Development of beef freshness detection device based on air flow and multi-point laser technique

Abstract: The quality and texture of meat are closely related to the viscoelasticity changes during storage. The current viscoelastic model can also be widely applied to assess the physical properties and chemical components in the meat. Therefore, the viscoelasticity and meat freshness can be indirectly related, where the internal properties of meat caused by spoilage can be expressed by viscoelastic characteristics. It is very necessary to detect the viscoelastic information for a better prediction of freshness. However, the current laser technique has been limited in the food industry, due to the airflow diffusion. In this study, an innovative detection device of beef freshness was developed using airflow and multi-point laser technique. The hardware system of the device mainly included an airflow control, a displacement information acquisition, a stage lifting, and an air chamber. Some design strategies were selected to obtain a stable air flow, including the air chamber and the nozzle with the contraction curve, while the specific circuits to control the solenoid and electro-pneumatic proportional valve. Some key parameters were also selected for the displacement information acquisition and stage lifting module. The viscoelasticity of beef samples with different freshness was first represented by the displacement of the beef under airflow. Then, the data set of displacement was preprocessed via the Savitzky-Golay smooth (SG), the First Derivative processing (FD), and the FD+SG. After that, a prediction model was established using the Total Volatile Basic Nitrogen (TVB-N) content. Finally, a systematic evaluation was also made using the Partial Least Squares Regression (PLSR) and Principle Component Regression (PCR). The results showed that the preprocessing was greatly contributed to the accuracy of the model, where the accuracy of the PLSR model was much higher than that of the PCR model. The best PLSR prediction model was also achieved, when the viscoelasticity information was pretreated by FD+SG with the correlation coefficients in the calibration and validation set of 0.891 and 0.859, respectively, and the root mean squared errors in the calibration and prediction set of 1.071 and 1.337 mg/100 g, respectively. It indicated that the accuracy and stability of the model were improved significantly, compared with the traditional. Particularly, the multi-point laser technique was superior to the traditional single-point one. In addition, the control software of the device was designed to implement using the QT application development framework. Subsequently, the prediction model was implanted in the software to realize the one-click operation of the device. Furthermore, an external prediction test was performed on the 13 beef samples, in order to verify the stability of the device. It was found that the correlation coefficient between the prediction and measurement value was 0.887, where the root mean square error was 1.385 mg/100 g. Consequently, an excellent performance of the device was achieved for the non-destructive detection of beef freshness. Furthermore, the new technique can be widely expected to comprehensively represent the deformation of the sample in the future. The finding can also provide a strong reference for the freshness detection of meat products.