运用自行设计的敲击振动检测系统对西瓜进行敲击振动检测,通过不同激励-接收组合发现,当瓜脐处为测量点、瓜蒂处为激励点时,声学特征和西瓜内部糖度的相关性较高,且当声学特征为第二共振峰频率f2时,与样本糖度建立的三次拟合模型的R2为0.877.采用逐步多元线性回归分析建立西瓜糖度的检测模型,回归模型为:B =20.371-0.038f2 +0.035Vf1-0.006f4,模型的R2为0.881,通过校正集验证,预测样本平均误差为7.48％,检测模型的校正均方根误差RMSEC和预测均方根误差RMSEP分别是0.663和0.796.该模型的建立为在线无损检测西瓜的内部糖度提供了理论参考.%The research was conducted by knocking on watermelon for its' vibration test with self-designed vibration test system.Through different incentive-receive combinations,it found that when the melon pedicle was measurement point and the melon terrier was excitation point,the internal correlation of acoustic characteristics and sugar content was higher,and when the acoustic characteristics was the second formant frequency,the three fitting model of sugar content was 0.877.Stepwise multiple linear regression analysis was adopted to establish the watermelon sugar content testing model,regression model was B =20.371-0.038f2 + 0.035 Vf1-0.006f4,the R2of this model was 0.881.By calibrating validation,sample's average forecasting error was 7.48％,detection model correction RMSEC root mean square error and the root mean square prediction error RMSEP was 0.663and 0.796 respectively.The establishment of the model provides a theoretical reference for the on-line nondestructive testing of watermelons' sugar content.
展开▼
