Detection And Contrastive Study Of Antioxidant Enzymes In Tan-sheep By Hyperspectral And Microscopic Hyperspectral Techniques

In this paper,hyperspectral and microscopic hyperspectral imaging techniques were used to detect the three chemical indexes of superoxide dismutase activity,peroxidase activity and soluble protein content in Tan-sheep muscle tissue.The chemical indexes were used to flise hyperspectral image and spectral information to establish the internal relationship.In the process of establishing the model,the abnormal values of muscle tissue samples were removed,the optimal pretreatment was selected,the effects of different modeling methods were compared,and the statistical analysis of characteristic wavelengths was carried out,and the visualization research was carried out.The main research contents and conclusions were as follows:(1)A micro-hyperspectral system was built and the suitable scanning parameters were determined as follows:?Microscopic Visible/Near Infrared(Micro-Vis 400)Hyperspectral:the scanning speed ahead was 0.1 cm/s,the retrograde speed was 0.2 cm/s,the scanning distance was 0.5 cm,the exposure time was 1.9 ms and the gain was 1.?Microscopic Near Infrared Hyperspectral(Micro-NIR 900):the scanning speed ahead was 0.0005 cm/s,the retrograde speed was 0.2 cm/s,the scanning distance was 0.03 cm,the exposure time was 15 ms,the frame speed was 40 Hz,the object distance was 300 mm,and the focal length was 30 mm.(2)The activity values of superoxide dismutase(SOD),peroxidase(POD)and soluble protein content were modeled by hyperspectral data analysis.According to the model,the hyperspectral modeling effects of the two enzymes were Micro-NIR 900>Micro-Vis 400>NIR 900>Vis 400.The optimum model of SOD activity was LSSVM model with 12 characteristic wavelengths extracted by SPA after SNV being pretreated,Rc was 0.8900,Rp was 0.8521.The optimal model of POD activity was LSSVM model with 14 wavelengths,RO was 0.8680,Rp was 0.8410,which was extracted by CARS after SG being pretreated.Four kinds of hyperspectral data of soluble protein content were modeled and analyzed,and the optimal spectral ranking was as follows:Vis 400>NIR 900>Micro-NIR 900>Micro-Vis 400.The optimal model was PLSR model with 13 characteristic wavelengths extracted by CARS after MSC being pretreated,Rc was 0.8702,and Rp was 0.8063.(3)The visualizations of SOD activity,POD activity and soluble protein content were realized by combining the three chemical indicators with hyperspectral data.The PLSR model based on the optimal pretreatment was established according to the chemical value index.The input variables of the spectrum were extracted by the method of P coefficient,and the visual imaging effect maps were obtained,which provided a reference for the further study of spectral technology in meat products detection.