基于视觉词袋模型的茶机智能化控制方法

Tea machine intelligent control method based on BOVW and BP neural network

因3类名优茶在颜色上并无太大区分,纹理特征也不明显,所以在茶机智能化控制方法中主要依靠人为的主观判断进行名优茶分类,准确度较低。针对这个问题,提出一种基于视觉词袋模型的分类方法,首先对采集的茶叶图片通过数字图像处理的方法调整对比度,然后利用视觉词袋模型提取每一张图像的SIFT特征并生成视觉单词,最后将每一张图像表示成一个包含视觉单词的向量,输入到BP神经网络中进行训练并分类。实验结果表明:该方法合理有效,可以被广泛应用到茶叶分类问题中。

Classification of famous tea has been a difficult problem precisely, because the three types of famous tea have obviously little distinction in color and ARTS features. So, practical applications mainly rely on subjective human judgment. To address this problem, a novel method based on bag of visual words was proposed. The authors first collected images through digital image processing methods to adjust the contrast,and then used the bag of visual words model,to extract SIFT features from each images and generated to visual words. Finally,each image was expressed as a vector containing a visual representation of the words. Then,the vectors were entered into the BP neural network to train BP and classify tea pictures. Experimental results show that the proposed method is reasonable and effective. This method can be widely applied to tea classification problems.