基于BP神经网络的烟叶颜色自动分级研究

Automatic classification research on tobacco leaf color based on BP neural network

在形状、身份、油分及颜色等因素中,颜色是影响烟叶等级的一个重要因子。为实现根据烟叶颜色特征对烟叶等级的自动区分,选取贵州黔南地区中部六个等级的烟叶作为实验样本,经过样本采集、预处理,提取出样本的颜色特征值,分析RGB、HIS及HSV三种颜色模型下各颜色分量对烟叶等级影响的研究;基于BP神经网络,依据烟叶等级间差异较大的颜色分量作为神经网络输入因子,从而实现烟叶等级的预测识别。结果显示,用该模型进行烟叶等级预测识别的准确率可达89.17%,耗时仅0.39s。说明根据设计的BP神经网络可实现通过颜色特征对烟叶等级较为准确的预测,可靠性高。

Color is an important factor among those factors to leaf grades：shape,identity,oil content and color.For realizing the automatic identify of tobacco leaf grade based on color features,taking six grades of tobacco leaf as test sample supplied by Central Qiannan area,Guizhou.After taking pictures and image preprocessing,got the color feature value of samples,by analyzing how color vectors influence leaf grade under RGB,HIS and HSV three color models.According the main color factors to tobacco leaf to forecast and classify tobacco leaf grade based on BP neural network.The result shows,with this model,we can have 89.17% accuracy,just taking 0.39 second.As a result,according to the design on BP neural network can be realized an accurate predictions to leaf grades through the color features,with a high reliability.