摘要:
为更加准确、客观、快速地评定柑橘汁胞粒化程度,改变目前主要依靠肉眼感官评级的现状。该文进行了机器打分研究,并与感官评价进行了相关分析。该研究将低场核磁成像及磁弛豫时间测试结果与柑橘汁胞含水率及汁胞粒化程度的感官分级等指标相对应,应用于柑橘汁胞粒化程度的人工智能鉴定和识别。相关分析、F检验和多重比较等数学分析表明,利用低场核磁横向弛豫时间的响应值可有效反应汁胞粒化程度。横向弛豫时间3个峰的响应值与不同汁胞粒化程度柑橘的感官评价等级及汁胞含水率、果实失重率等存在显著相关(P<0.05)或极显著相关(P<0.01)。经成对t检验分析,基于低场核磁建立的汁胞粒化评级模型与评价员得到感官平均评分的评价结果间没有显著差异(P=0.4982
Abstract:
Abstract: Citrus is one kind of the most important edible fruits, which has important nutrition and economic value. However, the phenomenon of juicy sac granulation, which causes Citrus quality to decline in the process of postharvest storage, seriously affects the Citrus' edible value and commodity value. Using the technical apparatus for automatic recognition and classification of the Citrus with juicy sac granulation has important practical significance and application potentiality. Low field pulsed nuclear magnetic resonance (LF-NMR) analysis is a kind of measurement technology that is growing, due to its advantages of nondestructive measurement, rapidity and good consistency. The degree of juicy sac granulation in Citrus is closely related to the existent form and content of water. The present study, using the LF-NMR to evaluate the juicy sac granulation of Citrus, researched the correlativity between LF-NMR intensity of different form of H2O and degree of juicy sac granulation. The degree of juice sac granulation was classified by 9 assessors' sensory evaluation, and validated by the Duncan test and the significance analysis of correlation with physiology indices. These physiology indices including moisture content, lignin content, vitamin C content, weight-loss ratio and respiratory strength, which were significantly correlated to the classification by sensory evaluation of 9 assessors, were analyzed by the correlation significance testing. It was found that the integral values (I21, I22 and I23) of 3 peaks (T21, T22 and T23) on transverse relaxation time (T2) inversion spectrum significantly reflected the juicy sac granulation and were correlated to its degree evaluated by human senses (P<0.05 or P<0.01). The reason was that the integral values (I21, I22 and I23) of 3 peaks (T21, T22 and T23) on transverse relaxation time (T2) represented the responses of 3 different forms of water (chemically combined water, immobile water and free water) in the Citrus fruit respectively. And by the linear fitting analysis with human's sensory evaluation results, a mathematic model with I21 and I23 as testing indices was established to judge juice sac granulation degree. In present study, 27 Citrus fruits with different juice sac granulation degrees were used to set up the equation for the prediction of Citrus' sac granulation degree, and 8 Citrus fruits were used for the validation of the effectiveness and accuracy of the fitting equation. However, because of its unsatisfying precision, the classification of juice sac granulation was simplified into acceptable or unacceptable on juice sac granulation based on the result of score, which was calculated according to the original fitting equation and the demand of Chinese National Standard. Therefore, the modified method got a good accuracy and practical effectiveness. The paired-samples t-test analysis indicated that there was no significant deference between artificial intelligence evaluation scores and sensory evaluation scores(P=0.4982