A novel quantitative analysis method for printability in 3D food printing for surimi

The success of 3D food printing (3DFP) relies heavily on material printability, defined by smooth extrusion and structural stability. However, many food materials struggle to meet these criteria, leading to challenges in shaping and printing precision. Conventional evaluation methods for evaluating key printability indicators like water holding capacity (WHC) and gel strength are often costly, require specialized equipment, and are destructive. To overcome these limitations, this study proposes a non-destructive, computer vision-based approach for printability evaluation using digital images. Grayscale images were analyzed to extract texture features based on the gray-level co-occurrence matrix (GLCM). These texture features, along with additive concentration data, then served as input for a Random Forest model. Three distinct models were developed: top view (TOP), side view (SIDE), and combined view (TS). The TOP model demonstrated the highest predictive performance for WHC (MAE = 0.674, R² = 0.875, RMSE = 0.831), while the TS model showed superior accuracy for gel strength (MAE = 0.391, R² = 0.691, RMSE = 0.556). This novel approach enables rapid, automated, and cost-effective assessments, thereby significantly aiding in 3DFP optimization and the development of customized food products.