Scholarlink

Modern manufacturing environments require adaptable, low-cost vision systems, particularly for robotic end-effectors that must handle a diverse range of workpieces. This study evaluates the ESP32-S3 microcontroller integrated with the OV2640 camera as a compact, wireless vision module compatible with ROS2. The system transmits JPEG-compressed images over Wi-Fi using the micro-ROS communication stack. Performance was analyzed across various image resolutions, and the video stream was processed on a remote ROS2 host using OpenCV and YOLOv8n. To identify performance bottlenecks, high-resolution timing measurements were conducted along the data pipeline. Results show that image acquisition and message preparation contribute negligible delay (< 0.1 ms), while most latency arises during image encoding, serialization, and wireless transmission, with a median delay of approximately 78 ms. This limits the achievable frame rate between 18.3 and 40 frames per second. The findings demonstrate the ESP32-S3’s suitability for motion detection and basic object recognition in low-cost, real-time robotic applications.

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

Timing Analysis of Low-Cost Edge Vision for Object Detection for Smart Factory Systems

  • 会议论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年

查看全文
  • Christopher Tofaute,
  • Tadele Belay Tuli,
  • Florian Schreiber,
  • Martin Manns

摘要

Modern manufacturing environments require adaptable, low-cost vision systems, particularly for robotic end-effectors that must handle a diverse range of workpieces. This study evaluates the ESP32-S3 microcontroller integrated with the OV2640 camera as a compact, wireless vision module compatible with ROS2. The system transmits JPEG-compressed images over Wi-Fi using the micro-ROS communication stack. Performance was analyzed across various image resolutions, and the video stream was processed on a remote ROS2 host using OpenCV and YOLOv8n. To identify performance bottlenecks, high-resolution timing measurements were conducted along the data pipeline. Results show that image acquisition and message preparation contribute negligible delay (< 0.1 ms), while most latency arises during image encoding, serialization, and wireless transmission, with a median delay of approximately 78 ms. This limits the achievable frame rate between 18.3 and 40 frames per second. The findings demonstrate the ESP32-S3’s suitability for motion detection and basic object recognition in low-cost, real-time robotic applications.