Comprehensive non-destructive assessment of beef quality: Focus on textural feature analysis and tenderness using AI and hyperspectral imaging

Abstract

This study developed a non-destructive method to predict beef tenderness using hyperspectral imaging (HSI), near-infrared (NIR) spectroscopy, and machine learning (ML). A total of 159 beef loin samples were analyzed to quantify moisture and collagen content using partial least squares regression ( R 2 = 0.973 and 0.953, respectively). Texture features were extracted from HSI-based chemical maps using gray-level co-occurrence matrix (GLCM) analysis. These features, combined with chemical indicators, were used to classify tenderness into three categories based on Warner-Bratzler Shear Force. The Random Forest model achieved the highest accuracy (F1-score = 1.00), followed by SVM (97 %). The proposed method offers accurate and efficient tenderness assessment suitable for large-scale industrial application. • The paper presents the Comprehensive AI (CAI) framework , designed to connect narrow AI, AGI, and human-like thinking through modular subsystems inspired by cognitive architectures. • Modular & Hierarchical : CAI is built like the human mind, with separate components for perception, reasoning, memory, learning, and self-monitoring, enabling it to handle a wide range of tasks. • Human-Inspired : It borrows ideas from neuroscience and cognitive science, incorporating abilities like attention, episodic memory, and self-reflection for more adaptable and context-aware AI. • Explainable & Ethical : CAI prioritizes transparency, trustworthy reasoning, and alignment with human values—key for safe AGI development. • Real-World Applications : This approach could improve AI safety and robustness in healthcare, education, robotics, and scientific research, boosting innovation while managing risks.