Kalibracioni model za blisku infracrvenu spektroskopiju (NIRS) za procenu sastava stočnog graška (Pisum sativum L.)

Abstract

Bliska infracrvena spektroskopija (NIRS) se koristi za analizu hemijskih svojstava poljoprivrednih proizvoda bez prethodne pripreme uzorka. U ovom radu razvijen je kalibracioni model za merenje proteina i vlage za celo zrno i brašno stočnog graška. Dobijeni spektri obrađeni su kombinovanjem matematičkih modela (PLSR, MSC, prvog i drugog izvoda) i računate kalibracione greške. Greške pri merenju vlage celog zrna su u granicama (od -0,8 do 0,8) i brašna (od -0,7 do 0,8); pri merenju proteina celog zrna greške su (od -1,6 do 1,8) i brašna (od -2 do 2). Ovim ispitivanjem se dokazalo da je moguće odrediti količine proteina i vlage pomoću NIRS metode a dobijene greške se tolerišu i kalibracioni model je primenjiv u praksi.

Near infrared spectroscopy (NIRS) is used to analyze the chemical properties of agricultural products with no sample prepation. In this paper, a calibration model was developed to measure protein and moisture of forage pea in the whole seed and flour. The obtained spectra were processed by combining mathematical models (PLSR, MSC, first and second derivative) and calculating calibration errors. Measuring errors of the whole seed moisture are from (-0.8 to 0.8) and flour from (-0.7 to 0.8); measuring errors of the whole seed protein are from (-1.6 to 1.8) and flour from (-2 to 2). This test proves the possibility to determine protein and moisture via NIRS, errors are tolerated and the calibration model is applicable.