Comparative study between hyperspectral imaging spectrometer and a discrete multispectral detection device for assessing meat tenderness

Abstract. For rapid and accurate detection of meat tenderness, this paper attempted to measure meat tenderness and compared the results using two technical methods of hyperspectral imaging spectrometer and a discrete multispectral detection device. Firstly, preprocessing algorithms of Smoothing (SG) and multiplicative scatter correction (MSC) were implemented to carry out data preprocessing based on hyperspectral imaging data, then methods of Successive Projections Algorithm (SPA) and Step Wise Algorithm (SWA) were used to extract the characteristic bands in the spectral range of 400~1100nm. According to the extract characteristic bands, a discrete multispectral detection device was designed which contained a mini probe as the detector, a silicon photodiode detector as the signal acquiring component and Light emitting diodes as the light sources which matched the extract characteristic bands from hyperspectral imaging spectrometer. After constructing the discrete multispectral detection device, 59 pork meat samples with different tenderness were used to verify the performance of this device, Multiple Linear Regression (MLR) model using discrete multispectral detection device was established. The results indicated that the Multiple Linear Regression (MLR) model using discrete multispectral detection device was slightly below PLS model with hyperspectral imaging system, its correlation coefficient (Rc)and square error(SEC)of calibration set was 0.8327 and 5.91 respectively, and the correlation coefficient(Rp) and square error (SEP) of prediction set was 0.8040 and 6.15 respectively. This work demonstrates that the discrete multispectral detection method has the potential to predict meat tenderness and can simplify related instruments design structure and reduce their development cost in future.