多传感器信息融合在药材分类系统中的应用  

Multi-sensor Information Fusion Based Medicinal Material Classification System

作　　者：刘长江[1,2] 杨添钧 艾莉[3] 张轶[1] 杨丹[1] 

机构地区：[1]四川大学视觉合成图形图像技术国防重点学科实验室,四川成都610065 [2]四川理工学院理学院,四川自贡643000 [3]成都中医药大学药学院,四川成都611137

出　　处：《四川大学学报（工程科学版）》2013年第S1期100-105,共6页Journal of Sichuan University (Engineering Science Edition)

基　　金：四川省科技支撑计划资助项目(2008SZ0142);四川省科技创新苗子工程资助项目(2012ZZ026);四川理工学院科研项目(2012KY07)

摘　　要：针对传统药材分类技术中存在的速度慢、效率低、易受人为因素影响等缺陷,设计出一个基于多传感器信息融合的药材分类系统,完成药材成品的分类和评级。该系统利用图像传感器、气体传感器、味道传感器对药材特征进行量化,提取高维特征向量,通过支持向量机(SVM)训练学习,建立特征数据库,实现对药材的分类和评级。该药材分类系统集图像采集、特征提取、特征分析、模型训练和药材分类等功能于一体,实验结果表明,系统能够区分不同类别的药材,并能对同一类药材进行等级的确定。Traditional medicinal material classification methods has suffered from inefficiency and being susceptible to human factor.A multi-sensor information fusion classification and identification system was proposed.Firstly,this system quantized medicinal material features from image sensors,gas sensors and taste sensors.Secondly,high dimension features from both training and testing samples of medicinal materials were extracted.In the end,feature matrix was created and SVM algorithm was adopted for recognition purpose.Through experiment and visual measurement,medicinal material classification system was multi-functional one including image acquisition,feature extraction,feature analysis,model building and classification.It could not only classify medicinal material between different classes,but also identify grade in the same class.

关 键 词：多传感器信息融合 药材分类 特征提取 支持向量机 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]