基于多颜色空间的烟叶颜色表征方法研究

Research on Tobacco Leaf Color Characterization Method Based on Multi-Color Space

【目的】实现云南产地烟叶颜色表征,输出真正意义上能够反映云南宣威复烤厂初烤烟叶颜色特征的信息。【方法】课题组利用工业相机采集烟叶图像,对烟叶图像进行预处理、彩色图像分割,将烟叶图像分别转换到RGB、HSV、Lab等多颜色空间,采用投影寻踪法,将多颜色空间下的各颜色分量信息进行融合,使用综合值对烟叶颜色进行表征,并分析了使用多颜色空间的表征方法是否优于单颜色空间。【结果】1)融合多颜色空间计算出的颜色表征值与专家排序结果的相关系数高达0.98;2)单独使用RGB、HSV、Lab颜色空间进行颜色表征的相关系数分别为0.95、0.62、0.76,这表明多颜色空间包含了更多颜色信息,其表征值更能反映烟叶颜色特征。【结论】该方法可快速把烟叶真实的颜色信息提取出来,并且与人工排序的符合度较高,可实现在线应用,并为烟叶外观颜色均质化加工提供控制指标和数据支撑。

[Objective]To achieve color characterization of tobacco leaves from Yunnan production areas and output information that truly reflects the color characteristics of the first roasted tobacco leaves from Yunnan Xuanwei Redrying Factory.[Method]The research group used industrial cameras to capture tobacco leaf images,preprocessed and segmented them into color images.The tobacco leaf images were converted into RGB,HSV,Lab and other multi-color spaces,used the projection tracing method to fuse the information of each color component under the multi-color space,used the integrated value to characterize the tobacco color,and analyzed whether the characterization method using the multi-color space is better than the single color space.[Result]1)The correlation coefficient between the color characterization values calculated by fusing multiple color spaces and the expert ranking results was as high as 0.98.2)The correlation coefficients of color characterization using RGB,HSV,and Lab color spaces alone were 0.95,0.62 and 0.76,respectively,this indicates that multiple color spaces contain more color information and their characterization values can better reflect the color characteristics of tobacco leaves.[Conclusion]The method can quickly extract the real color information of tobacco leaves and has a high degree of conformity with the manual sorting,which can be applied online to provide control indexes and data support for the color homogenization processing of tobacco leaf appearance.