Real-time detection of intact pork freshness based on near infrared spectroscopy

Abstract. Online and real-time detection and analysis of intact pork freshness is an urgent need for meat enterprises. Near infrared (NIR) spectroscopy has revealed great potential in real-time detection for its fast detection speed, no need of sample disposal and likelihood of simultaneous prediction of multiple components. In this paper, a real-time detection system coupled with NIR spectroscopy whose wavelength were 350~2500 nm, light source, optical fiber, controller and tablet computer was built to detect pH and total volatile basic nitrogen (TVB-N) content of intact pork sample. The online and real-time detection parameters such as speed of conveyor belt and distance between optical fiber probe and pork sample surface were determined by orthogonal design experiments. Five levels were selected for speed of conveyor (200, 225, 250, 275 and 300 mm/s) and three distances were chosen (10, 11 and 12 cm). Reflection spectral curves of 20 samples in each case were collected and the accuracy and repeatability were compared to determine the optimal acquisition parameter. On this basis, 50 pork samples were employed for on-line spectra collection and reference values determination. Savitzky-Golay (S-G) filter, standard normal variate (SNV), first derivative and their combination were employed to eliminate the spectral noise and improve the signal to noise ratio. Then partial least square regression (PLS) was employed to build prediction models between the spectra and reference values. The results demonstrated that both speed of conveyor belt and distance between optical fiber probe and pork sample surface have effect on the spectral curve to a greater or less extent. And the prediction models based on the optimal detection parameters yielded good prediction for pH and TVB-N content and correlation coefficient in the prediction was 0.9055 and 0.9219, respectively. This work sufficiently demonstrated the proposed system could realize simultaneous real-time detection of pork freshness and may also be applied into the online prediction of color, tenderness and other quality parameters.