Research On Portable Device For Detecting Multi-quality Parameters Of Shrimp And Models

Prawns are rich in nutrients.Due to their high water content,they are prone to spoilage and spoilage under the influence of enzymes and microorganisms,which adversely affects quality.Therefore,it is of innovative research value to find a fast,non-destructive,objective and accurate method for evaluating shrimp quality and to develop a portable non-destructive testing device for shrimp quality.This article uses Raman technology,based on embedded development board research and design portable detection device and explores traditional machine learning methods and deep learning methods to detect the quality of prawns.The main research contents are as follows:1?Research on portable Raman detection device.The embedded operating system is used to build a small-size,easy-to-carry Raman detection instrument that integrates a spectrometer,laser,Raman probe,Y-type optical fiber,and development board.The instrument designed and studied has the performance of quickly collecting sample Raman spectra.2?Establishment of a multi-index prediction model of shrimp quality.Use second-order differential,background subtraction,SG smoothing,standard normal variable transformation to process the spectrum,and PCA to reduce the dimensionality of the data,compare ridge regression,partial least squares regression,forward stepwise regression to construct shrimp quality multi-index prediction The prediction result of the model.The best model is forward stepwise regression model,and the best combination preprocessing method is second-order differential +background subtraction + SG smoothing + standard normal variable transformation.The results show that second-order differential + background subtraction + SG smoothing + standard normal variable transformation And SMLR to obtain the best prediction effect.The correlation coefficients of L,a *,b *,p H,and TVB-N in the prediction set are 0.863,0.850,0.859,0.900,and 0.916,respectively;the root mean square errors are 1.394,0.406,and 0.605,0.194,2.734.3?The establishment of a freshness discriminative model of shrimp based on deep learning.The accuracy of the constructed model is 100% for the correction set and 92.3% for the prediction set.The results show that the neural network model in the deep learning method can quickly identify and classify the quality of prawns.4?Reliability and accuracy verification of portable shrimp quality testing device.The physical and chemical index results of fresh shrimp samples obtained in the experiment showed that the deviations of L,a *,b *,p H and TVB-N were relatively low,which were 0.147,0.330,0.547,0.105 and 1.266.The predicted and physical and chemical values were compared.the device's prediction of shrimp quality indicators is relatively accurate.After 5s of the entire device,the human-machine interface displays the predicted values of different indicators and the Raman spectrum in real time,indicating that during the experiment,the detection time of the instrument is shorter,and the quality of the shrimp can be quickly detected.