¹School of Science, Sichuan University of Science & Engineering, Zigong 643000, China
²School of Foreign Language, Sichuan University of Science & Engineering, Zigong 643000, China
³School of Computer Science, Sichuan University, Chengdu 610065, China

---

Abstract—A neural network model for classifying the Chinese herbal medicine based on image feature has been developed. The key process of the approach is to extract medicinal target region (region of interest, ROI). Two different auto-segmentation methods were adopted according to whether there were hollow in the medicinal materials. Subsequently, color, texture and shape feature of ROI was computed. Feature vector composed of forgoing features was input to back propagation neural network (BPN), with the class label as the output of neural network. BPN was trained mutually to classify the medicinal materials. The results of the experiment on 8 kinds of herbal medicine show the algorithms are valid with 98 high precision, with the structure of a 39-12-1 back propagation net.

Index Terms—Chinese herbal medicine (CHM), classification, back propagation neural network (BPN), region growing, image processing

[PDF]

Cite: Changjiang Liu, Xuling Wu, Wei Xiong, "Chinese Herbal Medicine Classification Based on BP Neural Network," Journal of Software vol. 9, no. 4, pp. 938-943, 2014.