Sign in Subscribe
Machine learning

Artificial Intelligence Researchers Develop Machine Vision System to Address Microcontroller Wastage

Strategic Insights

— 2 min read

Researchers at Ateneo de Manila University have developed a Machine Vision System (MVS) that can automate the inspection of microcontroller units (MCU) and detect empty trays, addressing a serious issue in the manufacturing environment. The MVS uses a Convolutional Neural Network (CNN) deployed in OpenMV H7 Plus, a low-power, memory-constrained Embedded Machine Learning System (EMLS). According to the study, the CNN achieves 94.8% training accuracy and 86.67% testing accuracy, classifying 128-pin, 64-pin, and 48-pin TQFP MCU IC packages, tray pockets, and background in real time.

Key Takeaways:

  • The mixing of MCU parts leads to wastage of materials, dissatisfied customers, and non-value-adding activities, resulting in negative feedback, loss of confidence, and impact on business growth.
  • The root cause of the issue is the final testing of the manufacturing back-end process when reusing unemptied standard JEDEC matrix trays for good and bad units in the test handler.
  • The proposed solution is developing a Machine Vision System (MVS) to automate MCU Integrated Circuit (IC) tray inspection and detect empty trays.
  • The MVS uses a Convolutional Neural Network (CNN) deployed in OpenMV H7 Plus, a low-power, memory-constrained Embedded Machine Learning System (EMLS).
  • The CNN achieves 94.8% training accuracy and 86.67% testing accuracy, classifying 128-pin, 64-pin, and 48-pin TQFP MCU IC packages, tray pockets, and background in real time.
  • The results show that an EMLS-based MVS can address MCU IC mixing issues, and future work should collect more data and explore additional feature extraction and data augmentation techniques to enhance CNN accuracy further.
  • The research proposes the development of an MVS as a feasible solution to automate MCU IC tray inspection and detect empty trays.

Statistics:

  • 94.8% training accuracy achieved by the Convolutional Neural Network (CNN)
  • 86.67% testing accuracy achieved by the CNN
  • 128-pin, 64-pin, and 48-pin TQFP MCU IC packages classified by the CNN
  • 7 (Proceedings on Engineering Sciences, 2025,7():1083-1090) research articles on the development of Machine Vision Systems for microcontroller inspection published in 2025
  • The research was conducted by Mark M. Pallones and King Harold A. Recto from Ateneo de Manila University

Sources:

  • Classification of Microcontroller Integrated Circuit on The Pocket of Jedec Tray Using Convolutional Neural Network in Embedded Machine Learning System. Proceedings on Engineering Sciences, 2025,7():1083-1090.
  • University of Kragujevac (Publisher for Proceedings on Engineering Sciences)
  • Journal of Engineering, July 14, 2025, p 167.
  • NewsRx LLC (Copyright 2025)
  • Ateneo de Manila University, Quezon City, Philippines.

Read more