市場前線｜智慧製造中智慧機械的AI 視覺技術發展及趨勢分析： 從感知自動化到自主決策的演進

隨著全球製造業面臨供應鏈重組、勞動力短缺及少量多樣生產需求，人工智慧 (AI) 視覺技術已由傳統自動光 學檢測逐步發展為智慧製造體系中的核心感知技術。本文探討智慧機械AI 視覺之技術演進與產業發展趨勢，從演算 法、感測器與運算架構三大面向分析其由影像辨識、三維感知至自主決策之發展脈絡，並解析在線品檢、機械手臂 視覺引導、組裝驗證、物料追蹤、預測維護及安全監控等應用場景。進一步透過產業實務案例，探討邊緣 AI、生成 式 AI、基礎模型、多模態感知及數位雙生等技術之應用。未來，AI 視覺將朝向實體 AI、多感測器融合、生成式虛 擬場景及多種 AI 代理人協作發展，並成為智慧製造提升生產效率、品質與製程之關鍵技術。

As global manufacturing faces supply chain restructuring, labor shortages, and increasing demands for highmix, low-volume production, artificial intelligence (AI)-based vision technologies have evolved from conventional automated optical inspection toward a core perception technology for smart manufacturing. This article examines the technological evolution and development trends of AI vision for smart machinery, focusing on the synergistic advancement of algorithms, sensing hardware, and computing architectures. The evolution from image recognition and three-dimensional perception to autonomous decision-making is analyzed, together with key manufacturing applications including in-line quality inspection, robot guidance, assembly verification, material tracking, predictive maintenance, and safety monitoring. Industrial cases are further examined to illustrate the practical applications of edge AI, generative AI, foundation models, multimodal perception, and digital twins in manufacturing environments. The analysis indicates that AI vision technologies are increasingly evolving toward Physical AI, multisensor fusion, generative virtual environments, and multi-agent collaboration. This transition will enable smart machinery to progress from perception-based automation toward autonomous decision-making, while enhancing production efficiency, quality assurance, process adaptability, and manufacturing resilience. AI vision is therefore expected to become a critical enabling technology for the next generation of smart manufacturing systems.