This study established a near infrared reflectance spectroscopy models for exactly predicting the fat, protein and moisture of the ground and mince beef on line. Using our country' SupNIR-1000 near infrared spectrometer, the models were set up by artificial neural network (ANN). Related coefficient of calibration (rc) of fat model of mince was 0. 971 and related coefficient of prediction (rp) was 0.972. The protein' rc and RP were 0. 952 and 0. 949, respectively. The moisture' rc and rp were 0. 938 and 0. 927, respectively. Using ground beef established models, the fat' rc and rp were 0. 935 and 0. 810; the protein' rc and rp were 0. 954 and 0. 868; the moisture' rc and rp were 0.930 and 0. 913, respectively. So near infrared reflectance spectroscopy can better detect the fat, protein and moisture of mince than ground beef. But basically the ground beef model also can be used to quickly predict the chemical composition on line.%通过对整块牛肉和肉馅样品进行扫谱,测定其脂肪、蛋白和水分含量,采用国产 SupNIR-1000 近红外光谱仪,运用人工神经网络(ANN)分别建立肉馅和整块牛肉的脂肪、蛋白和水分的模型.肉馅样品的脂肪模型校正相关系数为0.971、预测相关系数为0.972;蛋白的校正相关系数为0.952、预测相关系数为0.949;水分的校正相关系数为0.938、预测相关系数为0.927.整块牛肉的脂肪模型校正相关系数为0.935,预测相关系数为0.810;蛋白的校正相关系数为0.954、预测相关系数为0.868;水分的校正相关系数为0.930、预测相关系数为0.913.比较可知近红外光谱能够更好的检测肉馅的脂肪、蛋白和水分含量.整块牛肉的模型也基本上可以满足牛肉化学成分的在线快速检测的要求.
展开▼
