BeamProfiler Computer Vision System
At Excelitas Technologies, I designed a basic computer vision system to screen short-arc xenon light sources for quality and to improve optical coupling through apertures. I wrote a program in Python 2.7 using OpenCV and called the software BeamProfiler.
BeamProfiler analyzes light reflected from a surface to determine the beam characteristics of a light source. It streams live video from a mounted DSLR camera and reads the assembly orientation of a light source from a rotary encoder. The program initially searches for four blue squares in its video stream to define the extents of the imaging surface and to determine the distance from the screen. Once calibrated, BeamProfiler performs geometric corrections on an incoming video stream.
A light source is mounted in an assembly which allows it to be rotated along its axis of symmetry freely during operation. Once the light source is turned on, and the program begins to collect data. As a source is rotated 360°, BeamProfiler determines how well the light will couple into an aperture at each angle. The program then recommends the best orientation to maximize coupling efficiency at the end of the test.
The entire process takes 30 seconds per part and can increase the overall light output through an aperture by 50% in some cases.
