Study On Hyperspectral Determination Of Water Content Of Pork And Identification Method Of Injected Meat

Watered-down meat is considered to have serious risk for human health.Meat from large livestock such as pork,mutton,chicken and beef is easy to be made into watered-down meat,which is mainly made by watering the living animals at a certain time before slaughter or injecting water into the raw meat after slaughter.Generally,the color of the injected meat will be lighter than the normal meat.The surface of the meat will not feel sticky when touch it.After being placed in the air for a certain time period,there will be reddish blood flowing out.The social problems caused by watered-down meat include animal cruelty,violation of food safety laws and regulations,damage to consumer’s rights and interests,reduction of the taste quality of meat,and injection of unsanitary water.According to national standard "GB 18394-2001 livestock and poultry flesh moisture set limit to" the content of the water content>77%of the pork,beef,chicken in the setting of livestock and poultry flesh moisture limit standard,mutton moisture content>78%,can be judged as water injection meat,or water content exceeds bid.Meat consumption has always been the main part of resident’s consumption in China.In these meat food consumption,people consume most is pork.Therefore,to protect the interests of the people and the country,we need to control the safety of pork food testing strictly.In recent years,many researchers have adopted the high spectral imaging system in the nondestructive testing of agricultural and livestock meat and the testing and application of meat safety problems.The system is characterized by high resolution,no need to damage the collected samples,simple operation.However,to compose a specular like technology still needs more research and application,because of high spectral instrument in the data content of the cube,huge data information,containing a lot of duplicate content,to deal with the data of time will produce large amount of calculation,it will cost too much time,which makes the model slow.Therefore,different data processing methods adopted will have different model precision.Different types of meat used for modeling will also have different model accuracy.In given situation,this study was carried out on normal pork and injected pork meat through the design of the experiment and data analysis,and put forward the difference between them,and can adequately distinguish the two kinds of meat.Furthermore,then applied research was carried out on the identification of injected meat,and the specific research content included the following aspects:(1)The set highlights as technology in the spectral range of 358 nm to 1021 nm band to spectral information collection of 100 samples of pork meat,research methods in the pretreatment of raw spectral information,through comparing the results under different pretreatment of different modeling methods,the results show that the best 358 nm to 1021 nm band pretreatment method for the multiple scattering correction method.(2)Adopt the hyperspectral quantitative analysis technology to detect the moisture content of the whole meat quicjly,take the moisture content of the meat as the judgment standard of the injected meat,establish the detection model of the moisture content of the pork based on the hyperspectral technology,so as to identify the injected meat.Under this part separately established the whole band and characteristic band of multiple linear regression(SMLR),neural network(BP)and partial least squares(PLSR)model.The results showed that the optimal modeling method was the characteristic band of SMLR,the training set Rc2 is 0.7978,the training set RMESC root mean square error is 0.4021,the test set Rp2 is 0.8288,the test set root mean square error is 0.3412.Then using the characteristic band of SMLR,the recognition rate is up to 96.4%.(3)Adopt the hyperspectral qualitative analysis technology,according to the hyperspectral characteristics of the injected meat and the normal meat,and use the hyperspectral pattern recognition method to establish the discrimination mode of the injected meat,to realize the rapid identification of the injected meat.In this part,the support vector machine(SVM)and neural network(BP)classifier models are established respectively.The results show that the best classification method is the support vector machine model,and the recognition rate is as high as 95.2%.This project studied the determination method of pork meat moisture content based on hyperspectral technology,which can effectively protect the tested sample from damage and can carry out follow-up experiments,providing a theoretical basis for the design and development of portable pork waterflood meat identification instrument in the future.Therefore,research on pork meat quality hyperspectral NDT methods,can accelerate the rate of detection of meat,improve the quality of detection,can very good use detection model,then to promote pork high quality research and application of nondestructive testing technology,promote the nondestructive testing of meat high spectrum detection system development and the development of portable testing equipment,as well as to promote industrialization of the pork production process has important scientific significance and broad application prospects.