高光谱成像技术检测冷却羊肉表面细菌总数

为了对冷却羊肉表面细菌总数进行无损检测,采用不同波段范围高光谱成像系统结合多种建模方法建立预测模型,进行理论分析和实验验证。分别在400nm~110nm和900nm~1700nm波长范围内获取冷却羊肉样本的高光谱图像信息,结合偏最小二乘和人工神经网络(反向人工神经网络和径向基人工神经网络)建立预测模型。结果表明,神经网络建模效果优于偏最小二乘;其中,径向基人工神经网络模型在400nm~1100nm和900nm~1700nm波长范围内相关系数分别为0.9872和0.9988,均方根误差分别为0.8210和0.2507,预测效果最好;而900nm~1700nm波长范围为最佳建模波长。这一结果说明利用高光谱图像技术对冷却羊肉表面细菌总数进行快速无损检测是可行的。

Measurement of total viable count on chilled mutton surface based on hyperspectral imaging technique

In order to obtain non-destructive assessment of total viable count (TVC) on chilled mutton surface, different kinds of recognition models were established based on different wavelength range hyperspectral imaging systems, then theoretical analysis and experiments were carried out. The hyperspectral imaging information of chilled mutton samples were collected in the region of 400nm~1100nm and 900nm~1700nm. The predictive models were established by partial least squares (PLS) and artificial neural network (back propagation artificial neural network and radial basis function artificial neural network (RBF-ANN)). The results show that the model which is on the basis of artificial neural network is better than PLS to predict TVC of chilled mutton surface. The best prediction result is based on the RBF-ANN model, the correlation coefficient and the root mean square error of prediction are 0.9872, 0.9988 and 0.8210, 0.2507 in the region of 400nm~1100nm and 900nm~1700nm. Meanwhile, the region of 900nm~1700nm is the best modeling wavelength. Therefore, hyperspecctral imaging technique can be used for the non-destructive detection of total viable count on chilled mutton surface.