Computer Vision

Pupil Detection for Maculogix

Pupil finding is not a new problem, but hardware processor constraints, illumination imperfections, movement blur, lens fog, and even dark makeup can take a simple problem and make it quite complicated. Below is an example of an accurate detection (bottom right) after measuring and weighting a system of algorithms to determine the most likely location of the pupil.

This project was completed on embedded Linux using OpenCV with minimal processor speed, low battery usage, and a small training set.

“Vacuum Motor Inspection System” for Hoover Corporation

When vacuum cleaner motor wire is wound, errors occasionally occur.  This inspection system processes images from six separate cameras to determine if the part passes or fails, and integrates into the existing assembly line framework to send the parts to the correct path.



“Virtual Elevator Control Panel Prototype” for Otis Elevator Company.

This product was created in a Windows environment, with MS Visual Studio, and utilized off-the-shelf cameras, lasers, and optical filters to “see” and translate user inputs and generate output data for the existing elevator controller.


“Virtual Keyboard/Mouse Prototype” for Virtual Devices

EME developed a virtual keyboard prototype using a picoprojector, laser fitted with a line generator and a camera fitted with an optical filter.  In this video you can see the software reading the user inputs and translating them to operating system commands.

 Computer Vision