论文部分内容阅读
以叶绿素含量为评价菠菜新鲜度的参考指标,开发菠菜采后品质无损检测方法。采用计算机视觉和电子鼻分别获取储藏期内菠菜的图像和气味信息。分别提取视觉、嗅觉信息的主成分作为模型的输入,以叶绿素含量的化学检测值作为模型的输出,采用误差反向传播神经网络建立菠菜叶绿素的定量预测模型。试验显示,以视觉信息为输入量的模型测试结果:训练集和测试集的均方根误差(RMSE)分别为0.1978 mg/g和0.2147 mg/g,相关系数(R)分别为0.8457和0.7995。以电子鼻信息为输入量的模型测试结果:训练、测试集的RMSE分别为0.3119 mg/g和0.3032 mg/g,R分别为0.7013和0.6905。以视觉和嗅觉融合信息为输入量的模型测试结果:训练、测试集的RMSE分别为0.1759 mg/g和0.2121 mg/g,R分别为0.8888和0.8736,精度比两个单一技术均有所提高。研究表明,利用计算机视觉和电子鼻技术预测菠菜叶绿素含量的方法是可行的,采用融合技术有助于提升模型的预测精度。
Taking chlorophyll content as a reference index for evaluating the freshness of spinach, a nondestructive testing method for quality of spinach was developed. Computer vision and electronic nose were used to obtain the images and smells of spinach during storage. The main components of visual and olfactory information were extracted respectively as the input of the model. The chemical detection value of chlorophyll content was taken as the output of the model. The quantitative prediction model of spinach chlorophyll was established by error back propagation neural network. The results showed that the RMSEs of training set and test set were 0.1978 mg / g and 0.2147 mg / g respectively, and the correlation coefficients (R) were 0.8457 and 0.7995, respectively. The model test results using electronic nose information as inputs: the RMSE of training and test set were 0.3119 mg / g and 0.3032 mg / g, respectively, and R were 0.7013 and 0.6905, respectively. The model test results based on visual and olfactory fusion information: RMSE of training and test set were 0.1759 mg / g and 0.2121 mg / g respectively, and R were 0.8888 and 0.8736, respectively. The precision was higher than that of two single techniques. The research shows that it is feasible to predict the content of chlorophyll in spinach by using computer vision and electronic nose technology. The fusion technique can help improve the prediction accuracy of the model.