Inspiration

In manufacturing environments, camera-based vision systems are crucial for quality control. When cameras are bumped or lighting changes, detection can fail and a line can be stopped. Ford's MHacks track asked teams to address manufacturing quality control problems that reduce uptime. We built VisionAlign Tools so any floor worker can realign a camera system quickly, without specialist training, reducing downtime and rework.

What it does

VisionAlign is two complementary tools in one repo:

• Web-based operator tool (cli_main.py): browser UI, click-to-define target region, live MJPEG stream, visual guidance arrows, size/zoom advice, and a metrics panel. Designed for non-expert operators to align objects inside a bounding region.
• OpenCV frame-alignment runner (frame-alignment/alignment_runner.py): local, high-control tool intended for deployed cameras. Provides robust multi-method matching, image-quality metrics, and automatic exposure adjustment to maintain consistent detection under changing lighting. More useful for aligning a frame or camera that has been nudged.

Common capabilities:

• Click-to-define target region (web or OpenCV)
• Multi-method matching
• Visual guidance: arrows, distance indicators, status overlays
• Image-quality metrics and auto-exposure to compensate for lighting shifts
• Works with Basler Pylon cameras and standard webcams

How we built it

• Python + OpenCV for processing and visualization
• Flask for the web UI and streaming (cli_main.py)
• Basler pypylon SDK support for industrial cameras
• Modular auto exposure fitting and metric utilities in ae_hud.py

Challenges we faced

• Achieving real-time performance with multiple detection methods
• Creating an interface simple enough for any user
• Handling variable factory lighting conditions
• Balancing detection sensitivity vs false positives

Accomplishments

• Operator-friendly web tool for region-based alignment
• A dedicated runner with AE and HUD for deployed camera tuning
• Auto-exposure controller and image metrics that reduce detection drift
• Designed to directly address Ford’s manufacturing alignment & QC challenge

Built With

• basler
• flask
• opencv
• pypylon
• python