多源信息融合技术的猪肉新鲜度检测方法研究

Measurement of Pork Freshness Based on Multi-sensor Information Integration Technology

随着猪肉产量的提高和人们对食品安全的重视,快速准确地检测肉新鲜度越来越有着重要的现实意义。针对猪肉腐败过程中气味与颜色的变化,本文设计了猪肉图像采集装置和气体采集装置,对10个不同时间段采集到的各240份猪里脊肉样品图像信息和气体信息进行特征层的融合,基于最小二乘支持向量机方法(LS-SVM)建立多源信息融合的猪肉新鲜度评价模型,结合二步格点搜索法(Grid Search-ing Technique)和交叉验证方法(Cross Validation),对该模型参数γ和σ2进行选择和优化,分析比较了机器视觉系统、电子鼻系统及其两者融合系统所建立的猪肉新鲜度评价模型,3个模型对猪肉新鲜度的识别率分别为达到77.33%、91.67%和97.33%。结果表明,基于机器视觉和电子鼻多源信息融合系统可显著提高猪肉新鲜度的识别率。

It is important to develop a fast and precise method for detection of meat freshness with the annually increasing output of meat and concerns on food safety.In the light of color and odour change during meat taint,a meat freshness test system device for image acquisition and odour collection was designed,and the evaluation models were developed with 240 pork samples data fusion of color and odour characteristic parameters at feature level based on least-square support vector machine.To enhance the performance of least squares support vector machines（LS-SVM）,two parameters（γ and σ2） of the least squares support vector machine model were optimized by combination of two-step Grid Searching Technique and Cross-Validation.Then,the different models were established to assess pork freshness based on machine vision,electronic nose and combination of the two,giving recognition rates of 77.33%,91.67% and 97.33%,respectively.The results showed that multi-source information fusion system based on machine vision and electronic nose could significantly improve the recognition rate of pork freshness.