【Objective】The present paper aimed to explore the relationship between the conventional chemical composition of tobacco and the evaluation indexes of the individual appearance quality,thus to promote the intelligent development of appearance quality evaluation.【Method】The measurement of the conventional chemical composition and the appearance quality assessment of the selected samples of the selected tobacco leaves in Hunan tobacco area in 2017 were conducted. Through factor analysis, we selected the regular chemical components as input variables of BP neural network, and constructed a single appearance quality index prediction model with 7-10-1 topology.【Result】The statistical analysis of the conventional chemical composition and the appearance quality of the selected tobacco samples accorded with the normal distribution. The network model of the sample training results showed that the evaluation index of each individual appearance quality prediction model, the proportion of the error between the actual value and the target value of the interval in the range of 0-0.5 were more than 60 % network simulation, the proportion of the sample in the range of 0-1.0 error interval reached more than 90 %. The R~2 of maturity and chromaticity reached a significant level, and the R~2 of color, identity, oil and leaf structure reached a very significant level.【Conclusion】The designed BP neural network model can accurately predict the appearance quality through the conventional chemical composition of the tobacco.
Southwest China Journal of Agricultural Sciences
BP neural network
Conventional chemical composition