Research On Rapid And Non-Invasive Determination For Ouality Of Fenneropenaeus Chinensis Based On Hyperspectral Imaging Technology

Prawn is one of the most important aquatic products in the world. There are variousnutrition contained in prawn, such as proteins, vitamin, iodine, zinc and astaxanthin, and therefore is welcomed by many people. With economic development and living standard’s improvement in China in recent years, there is a greater demand for consuming prawn in people" s daily life. It needs to product prawn in to many different kinds of products through different process and methods which include physical, chemical and microorganism process. All the product process affect the quality of prawn. What’s more some fraudulent phenomena have increased as well. In recent years, many incidents have been reported that prawns were injected with gelatin-like chemicals to increase weight in aquaculture markets. The gelatin was injected into the head and belly of the prawns to make them weigh more and look plumper when the frozen prawns thawed. It is hard for people to determin the qulity of prawn with naked eyes. Traditional determin methods are time-consuming, laborious, tedious and inconsistent. Electronical nose technology needs more reaction time and it cannot obtain image information. Near-Infrared sperctral(NIS) technology cannot get rid of the interference of background information and it can not gain spatial message. Based on this situation, this paper aims at developing a rapid and thoroughly determination method for quality of prawn using hyper spectral imaging technology systerm. The major research contents and conclusions are as follows:1. Based on Hyper Spectral Imaging technology obtain prawn’s different dehydration periods’ image, segment the image and extract the spectral data. Using LS-SVM model calibrate the VIS and NIR spectral data respectively, and use SPA, UVE and UVE-SPA select the optimal wavelength. Between LS-SVM、SPA-LS-SVM、 UVE-LS-SVM、UVE-SPA-LS-SVM model, UVE-SPA-LS-SVM model had the best performance with RPD>3.5, thus, it is more suitable for predict prawn’s moisture content.2. Studied the application prospect of Hyper Spectral Imaging technology in the determination of refrigeration of prawn. Established linear calibration algorithms named LS-SVM based on VIS and NIR spectral data respectively, Rp2>0.82. Using SNV pretreat the spectral data and SPA, UVE and UVE-SPA selected optimal wavelength and establish LS-SVM. We got the results that SNV-UVE-LS-SVM model had better performence for VIS and for NIR,SNV-UVE-SPA-LS-SVM model had better performence.3. Studied the application prospect of Hyper Spectral Imaging technology in the determination of freezing time of prawn. Established linear calibration algorithms named LS-SVM based on VIS and NIR spectral data respectively, Rp2>0.9. Using SNV pretreat the spectral data and SPA, UVE and UVE-SPA selected optimal wavelength and establish LS-SVM. We got the results that SNV-UVE-SPA-LS-SVM model had better performence for VIS and NIR. What’s more, SNV-UVE-SPA-LS-SVM model for NIR was better than that for VIS. Thus, NIR spectrum was more suitable for freezing time of prawn determination.4. Studied the reliability and accuracy of hyperspectral imaging technique with multivariate analysis was investigated for determination and authentication of gelatin inprawn. Hyperspectral images of prawns were acquired with different injection contents ofgelatin. The spectra of prawns were then extracted from hyperspectral images. LS-SVM was used to model the gelatin concentrations of prawn samples with their corresponding spectral data quantitatively. The combination of UVE and successive projections SPA was applied for the optimal wavelength selection. A total of thirteen wavelengths were selected as optimal ones from the whole spectral wavelengths462wavelength bands for gelatin determination. The UVE-SPA-LS-SVM model led to a coefficient of determination Rp2of0.965with root mean square error of prediction RMSEP of0.329, and was finally transferred to every pixel in the image for visualizing gelatin in all portions of the prawn. The results demonstrated that the combination of hyperspectral imaging, multivariate analysis and image processing has a great potential for determination and authentication of gelatin in prawn in a reasonable accuracy.5. The shape of prawn is not round, struments cannot measure the spectra of the whole prawn with-out containing the background information, therefore an online hyperspectral imaging system was developed to determine the moisture content.This reaserch studied the imaging automatic segmention. From segmented image extract prawn’s spectral data. The results suggested that hyperspectral imag-ing together with SPA held the advantages to be a fast, accurate, objective and non-invasive tool for the prediction of moisture content.Commparied the imaging automatic segmention technology with imaging mannual segmention technology. The imaging automatic segmention technology has not only better results but less physical work. Transferred the image of every pixel for visualizing moisture content in all portions of the prawn. At last, a map of moisture content distribution visualization of prawn obtained.