Rapid Determination Of The Total Bacteria In Marine Fish By Portable Near Infrared Spectrometer

Microbial contamination is considered one of the most important hazards forfishery products. Traditional techniques for determination of total bacteria in foodare usually based on GB4789.2-2010. They are time-consuming which neededcultured (72±3) h, destructive in nature, and complicated operation which makesthem unsuitable for onsite monitoring of large numbers of samples. The currentlydeveloped detective technologies include impedance method, micro-biochemicalmethod and immunological method. Although the detection efficiency can beimproved, the pretreatment and cultured processes are still complex andcumbersome. They can hardly reach the on-site and on-line detection requirements.However, the near-infrared spectroscopy technology, aimed at food and otherbiological samples, can be qualitative or quantitative analysis rapidly,non-destructive and efficiently. In this research, a new strategy for determination oftotal bacteria in white fish (flounder fillets, perch fillets) and red fish (salmon fillets)is established using a portable near infrared spectrometer (FQA-NIR GUN). Incomparison to traditional exploited partial least squares regression (PLS), acombination of genetic algorithm (GA) and back-propagation artificial neuralnetwork (BP-ANN) to forecast the total bacteria of fish products. And actualsamples are measured. This research aims to develop a more rapid and accuratedetection of microorganisms which is important for the control of aquatic products’quality and safety. Main results are as follows:(1) In the progress on the application of near infrared spectroscopy in thedetermination of total bacteria and the pretreatment of the spectrum, the mainexperimental device was a portable NIR spectrometer. It connected with computerand two pieces of glass slides which close to the probe of NIR. Fish samples couldbe scanned against the slide. This portable assembly process was more appropriatefor onsite monitoring of total bacteria in fishery products. The spectrum whichscanned in fish should be analyzed by wavelet transform of Daubechies (db5). The second layer of detail coefficients was chosen. With such a pretreatment, thelow-frequency background could be effectively removed while most of usefulsignals could be preserved. It improved accuracy and precision of the analysis.(2) A new model for determination of total bacteria in cold flounder fillets wasestablished by rapid non-destructive analytical technique of near infraredspectroscopy analysis. PLS and combined GA and BP-ANN were used toestablished the model. R and RMSE were used to evaluate the model. The resultshowed that combining GA and BP-ANN was more appropriate to modeling thanPLS. It was validated in flounders from different time, places and freshness. Sothese results allowed us to suggest a new promising potential of the establishedtechnique for rapid (which just cost1min in whole testing process) and onsitemonitoring of total bacteria in fishery products.(3) The model of predicting the total bacteria of flounders was applied to otherwhite fish (Lateolabrax japonicus) and red fish (salmon). The result showed that it isfeasible through modeling, verification, and verification of other actual sampleperformance.Using a portable near infrared spectrometer, combined with neural networksystem modeling, a new technique for rapid and onsite monitoring of total bacteria insome fishery products was established. Compared with foreign papers of N.B. Tito etal, Yao-Ze Feng et al, Dimitris Alexandrakis, Yankun Peng and present progress, thismethod has an advantage on accuracy and precision. This on-site, rapid andnon-destructive method is meaningful for detection of freshness, food processing,trsportation and storage.