抄録
We develop an online real-time near-infrared (NIR) spectroscopic sensing system for the determination of progesterone concentration during milking. Enzyme immunosorbent assay was used as the reference (chemical) method. During milking, the milk spectra data of the range of 700–1050 nm wavelength were recorded at 1-nm interval every 20 s. We used partial least squares statistical technique to generate statistical models. The results obtained indicated very high precision and accuracy for the prediction of progesterone concentration. Consequently, this NIR spectroscopic sensing system installation into an automatic milking system could assess the concentration of milk progesterone for one milking with almost identical accuracy as the chemical assay result. Thus, this sensing system could monitor each cow`s estrus status and early pregnancy diagnosis.
