基于支持向量机算法的乳制品分类识别技术研究

本文针对乳制品分类快速识别技术依旧相对匮乏的现状,获取了样品的拉曼光谱,以此作为表征样品的质量特性数据,输入支持向量机判别模型,构建高效识别技术。结果显示,乳制品拉曼光谱数据采集迅速,含水样品可直接上样测试,单个样品的数据采集时间仅需2.5min,计算机处理时间在10s以内,参数优化条件分别为小波软阈值降噪(db1小波基,分解层数N=3)、归一化处理([-1,1]区间),通过主成分分析提取80个主成分(累计贡献率99%以上),支持向量机算法(径向基核函数,惩罚系数c=32,核函数参数g=0.022097),测试集最佳识别率可达到100%。由此可见,本文所建立的高效识别方法,具有分析速度快、流程便捷等多项优点,能够为乳制品质量安全监管提供技术参考。

Research on Classification and Recognition of Dairy Products Based on Support Vector Machine Algorithm

In view of the current situation that the classification and rapid recognition technology of dairy products is still relatively scarce.The Raman spectra of the samples were obtained,which were employed as their quality characteristic data,then these data were input into the support vector machine model and an efficient recognition technology was constructed in this work.The results showed that the data collection process of Raman spectra of dairy products was rapid,the samples containing water could be tested directly,and the data acquisition time of a single sample was only 2.5 minutes.The processing time of computer was less than 10 seconds.The optimization conditions of the parameters were wavelet soft threshold denoising(db1 wavelet basis,decomposition layer N=3),normalization processing([-1,1]interval),principal component analysis extracting 80 principal components(cumulative contribution rate over 99%),support vector machine algorithm(radial basis kernel function,penalty coefficient c=32,kernel function parameter g=0.022097).The best recognition rate of test set could be achieved to 100%.Therefore,the efficient identification method established in this paper has many advantages,such as fast analysis speed,convenient process and so on,which provides a technical reference for dairy product quality and safety supervision.