基于NACA和LS—SVM的蜂蜜真伪识别
2014-05-04梁秀英
梁秀英
摘要:针对目前蜂蜜掺假严重且传统检测方法耗时等问题,研究了基于近红外光谱
关键词:蜂蜜;近红外光谱(NIR);小生境蚁群算法(NACA法);偏最小二乘支持向量机(LS-SVM)
中图分类号:S896.1;TS207.3 文献标识码:A 文章编号:0439-8114(2014)02-0430-04
Identification of Honey Authenticity based on LS-SVM and NACA
LIANG Xiu-ying
(College of Engineering,Huazhong Agricultural University, Wuhan 430070, China)
Abstract: As honey has become the target of adulteration with cheaper sweeteners and the traditional methods for detecting honey adulteration are time-consuming and inaccurate. A new detection method of honey authenticity based on near infrared spectroscopy(NIR) was developed. The grid method and the ant colony algorithm(NACA) method were compared to optimize partial least squares support vector machine(LS-SVM) model parameters of γ and σ2 with RBF kernel. The results showed that two optimization methods could make the LS-SVM model to achieve 100% correct identify rate, but NACA can get the optimization solution in shorter time than that of the grid method. Using the standard normal variate calibration(SNV) as honey spectral data pretreatment and spectral data compressed by the partial least squares(PLS), the LS-SVM model is constructed to distinguish between unadulterated honey and adulterated honey samples in the range of 5 303~6 591 cm-1, 7 012~10 001 cm-1 and full spectrum range, respectively. The results showed that the 100 percent identification can be achieved in the range of 5 303~6 591 cm-1 and 7 012~10 001 cm-1.
Key words: honey; near infrared spectroscopy(NIR); ant colony algorithm(NACA); partial least squares support vector machine(LS-SVM)
蜂蜜是一种天然营养食品,含有丰富的糖类、蛋白质、氨基酸、多种维生素和矿物质等营养物质,经现代医学临床应用证明,服用蜂蜜可促进消化吸收,增进食欲,镇静安眠,提高机体的免疫力[1,2]。根据国际仪器法典、欧盟蜂蜜标准等国际标准规定,蜂蜜中不允许添加任何物质,也不允许从蜂蜜中去除某些成分[3]。然而,近年来由于经济利益的驱使,在蜂蜜的生产、加工、销售过程中,往往被加入其他价格低廉的甜味物质,如果糖、工业糖浆等,这些都制约了我国蜂蜜的出口,严重影响了国内蜂农和消费者的利益。因此研究快速、准确的方法来鉴别蜂蜜真伪,对中国蜂业的健康发展具有重要意义。
蜂蜜的主要成分为糖类,其中60%~80%是人体容易吸收的葡萄糖和果糖。在蜂蜜中掺入廉价的果葡糖浆是目前国内市场上最常见的蜂蜜掺假手段之一,尽管糖浆与蜂蜜中的葡萄糖、果糖成分类似,且当添加量控制在一定范围内时,难以用感官和常规化学方法检出,但在品质和营养价值方面与纯蜂蜜差异很大。
近红外光谱(NIR)分析技术具有快速、非破坏、无污染、可在线测量[3]等优点,已广泛应用于医药、食品、农业等领域。
蚁群算法又称蚂蚁算法,是Marco 1992年提出的模拟进化算法,是一种求解组合最优化问题的新型通用启发式方法,该方法具有正反馈、分布式计算和富于建设性的启发式搜索的特点,目前已被成功应用于人工智能、交通等领域。……
