Scholarlink

In industrial environments, detecting abnormalities in production equipment is a critical issue. While vibration-based measurement techniques are widely used, they typically require direct contact sensors, which poses certain limitations. Moreover, existing non-contact vibration measurement approaches, such as laser Doppler vibrometers (LDVs) or high-speed cameras, face challenges related to high costs and restrictive operational conditions. In this study, we address these challenges by utilizing an event camera for non-contact vibration measurement to enable effective abnormality detection. We conducted experiments involving multiple targets and evaluated abnormality detection using a simple threshold-based method. Additionally, we implemented time-series prediction with an LSTM in a sequence of operational steps to detect anomalies. Our results demonstrated the effectiveness of the proposed approach, indicating that non-contact abnormality detection is feasible using an event camera for vibration measurement.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Anomaly Detection Through Remote Vibration Measurement Using Neuromorphic Event-Based Camera

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Yosuke Ishikawa,
  • Tomomi Sato,
  • Kenji Iwata,
  • Yutaka Satoh

摘要

In industrial environments, detecting abnormalities in production equipment is a critical issue. While vibration-based measurement techniques are widely used, they typically require direct contact sensors, which poses certain limitations. Moreover, existing non-contact vibration measurement approaches, such as laser Doppler vibrometers (LDVs) or high-speed cameras, face challenges related to high costs and restrictive operational conditions. In this study, we address these challenges by utilizing an event camera for non-contact vibration measurement to enable effective abnormality detection. We conducted experiments involving multiple targets and evaluated abnormality detection using a simple threshold-based method. Additionally, we implemented time-series prediction with an LSTM in a sequence of operational steps to detect anomalies. Our results demonstrated the effectiveness of the proposed approach, indicating that non-contact abnormality detection is feasible using an event camera for vibration measurement.