Nondestructive Detection of Slight Mechanical Damage of Apple by Hyperspectral Spectroscopy Based on Stacking Model
ZHANG Yue1, 2, LI Yang1, 2, SONG Yue-peng1, 2*
1. College of Mechanical and Electrical Engineering, Shandong Agricultural University, Taian 271018, China
2. Shandong Province Key Laboratory of Horticultural Machinery and Equipment, Taian 271018, China
Abstract:To detect minor mechanical damage to apples without damage, the most common Fuji apple in China was used as the research object. The spectral information of intact, just damaged and 1, 3, 6 and 24 h after damage were collected using the hyperspectral imaging. Competitive adaptive reweighted sampling and continuous projection algorithms were used to extract the feature wavelengths of apple hyperspectral data. The extracted feature wavelength image data were compressed using the minimum noise fraction transform to study damage detection of Fuji apples. Taking random forest, Support Vector Machine, and Spectral Angle Mapper Classifier algorithm as primary learners and logistic regression as secondary learners, a new Stacking model, is established to extract the slight damage area of the apple. Its performance is evaluated by establishing a training set and prediction set and comparing it with three single algorithms in primary learners. The results show that: (1) for the classification detection of damaged fruits, the detection accuracy of the stacking model for damaged samples is 100%, for intact samples, the detection accuracy is 96.67%, and the overall detection accuracy is 99.4%, indicating that the model can be effectively applied to the classification detection of Apple damage in different damage periods. (2) The stacking model is compared with the other three single algorithms for detecting damaged areas. It is found that for the newly damaged fruits, the classification accuracy of the support vector machine algorithm and the triangular algorithm is poor, both of which are less than 60%, and the classification accuracy of the random forest algorithm is relatively good, reaching more than 75%, The classification accuracy of stacking model for damaged and undamaged fruit areas reached 90.2% and 92.3% respectively. For the fruits damaged for 1~6 hours, the classification accuracy of the stacking model for the two fruit regions reached more than 92%, which was significantly better than other classification models. For the fruits damaged for 24 hours, there is little difference among the four models, all of which have a good classification effect, and all of them have a classification accuracy of more than 97%, indicating that the stacking model can extract the slightly damaged area of Apple relatively accurately. It has a high reference value for the follow-up study of fruit damage based on Hyperspectral.
张 悦,李 扬,宋月鹏. 苹果轻微机械损伤高光谱无损检测Stacking集成学习模型构建[J]. 光谱学与光谱分析, 2023, 43(07): 2272-2277.
ZHANG Yue, LI Yang, SONG Yue-peng. Nondestructive Detection of Slight Mechanical Damage of Apple by Hyperspectral Spectroscopy Based on Stacking Model. SPECTROSCOPY AND SPECTRAL ANALYSIS, 2023, 43(07): 2272-2277.
[1] Che W, Sun L, Zhang Q, et al. Computers and Electronics in Agriculture, 2018, 146: 12.
[2] Cavallo D P, Cefola M, Pace B, et al. Computers and Electronics in Agriculture, 2019, 156: 558.
[3] Pan X, Sun L, Li Y, et al. Journal of the Science of Food and Agriculture, 2019, 99(4): 1709.
[4] Rivera N V, Gómez-sanchis J, Chanona-pérez J, et al. Biosystems Engineering, 2014, 122: 91.
[5] Ferrari C, Foca G, Calvini R, et al. Chemometrics and Intelligent Laboratory Systems, 2015, 146: 108.
[6] Zhu Q, Guan J, Huang M, et al. Postharvest Biology and Technology, 2016, 114: 86.
[7] Rivera N V, Gómez-sanchis J, Chanona-pérez J, et al. Biosystems Engineering, 2014, 122: 91.
[8] Che W, Sun L, Zhang Q, et al. Computers and Electronics in Agriculture, 2018, 146: 12.
[9] GUO Zhi-ming, ZHAO Chun-jiang, HUANG Wen-qian, et al(郭志明,赵春江,黄文倩,等). Transactions of the Chinese Society for Agricultural Machinery(农业机械学报), 2015, 46(7): 227.
[10] SHEN Yu, FANG Sheng,ZHENG Ji-ye, et al(沈 宇,房 胜,郑纪业,等). Shandong Agricultural Sciences(山东农业科学), 2020, 52(2): 144.
[11] Li H, Liang Y, Xu Q, et al. Analytica Chimica Acta, 2009, 648(1): 77.
[12] Zhang P, Shen B, Ji H, et al. Food Analytical Method, 2022, 15(5): 1397.
[13] Wolpert D H. Neural Network, 1992, 5(2): 241.
[14] CHEN Zhi-jun, ZHU Zhen-chuang, SUN Shi-jun, et al(陈志君,朱振闯,孙仕军,等). Transactions of the Chinese Society of Agricultural Engineering(农业工程学报), 2021, 52(7): 195.
[15] ZHANG Hong-ming, CHEN Li-jun, LIU Wen, et al(张宏鸣,陈丽君,刘 雯,等). Transactions of the Chinese Society for Agricultural Machinery(农业机械学报), 2021, 52(7): 195.