Bachelor's Thesis Project
The aim of this project was to design and construct a welding helmet utilizing Augmented Reality (AR) technology in Video Pass-Through mode. The system eliminates the risk of flash blindness by replacing the traditional optical path with a digital vision system using a Global Shutter camera.
The device features a custom-made PCB, a hybrid power supply with Energy Harvesting (solar panel), and environmental sensors (gas, light). The software runs on a Raspberry Pi Zero 2 W, utilizing direct framebuffer access to achieve low-latency stereoscopic vision.
- SBC: Raspberry Pi Zero 2 W
- Camera: Raspberry Pi Global Shutter Camera (Sony IMX296) + ND 2-1000 Filter
- Display: 5.5" HDMI AMOLED (1920x1080)
- ADC: MCP3008 (SPI interface)
- Gas Sensor: MQ-07 (Carbon Monoxide)
- Light Sensor: Photoresistor (Trigger for AGC)
- Energy Harvesting: Polycrystalline Solar Panel
- Power: Li-Ion 18650 Cells + BMS
The electrical connection between the Raspberry Pi and the analog sensors via the MCP3008 ADC.
The motherboard was designed in Autodesk Fusion 360 and manufactured using a custom laser ablation method combined with chemical etching.
The control software is written in Python and optimized for the limited resources of the RPi Zero 2 W.
- Direct Framebuffer Access: Writing directly to
/dev/fb0to bypass X11 overhead and minimize latency. - Adaptive Exposure Control (AEC): Custom PID-like algorithm to adjust exposure time and gain in <100ms during arc ignition.
- Multithreading: Separated threads for image capture, data processing, and HUD rendering.
- Stereoscopy: Split-screen side-by-side rendering for VR optics compatibility.
opencv-python-headless
picamera2
spidev
numpy
The physical construction integrates standard welding protection with custom digital systems. The layout of the components is designed for optimal weight distribution and sensor accuracy.
| Front View (External Sensors) | Internal Electronics & Optics |
|---|---|
 |
 |
Front Panel (External):
- A) Camera Optical Assembly: Raspberry Pi Global Shutter camera with UV protection and adjustable ND 2-1000 filter.
- B) Solar Panel: Polycrystalline silicon cell for Energy Harvesting.
- C) Photoresistor: Light intensity sensor acting as a trigger for exposure control algorithms.
- D) MQ-07 Gas Sensor: Carbon Monoxide monitoring inlet.
Internal Layout:
- E) Central Processing Unit: Raspberry Pi Zero 2 W.
- F) Custom PCB: Motherboard with integrated ADC, manufactured via laser ablation.
- G) Power Management (BMS): Module for energy distribution and solar charging control.
- H) Energy Storage: High-capacity Li-Ion 18650 battery pack.
- I) VR Lens Assembly: Biconvex lenses for comfortable eye accommodation at close display proximity.
- J) 5.5" LCD Display: Full HD panel responsible for rendering the AR HUD interface.
Visualization of the real-time data overlaid on the welder's field of view.
| Arc Ignition (Initial Flash) | Stabilized View (Active AGC) | Gas Alarm (Safety Warning) |
|---|---|---|
 |
 |
 |
System operation in real-time. The GIF below demonstrates the Adaptive Gain Control (AGC) reacting to the welding arc.
Note: The high-quality raw video file is available in the repository at:
media/video/demo.mp4
Key metrics measured during validation tests:
| Metric | Result | Notes |
|---|---|---|
| Glass-to-Glass Latency | 60 - 90 ms | Dependent on exposure time. Acceptable for MIG/MAG welding. |
| Frame Rate | 18 - 20 FPS | Stable limit of RPi Zero 2W with dual-view rendering. |
| Solar Charging | ~150 mA | Measured under welding arc simulation. Acts as a Range Extender. |
| Gas Response Time | < 2.0 s | Tested with butane/propane mixture. |
The entire mechanical structure, including electronics enclosures and custom mounts, was designed in Autodesk Fusion 360. All components were optimized for 3D printing using PET-G and PLA filaments.
This project is open-source and available under the MIT License.
- Author: Jakub Antonowicz
- University: Silesian University of Technology (Politechnika Śląska)
- Year: 2026