RoboCam Autofocus

Context

RoboCam was used for long-duration Drosophila imaging, but focus drift regularly compromised datasets during 24-hour runs. The system needed a reliable way to recover focus without constant manual intervention.

Contribution

• Built an autofocus workflow for a 3D-printer-based microscope used in reproductive-toxicity studies.
• Applied OpenCV, Gabor filters, and z-stack analysis to identify the sharpest focal plane automatically.
• Designed modular 3D-printed mounts for the camera and sample plate to keep optical alignment stable across runs.

Approach

The autofocus routine evaluated 5 to 20 images at each 0.1 mm z-step and selected the sharpest plane during setup and follow-up corrections. Initial focusing completed in about 30 seconds, while later corrections took about 15 seconds. I paired the software work with mechanical improvements so the imaging geometry stayed consistent during long recordings.

Result

The system captured continuous 1080p video at 24 fps and stabilized focus across more than 40 imaging runs.