Potable Protein Detector And The Analysis Of Its Testing Precision

This paper is mainly using the regression method to establish regression function model between the spectrophotometry data of test card and protein concentration. One of the difficulties and key technology is the feature of election and extraction in modeling. First of all, pretreat noise reduction and anticipate hypothesis of absorbency curves and process data smoothly through average algorithm. Then using the combination selection method of inspiring method to optimize and extracting the characteristic variables, and finally got the optimized regression model. Using nonlinear method to linearize logarithm data and the log (c) model finally the RMSE of the model is reduced to around0.14. The relative error is within50%. Estimated modeling ranges from19to2500miu. The relationships of certainty dependency between protein concentration and absorbency data are significant. But the uncertainty of the data is also high. It suggests that we should increase the technology of protein card itself, reduce uncertainty further more and improve the precision of the relative quantitative detection.