支持向量机在机器视觉识别茶叶中的应用研究

Study on identification of tea using computer vision based on support vector machine

本文针对茶叶的感官评定分类存在主观性强和一致性差等缺点,提出了一种新的茶叶识别分类方法,该方法是在机器视觉技术定量描述茶叶的颜色特征的基础上,根据支持向量机模式识别原理分别为碧螺春、龙井和祁红等3种茶叶建立了各自的分类识别模型。在RBF核函数下,所建立的模型最佳,3个模型的回判率都达到100%;对未知样本进行验证时,模型的识别率分别为95%、90%和100%。实验结果表明,基于支持向量机的机器视觉技术识别茶叶色泽类型是可行的。

Aiming at the deficiencies of tea classification by sensory evaluation such as result subjectivity and poor consistency, a new method of tea identification was proposed,which is based on pattern recognition theory of support vector machine （SVM） and uses computer vision to quantitatively depict tea color characteristic. The identification models for Biluochun tea,Longjing tea,and Qihong tea were built. With RBF kernel function, the back estimation rates of three models are all 100 %; while predicting unknown tea samples, the identification rates of three models are 95%,90%,and 100% respectively. The experimental results show that the proposed method is feasible.