摘 要：舌象诊断是临床决策中非常重要的一个环节。研究人员提出了自动化的舌象诊断方法。他们通常从图片提取舌象，然后通过特征工程或深度学习方法，提取出相关的特征并分类，取得了不错的效果。然而，使用特征工程设计舌头特征需要很大工作量，另外仅使用手工特征或深度特征，无法较好地表示舌头的特征，特别是在舌头处于非统一光源和姿态下。因此，文章首先设计了基于Faster-RCNN 的检测框架对舌象进行预处理，然后使用了多特征融合的方法，对底层特征和高层语义特征进行特征融合，使用该方法来对舌象进行分类。结果表明，该算法具有更好的诊断效果。
关键词：舌头图像；检测框架；多特征融合；深度特征；小样本舌象
中图分类号：TP391.41 文献标识码：A 文章编号：2096-4706（2020）01-0063-04
Tongue Image Diagnosis Algorithm Based on Deep Learning and Multi-feature Fusion
QIU Tong
（Department of Computer Science and Technology，Tsinghua University，Beijing 100084，China）
Abstract：Tongue image diagnosis is a very important step in clinical decision-making. The researchers proposed an automated method of tongue image diagnosis. They usually extract tongue images from images，and then extract relevant features and classify them through feature engineering or deep learning methods，and achieve good results. However，it takes a lot of work to use feature engineering to design tongue features. In addition，only manual features or depth features are used to represent tongue features，especially when the tongue is in a non-uniform light source and posture. Therefore，this paper first designs a Faster-RCNN based detection framework to preprocess the tongue image，and then uses the method of multi feature fusion to fuse the features of the underlying features and the highlevel semantic features，using this method to classify the tongue image. The results show that the algorithm has better diagnosis effect.
Keywords：tongue image；detection frame；multi-feature fusion；depth feature；small sample tongue image
参考文献：
[1] KANAWONG R，OBAFEMI-AJAYI T，MA T，et al. Automated Tongue Feature Extraction for ZHENG Classification in Traditional Chinese Medicine [J].Evidence-Based Complementary and Alternative Medicine，2012.
[2] WANG X. Statistical Analysis of Tongue Images for Feature Extraction and Diagnostics [J].Image Processing，IEEE Transactions on Image Processing，2013，22（12）：5336-5347.
[3] ZHANG B，ZHANG H .Significant Geometry Features in Tongue Image Analysis [J/OL]. Evidence-Based Complementary and Alternative Medicine，2015：897580 （2015-07-13）.https://scite.ai/reports/10.1155/2015/897580.
[4] HUO C M，ZHENG H，SU H Y，et al. Tongue shape classification integrating image preprocessing and Convolution Neural Network [C]// Intelligent Robot Systems.IEEE，2017.
[5] HOU J，SU H Y，YAN B，et al. Classification of Tongue Color Based on CNN// [C].Proceedings of 2017 IEEE 2nd International Conference on Big Data Analysis（ICBDA 2017）.Institute of Electrical and Electronics Engineers，2017：745-749.
[6] MENG D，CAO G，DUAN Y，et al. A deep tongue image features analysis model for medical application [C]//2016 IEEE International Conference on Bioinformatics and Biomedicine （BIBM）.IEEE，2016.
[7] REN S，HE K，GIRSHICK R，et al. Faster R-CNN：Towards Real-Time Object Detection with Region Proposal Networks [J].IEEE Transactions on Pattern Analysis & Machine Intelligence，2015，39（6）：1137-1149.
[8] DALAL N，TRIGGS B. Histograms of Oriented Gradients for Human Detection [C]// 2005 IEEE Computer Society Conference on Computer Vision and Pattern Recognition （CVPR’05）.IEEE，2005.
[9] OJALA T，PIETIKAINEN M，MAENPAA T. Multiresolution gray-scale and rotation invariant texture classification with local binary patterns [J].IEEE Transactions on Pattern Analysis and Machine Intelligence，2002，24（7）：971-987.
[10] HE K，ZHANG X，REN S，et al. Deep Residual Learning for Image Recognition [C]//2016 IEEE Conference on Computer Vision and Pattern Recognition （CVPR）.IEEE Computer Society，2016.
作者简介：邱童（1994.11-），男，汉族，江苏徐州人，硕士在读，研究方向：数据分析。