论文部分内容阅读
目的通过对近红外广谱数据进行神经网络系统训练,讨论近红外广谱技术对冷藏三文鱼菌落总数快速预测的可行性。方法针对三文鱼在4℃贮藏过程中的微生物变化,利用手持式近红外光谱仪,通过小波分析对于光谱进行预处理,之后结合遗传算法和BP神经网络系统方法建立预测和检测模型。结果该模型与传统平板计数方法的相关系数为0.981,均方根误差为0.097,验证模型的相关系数为0.960,均方根误差为0.098,具有良好精确度、准确度。结论该方法能够用于冷藏三文鱼菌落总数的无损、现场检测。
Objective To investigate the feasibility of rapid prediction of the total number of frozen salmon colonies by near infrared spectroscopy (NIRS) through neural network training of NIR data. Methods The changes of salmon during storage at 4 ℃ were studied. The spectrum was preprocessed by hand-held near-infrared spectrometer (WNIR) and analyzed by wavelet analysis. Then the prediction and detection models were established by genetic algorithm and BP neural network system. Results The correlation coefficient between the model and the traditional plate counting method was 0.981, the root mean square error was 0.097, the correlation coefficient of the validation model was 0.960, and the root mean square error was 0.098 with good accuracy and accuracy. Conclusion This method can be used for non-destructive on-site detection of the total number of salmon chilled colonies.