基于ELM算法的羊肉水分含量的快速无损检测

在近红外光谱900-1700nm的波长范围内采集南疆羊肉的光谱数据,来研究水分含量的快速无损检测.为减弱非目标因素对光谱的影响,采用SNV和去趋势法对光谱数据进行预处理.为降低建模的复杂度,去除共线性的影响,采用连续投影算法和相关系数法相结合选取8个特征波长变量,最后使用PLS和ELM算法分别进行建模.实验表明,与采用全光谱波段建模相比,采用特征波长变量建模,PLS和ELM算法的运行时间都大大缩短,并且在运行时间和预测精度上,ELM算法均优于PLS算法.ELM算法采用8个特征波段变量建模,预测精度达到0.9768,均方误差为4.4291e-04,相关系数为0.7603,运行时间可控制在1e-04s之下,这可为研发羊肉水分含量的便携式检测装置提供理论参考.

Fast Non-destructive Detection of Moisture Content in Mutton Based on ELM Algorithm

In order to achieve rapid non-destructive testing of moisture content,the spectral data of Southern Xinjiang mutton were collected within the wavelength range of 900-1700 nm in the near-infrared spectrum.To eliminate the effect of non-target factors on the spectrum,pretreat the spectral data by SNV and de-trend method In order to reduce the complexity of modeling and remove the influence of colinearity,eight characteristic wavelength variables were selected by using successive projection algorithm and correlation coefficient method.Finally,PLS and ELM algorithms were used for modeling respectively.Experiments show that compared with full-spectral band modeling,the operating time of PLS and ELM algorithms is greatly shortened,and the ELM algorithm is superior to the PLS algorithm in terms of running time and prediction accuracy.The ELM algorithm uses eight feature band variable modeling,with a predicted accuracy of 0.9768,an mean square error of 4.4291e-04,and a correlation coefficient of 0.7603.The operating time can be controlled under 1e-04 s,which can provide a theoretical reference for the development of portable testing device for the water content of mutton.