Noninvasive sensing of thermal treatments of Japanese seafood products using imaging spectroscopy

The potential of imaging spectroscopy for noncontact sensing of thermal treatments experienced on Japanese kamaboko was investigated. Samples were thermally treated at 100, 120, 140, 160, 180 and 200 °C to core temperatures of 45, 50, 55, 60, 65, 70, 75 and 80 °C and then promptly cooled and imaged in the short‐wave near infrared spectral range of 900–2500 nm. Partial least square (PLS) regression models were developed using the whole spectral range as well as using the most important wavelengths to predict the core temperature (T_C) and thermal history (TH) yielding a reasonable level of accuracy of ( = 0.86 and RMSEP = 3.9 °C) and ( = 0.83 and RMSEP = 0.29 min), respectively. Moreover, a stepwise linear discriminant analysis (LDA) model was developed for identifying samples whose core temperatures reached a threshold of 65 °C. The LDA model yielded overall classification accuracy of 93.75% in both calibration and validation sets. The resulting discrimination function was then applied in a pixel‐wise manner to produce understandable classification maps to exhibit the difference among samples with high accuracy.