基于支持向量机原料乳细菌总数分类检测研究  

Research on classificiation detection of bacteria in raw milk based on support vector machine

作　　者：何桢[1] 高雪峰[1] 周延虎[1] 张炎[1] 

机构地区：[1]天津大学管理学院,天津300072

出　　处：《计算机工程与应用》2007年第22期186-190,213,共6页Computer Engineering and Applications

基　　金：国家自然科学基金(the National Natural Science Foundation of China under Grant No.70572044);国家教育部新世纪人才支持计划 (the New Century Excellent Talent Foundation from MOE of China under Grant No.NCET-04-0240)。

摘　　要：基于抽样技术研制了获取原料乳载玻片,建立了基于支持向量机(Support Vector Machine,SVM)分类技术的原料乳细菌总数分类检测系统,其中采用"一对一"及打分策略解决多类别的SVM分类问题,同时通过实验获取了合理的核函数及相关参数,使用测试集进行验证,结果表明,基于SVM算法的原料乳细菌总数分类检测系统分类的准确率可以高达96.42%。最后经过现场试验,结果表明该检测系统的一致性非常好,而且这种检测方法经济、操作简单,12分钟即可给出检测结果,比传统平皿菌落计数法(48h)有很大提高,能达到对原料乳按照细菌在线检测分级的目的,完全符合现代化乳品加工业的要求。Firstly,microscope glass slide in obtaining raw milk smear is designed based on sampling technology by ourselves,and then the classificiation detection system of total bacteria in raw milk based on SVM classification technology is established.The ＂one-against-one＂ approach and voting strategy are used to solve multi-class classification problems,reasonable kernel function and model parameters are obtained by a series of experiments,and the model is test by a testing set.Using this new method,the elassificiation deteetion system of total bacteria in raw milk based on SVM classification technology eomes up to accuracy of 96.42%.On the basis of practice,our experimental conclusion shows that not only detection system＇s agreement is very good,this method is easy to operate,it can give the result within twelve minutes,This method is faster than the plate counting （48 h）.But also it is economic,thus the method can greatly accelerate the judgment of the raw milk quality.It can meet completely the requirement of modernization dairy industry.

关 键 词：支持向量机 检测 原料乳 细菌总数 品质 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]