针对传统滤波方法在茶叶特征选择上都存在一定的盲目性,以及茶叶类别数不确定等问题,提出为每一类茶叶都配备一个专属0-1分类器的验真方法。其中正样本是目标茶叶本身,标签为1,负样本是其余茶叶类型,标签为0,训练过程中迫使模型自动提取出最适合于区分目标茶叶的隐式特征进行验真或验假,同时使用孪生网络对负样本进行筛选,缓解了正负样本不平衡的问题。试验结果表明,该方法很好的适应了茶叶类别数不确定因素的干扰,具有较强的鲁棒特性,是一种有效可行的方法。
Owing to blindness in the selection of tea features in traditional filtering methods and the uncertainty of tea categories, a verification method was proposed to equip each type of tea with a exclusive 0-1 classifier. The positive sample is the target tea itself, the label is 1. The negative sample is the remaining tea type, and the label is 0. During the training process, the model is forced to automatically extract the implicit features that are most suitable for distinguishing the target tea for true and false. This method uses the Siamese network to screen the negative samples, which alleviates the problem of imbalance between positive and negative samples. The experimental results show that this method is well adapted to the disturbance of uncertainty of tea categories and has strong robustness. It is an effective and feasible method.
[1] Lee J A, Yow K C.Optimizing image matches via a verification model[J]. International Journal of Intelligent Systems, 2010, 25(11): 1103-1120.
[2] Fischler M A, Bolles R C.Random sample consensus: a paradigm for model fitting with applications to image analysis and automated cartography[J]. Communications of the ACM, 1981, 24(6): 381-395.
[3] Prasetyo H, Wiranto W, Winarno W.Image-based vehicle verification using steerable gaussian filter[J]. Journal of Telecommunication, Electronic and Computer Engineering, 2018, 10(2/3/4): 65-69.
[4] Zhang D, Oreifej O, Shah M. Face verification using boosted cross-image features [J]. Computer Science, 2013. https://arxiv.org/pdf/1309.7434v1.pdf.
[5] Wang F, Xiang X, Cheng J, et al.NormFace: L2 Hypersphere embedding for face verification[C]//ACM Conference on MultiMedia. 2017: 1041-1049
[6] 余洪. 基于计算机视觉技术的茶叶品质分级研究[D]. 南昌: 江西农业大学, 2017.
Yu Hong.Research on classification of tea quality based on computer vison technology [D]. Nanchang: Jiangxi Agricultural University, 2017.
[7] Wang J, Perez L. The effectiveness of data augmentation in image classification using deep learning [J]. Computer Science, 2017. https://arxiv.org/pdf/1712.04621v1.pdf.
[8] Sumit C, Raia H, Yann L C.Learning a similarity metric discriminatively, with application to face verification[C]//Proceedings of the 2005 IEEE Conference on Computer Vision and Pattern Recognition, 2005: 202-209.
[9] Melekhov I, Kannala J, Rahtu E.Siamese network features for image matching[C]//International Conference on Pattern Recognition, 2016: 378-383.
[10] Gregory K, Richard Z, Ruslan S.Siamese neural networks for one-shot image recognition[C]//International Conference on Machine Learning, 2015: 37-44.
[11] Simonyan K, Zisserman A. Very deep convolutional networks for large-scale image recognition [J]. Computer Science, 2014. https://arxiv.org/pdf/1409.1556v1.pdf.
[12] Szegedy C, Vanhoucke V, Ioffe S, et al. Rethinking the inception architecture for computer vision [J]. Computer Science, 2015. https://arxiv.org/pdf/1512.00567v1.pdf.
[13] Pan S J, Qiang Y.A survey on transfer learning[J]. IEEE Transactions on Knowledge & Data Engineering, 2010, 22(10): 1345-1359.
[14] 陆江峰, 单春芳. 基于数字图像的茶叶形状特征提取及不同茶叶鉴别研究[J]. 茶叶科学, 2010, 30(6): 453-457.
Lu J F, Shan C F.Shape extraction and varietial discrimination of tea based on digital image[J]. Journal of Tea Science, 2010, 30(6): 453-457.
浙公网安备 33019902000101号