970 7 0.113 5 0.742 3 0.165 2 SPA 10 0.690 6 0.409 3 0.630 2 0.298 9 表 4 成熟度预测的样本划分及预测结果
成熟度 校正集样本数/个 预测集样本数/个 预测集误判数/个 预测集判别正确率/% 八成熟及以上 33 6 2 66.67 七成熟 52 22 1 95.45 合计 85 28 3 89.29 -
[1] SULTAN M Z. Loquat (Eriobotrya Japonica Lindl. ), Fruit and Vegetable Phytochemicals: Chemistry and Human Health[M]. 2nd Edition. West Sussex: John Wiley & Sons Ltd, 2018. [2] ALOS E, MARTINEZ-FUENTES A, REIG C, et al. Involvement of Ethylene in Color Changes and Carotenoid Biosynthesis in Loquat Fruit (Eriobotrya Japonica Lindl. Cv. Algerie)[J]. Postharvest Biology and Technology, 2019, 149: 129-138. doi: 10.1016/j.postharvbio.2018.11.022 [3] 邓朝军, 吴琼, 许奇志, 等. 不同成熟度'贵妃'枇杷果实色泽与糖酸含量关系[J]. 热带作物学报, 2016, 37(9): 1747-1751. doi: 10.3969/j.issn.1000-2561.2016.09.015 [4] 许奇志, 陈秀萍, 邓朝军, 等. 不同成熟度"新白8号"枇杷果实品质特性分析[J]. 中国南方果树, 2019, 48(2): 71-74, 81. doi: https://www.cnki.com.cn/Article/CJFDTOTAL-FRUI201902017.htm [5] CAÑETE M L, HUESO J J, PINILLOS V, et al. Ripening Degree at Harvest Affects Bruising Susceptibility and Fruit Sensorial Traits of Loquat (Eriobotrya Japonica Lindl. )[J]. Scientia Horticulturae, 2015, 187: 102-107. doi: 10.1016/j.scienta.2015.03.008 [6] CUBERO S, ALEIXOS N, MOLTÓ E, et al. Advances in Machine Vision Applications for Automatic Inspection and Quality Evaluation of Fruits and Vegetables[J]. Food and Bioprocess Technology, 2011, 4(4): 487-504. doi: 10.1007/s11947-010-0411-8 [7] 高迎旺, 耿金凤, 饶秀勤. 果蔬采后内部损伤无损检测研究进展[J]. 食品科学, 2017, 38(15): 277-287. doi: 10.7506/spkx1002-6630-201715044 [8] 张义志, 王瑞, 张伟峰, 等. 高光谱技术检测农产品成熟度研究进展[J]. 湖北农业科学, 2020, 59(12): 5-8, 12. doi: https://www.cnki.com.cn/Article/CJFDTOTAL-HBNY202012002.htm [9] GOWEN A A, O'DONNELL C P, CULLEN P J, et al. Hyperspectral Imaging-an Emerging Process Analytical Tool for Food Quality and Safety Control[J]. Trends in Food Science & Technology, 2007, 18(12): 590-598. [10] RAJKUMAR P, WANG N, EIMASRY G, et al. Studies on Banana Fruit Quality and Maturity Stages Using Hyperspectral Imaging[J]. Journal of Food Engineering, 2012, 108(1): 194-200. doi: 10.1016/j.jfoodeng.2011.05.002 [11] XIE C Q, CHU B Q, HE Y. Prediction of Banana Color and Firmness Using a Novel Wavelengths Selection Method of Hyperspectral Imaging[J]. Food Chemistry, 2018, 245(15): 132-140. [12] MA T, XIA Y, INAGAKI T, et al. Rapid and Nondestructive Evaluation of Soluble Solids Content (SSC) and Firmness in Apple Using Vis-NIR Spatially Resolved Spectroscopy[J]. Postharvest Biology and Technology, 2021, 173: 111417. doi: 10.1016/j.postharvbio.2020.111417 [13] 邵园园, 王永贤, 玄冠涛, 等. 基于高光谱成像的肥城桃品质可视化分析与成熟度检测[J]. 农业机械学报, 2020, 51(8): 344-350. doi: https://www.cnki.com.cn/Article/CJFDTOTAL-NYJX202008038.htm [14] LI J B, RAO X Q, YING Y B. Detection of Common Defects on Oranges Using Hyperspectral Reflectance Imaging[J]. Computers and Electronics in Agriculture, 2011, 78(1): 38-48. doi: 10.1016/j.compag.2011.05.010 [15] LIU F, JIN Z L, NAEEM M S, et al. Applying Near-Infrared Spectroscopy and Chemometrics to Determine Total Amino Acids in Herbicide-Stressed Oilseed Rape Leaves[J]. Food and Bioprocess Technology, 2011, 4(7): 1314-1321. doi: 10.1007/s11947-010-0445-y [16] 何凤芸, 杨晓东, 周胜灵, 等. 基于FT-MIR的三七粉末掺伪成分的定量检测模型研究[J]. 西南大学学报(自然科学版), 2022, 44(4): 54-61. doi: http://xbgjxt.swu.edu.cn/article/doi/10.13718/j.cnki.xdzk.2022.04.007 [17] WOLD H. Estimation of Principal Components and Related Models by Iterative Least Squares[J]. Journal of Multivariate Analysis, 1966, 1: 391-420. [18] 吴习宇, 祝诗平, 王谦, 等. DPLS和SVM的掺假花椒粉近红外光谱定性鉴别[J]. 光谱学与光谱分析, 2018, 38(8): 2369-2373. doi: https://www.cnki.com.cn/Article/CJFDTOTAL-GUAN201808012.htm [19] 杨晓玉, 刘贵珊, 丁佳兴, 等. 灵武长枣VC含量的高光谱快速检测研究[J]. 光谱学与光谱分析, 2019, 39(1): 230-234. doi: https://www.cnki.com.cn/Article/CJFDTOTAL-GUAN201901041.htm