烟草科技2026，Vol.59Issue(5)：44-55,12.DOI:10.16135/j.issn1002-0861.2025.0347
基于图像数据的烘烤过程烟叶颜色值估测模型构建与应用
Construction and application of tobacco leaf color quantization model based on image data from curing process
摘要
Abstract
In order to establish a method to quantify tobacco leaf color during the curing process,image data from the tobacco leaf samples in the curing process in Qujing and Sanmenxia producing areas were collected,and the color CIELab true values of tobacco leaves were calibrated.The image data during the leaf curing process were collected synchronously,and the color parameters were extracted based on the color models such as RGB and YUV.A stepwise regression method was used to establish the prediction model of the color CIELab true values based on the image data and was applied for verification.The results showed that:1)Based on the leaf color calibration results by dispersive colorimeter,the brightness value L and yellowness value b of tobacco leaves increased first and then stabilized or decreased slightly during the curing process.And the redness value a raised sharply first and then increased slightly.2)Quantitative analysis of the leaf images based on the different color models revealed that during curing,the values of lightness(L),yellowness(b),blue component(B),luminance(Y)and blue-difference chrominance component(U)initially increased and then slightly decreased or stabilized.In contrast,the values of redness(a),red component(R)and red-difference chrominance(V)increased sharply at the early stage and subsequently raised more gradually,while the green component value(G)continuously declined.3)Regression models for predicting the true leaf color values were established based on image-extracted color parameters.The prediction models for L,a,and b achieved coefficients of determination(R²)of 0.89,0.97 and 0.77 respectively,with relative prediction errors of 2.3%and 3.6%,15.9%and 16.8%,and 4.9%and 6.0%for the two production regions.4)Validation trials using extensive image data from the curing process showed that the variation pattern of the predicted leaf color values were consistent with that of the actual values,with L value ranging from 40 to 75,and a and b values ranging from-15 to 17 and 20 to 57,respectively.Based on the quantitative analysis of leaf images during the curing process,the rapid and accurate inversion of actual values of tobacco leaf color could be realized.This study provides a method for accurately quantifying tobacco leaf color and identifying tobacco leaf status using image analysis.关键词
烟叶烘烤/烟叶颜色/烘烤图像/CIELab真实值/智能烘烤Key words
Tobacco curing/Tobacco leaf color/Curing image/CIELab true value/Intelligent curing分类
轻工纺织引用本文复制引用
师永刚,王远辉,徐嫱,过伟民,刘剑君,周硕野..基于图像数据的烘烤过程烟叶颜色值估测模型构建与应用[J].烟草科技,2026,59(5):44-55,12.基金项目
中国烟草总公司重大科技项目"烟叶高质量科学图像数据集构建关键技术研究"(110202201051) （110202201051）
河南省烟草公司重点科技项目"基于图像识别技术的热泵烤房智能烘烤技术研究与应用"(2024410000240029). （2024410000240029）
