智慧製造環境需要擷取大量的現場生產資訊，然而在真實的工廠中，仍有許多舊設備沒有具備數位化資料擷取和傳輸的介面，或是無法相容於標準化資料擷取和交換等問題，本研究因而設計一個嵌入式儀表辨識系統，用以自動擷取傳統儀表上的數字、指針與波形，將其轉換為數位資訊，從而解決設備資訊整合的問題。虛擬儀表系統分別對數字、指針和波形從事對應的影像處理和資訊擷取，同時以神經網路辨識數字資訊。我們使用一個基於ARM Cortex-M7核心的極小化硬體資源的嵌入式平台來進行實驗，其中數字、指針、波形辨識的正確性分別達到76.4%、73.68%和64.7%。驗證了此一嵌入式虛擬儀表系統的可用性。;To facilitate a smart manufacturing environment, considerable on-site production information must be extracted. However, in real factories, many old equipment and devices are not equipped with interfaces to extract and transmit digital data, or have compatibility problems against standardized data extraction and exchange. Therefore, this study designed an embedded instrument identification system to automatically extract digits, pointers, and waveforms on a conventional instrument and digitalize these data to solve the problem of equipment information integration. The virtual instrument system performs corresponding image processing and extracts data on digits, points and waveforms, and recognizes digits by neural network. We used an embedded platform based on ARM Cortex-M7 core with minimized hardware resources. The accuracy rates of the system in identifying digits, pointers, and waveforms reached 76.4%, 73.68%, and 64.7%, respectively. The results verified the feasibility of the proposed embedded virtual instrument system.