基于计算机视觉和人工神经网络预测挤压食品的质构特征

Prediction of texture characteristics in extrusion food based on computer vision and artificial neural networks

应用计算机视觉系统分别提取不同配方的挤压食品和同一样品不同部位的颜色值(HSI和L*、a*、b*),同时用质构分析仪测定样品质构特征。借助线性拟合模型通过样品的颜色对挤压食品的质构特征进行相关性分析,并利用BP神经网络模型通过颜色预测挤压食品的质构。线性拟合模型显示,硬度和胶粘度分别与a*值和对比度之间高度相关。两组实验中硬度与a*值之间的R2分别为0.9558、0.9429;胶粘度与对比度之间的R2分别为0.9741、0.9619。弹性与a*值和对比度之间具有一定的相关性,两组实验中弹性与a*值和对比度之间的R2分别为0.8675和0.8320。利用实验所得硬度、胶粘度、a*值以及对比度数据优化含有2个隐层的BP神经网络,得到两组实验对应最优网络模型结构,即每层所含神经元的数量分别为20、20,均方根(RMS,%)为4.25;20、40,均方根(RMS,%)为3.85。利用最优神经网络运用a*值和对比度对两组实验中的硬度和胶粘度进行模拟,得到的相关系数高于线性拟合模型拟合结果,两组实验中硬度与a*值之间的R2分别为0.9671、0.9770;胶粘度与对比度之间的R2分别为0.9766和0.9856。采用最优网络模型用颜色信息对挤压食品硬度和胶粘度的预测和验证结果表明,利用计算机视觉系统所提取的颜色值可以通过人工神经网络快速准确预测挤压食品的质构特征。

The surface color values（ HIS ＆ L＊, a＊, b＊ ）and texture characters of extrusion products were measured by using computer vision system and texture profile analysis.The correlation between color values and texture characteristics was analyzed by using linear fitting model as well as predicted texture characteristics from color value based on back-propagation neural networks.Results obtained from linear fitting model between the texture and surface colors values showed that the hardness and gumminess scores were indirectly reflected through a＊ value and Intensity due to their high correlation coefficients of 0.9558,0.9429,0.9741 and 0.9619, respectively.The coefficients（~） of springiness with a＊ value and Intensity were over 0.8675 and 0.8320, respectively.A back- propagation artificial neural network（BP-ANN） with 2 hidden layers,which derived from hardness, gumminess, a＊ and Intensity,was trained for simulating and predicting.The number of neural in each hidden layers were 20,20 （ RMS = 4.25% ） and 20,40 （ RMS = 3.85% ） in two experiments, respectively.A desirable and accurate prediction BP-ANN between hardness, a＊ and gumminess, Intensity were established with R2 = 0.9671,0.9770 and 0.9766, 0.9856 in two experiments.Results of this study indicated that texture characteristics of extrusion food could rapidly and accurately predicted through their surface color values by using computer vision system and artificial neural networks.