烤烟烟叶反射和透射图像的同步图像分割

A Synchronous Background Segmentation Method for the Transmission and Reflection Images of Tobacco Leaves

采用自行设计的烟叶图像采集灯箱系统,对同一片烤烟烟叶在不移动位置的情况下分别采集整片烟叶的反射和透射图像。依据烤烟烟叶透射图像烟叶区域敏感吸收蓝色光的特性,提出了一种利用烟叶自身特征的图像轮廓分割法,可实现同一片烤烟烟叶反射、透射图像的同步背景扣除。结合点统计分割方法和区域统计分割方法,能成功获取具有相同有效像素轮廓的反射和透射成对烟叶的纯净图像。经研究验证,分割精度比单独使用点统计分割方法或区域统计分割方法高,并且分割边界光滑而连续,为后续的基于反射与透射图像相结合的烟叶数字分级奠定了坚实的基础。

This paper describes the design of a lamp box device that captures, from each flue-cured to- bacco leaf, transmission and reflection images in the same image position. According to the transmission characteristics of tobacco leaves namely that the blue vector value in the leaf part of its RGB transmission image is dramatically absorbed. Thus, this study proposes a novel segmentation method to synchronously treat the transmission and reflection images of the same flue-cured tobacco leaf, combining it with a point-oriented and area-oriented statistical segmenting method to successfully obtain leaf parts in pairs which have the same effective pixel quantities and the same contours. The results show that segmentation accuracy is higher than when only using point-oriented or area-oriented statistical segmenting methods. Moreover, the boundary is smooth and continuous. These findings are very helpful for digitally rating tobacco quality based on the combination of transmission and reflection images.