Use of spectroscopic sensors in meat and livestock industries

Abstract

Near-infrared spectroscopy (NIRS), Raman spectroscopy, and hyperspectral (HS) sensors were trialled in the present thesis. The four experimental chapters were conducted to investigate potential applications in the meat industries. The first study compared a handheld NIRS sensor connected to a smartphone (NIRvascan) against a benchtop NIRS sensor to measure chemical composition (pH, moisture, fat, protein) of beef and lamb in three different sample presentations (fresh intact, freeze-dried, and freeze-dried-oven-dried). Both sensors showed similar, moderate precision in predicting fat concentration on processed meat (r2 = 0.78–0.81, RPD = 2.1–2.3). However, the predictions on fresh intact meat were insufficient for even rough screening in the industry (r2 < 0.67, RPD < 2.0). The second study compared the NIRvascan sensor against a Raman spectrometer to differentiate grass-fed from grain-fed retail beef cuts. The NIRvascan was more accurate than the Raman when scanning lean tissue, whereas the Raman was more accurate on fat tissue (both >90% accuracy). This study showed NIRvascan to be cheaper and more practical for industry and consumer use compared to larger and more expensive instruments. The third study used a multi-sensory platform containing two HS sensors (visible – VIS: 400–900 nm, and short-wave infrared – SWIR: 900–1700 nm) to classify beef and sheep organs by organ type (heart, kidney, liver, or lung). Hearts and livers were the more accurately identified (accuracy >90%) than kidneys and lungs (50–70%). The fourth study used the same multi-sensory platform to identify sheep organs rejected or fit for human consumption. Hearts and livers were more accurately (78–96%) discriminated as diseased or healthy compared to kidneys and lungs (60–91%). In the latter two chapters, the SWIR and VIS HS sensors achieved similar accuracy although VIS was slightly more accurate in differentiating organ type and SWIR was slightly more accurate in differentiating organs rejected or not for human consumption. In conclusion, three of the four chapters showed the potential of novel spectroscopic sensors for applications in the meat industries and encouraged further studies with larger sample sizes. Furthermore, the multi-sensory platform showed excellent potential as it is non-contact and can run at chain speed.