Abstract In order to realize the rapid monitoring of quality change during tomato storage, 300 tomato hyperspectral images with different storage data were collected by hyperspectral imaging technology (HSI), and a Mahalanobis distance monitoring model for quality change during tomato storage was constructed based on extracting feature wavelengths from the defined effective wavelength bands. Then the quality change monitoring during tomato storage was realized. Firstly, the hyperspectral raw data is preprocessed using multiplicative scatter correction (MSC) combined with Savizky-Golay convolutional smoothing (SG) to eliminate the effects of baseline drift and noisy signals and so on. Secondly, based on the variation trend of the spectral curve in different bands and combined with the weight coefficient of principal components corresponding to the minimum value of Wilks Λ statistic in the whole band, the effective band that can highlight the quality change in the tomato storage process is defined. Thirdly, three feature wavelength screening methods, namely, successive projections algorithm (SPA), competitive adaptive reweighted sampling (CARS) and principal component analysis based on Wilks Λ statistics, are used to extract feature wavelengths in the full band and effective band, respectively; By comparing and analyzing the number of feature wavelengths extracted by the three methods, it is pointed out that the fusion principal component analysis based on Wilks Λ statistics can effectively reduce the data dimension and simplify the operation process. Then, the principal components screened in the full band, and effective band based on Wilks Λ statistics coupled with principal component analysis are analyzed. It is pointed out that the extraction of feature wavelength based on Wilks Λ statistics fusion principal component analysis in the effective band can avoid the masking effect of redundant information on effective information effectively and further reduce the data dimension. Finally, the advantages and disadvantages of the Mahalanobis distance monitoring model for quality change during tomato storage constructed based on the first storage day and the critical day of tomato spoilage are analyzed, and it is pointed out that the model constructed by the first storage day as the monitoring benchmark has higher effectiveness and reliability. The results show that the number of feature wavelengths extracted based on Wilks Λ statistic combined with principal component analysis under the effective band is the least (5 feature wavelengths), and the selected principal components can effectively represent the difference between the quality change during tomato storage. At the same time, it also provides an effective feature wavelength extraction method for monitoring the quality changes during tomato storage by hyperspectral imaging technology.
JIA Meng-meng,YIN Yong,YU Hui-chun, et al. Hyperspectral Imaging Combined With Feature Wavelength Screening for Monitoring the Quality Change of Tomato During Storage[J]. SPECTROSCOPY AND SPECTRAL ANALYSIS, 2023, 43(03): 969-975.
[1] MA Yong-qiang, XIU Wei-ye, LI Chen-chen, et al(马永强，修伟业，黎晨晨，等). Food Science(食品科学), 2021, 42(3): 121.
[2] LI Jian-xin, WANG Wen-ping, HU Zhang-jian, et al(李建鑫，王文平，胡璋健，等). Chinese Agricultural Science(中国农业科学), 2021, 54(8): 1728.
[3] DENG Li-li, SHEN Lin, SHENG Ji-ping(邓丽莉，申 琳，生吉萍). Food Science(食品科学), 2013, 34(19): 351.
[4] XU Li-jia, CHEN Ming, WANG Yu-chao, et al(许丽佳，陈 铭，王玉超，等). Spectroscopy and Spectral Analysis(光谱学与光谱分析), 2021, 41(7): 2188.
[5] Yu Haidong, Qing Liwei, Yan Danting, et al. Food Chemistry, 2021, 348(6): 129129.
[6] Pullanagari R R, Li Mo. Journal of Food Engineering, 2021, 289(2): 110177.
[7] Mohammadi-Moghaddam T, Razavi S M A, Taghizadeh M, et al. Journal of Food Measurement and Characterization, 2018, 12(3): 1493.
[8] Zhu H Y, Chu B Q, Zhang C, et al. Scientific Reports, 2017, 7(1): 4125.
[9] Jiang J L, Cen H Y, Zhang C, et al. Postharvest Biology and Technology, 2018, 146: 147.
[10] Xue Shuning, Yin Yong, Wang Zhihao, et al. Journal of Food Measurement and Characterization, 2021, (6): 4061.
[11] SUN Shi-peng, PENG Jun, LI Rui, et al(孙世鹏，彭 俊，李 瑞，等). Food Science(食品科学), 2017, 38(2): 301.
[12] Ariana Diwan P, Lu Renfu. Computers and Electronics in Agriculture, 2010, 74(1): 137.
