[John] is keeping the neighborhood safe by keeping an eye out for speeders. Well, he’s really keeping a webcam out for speeders. His technique doesn’t use radar or lasers. He’s processing webcam frames in Python to calculate speed.
It comes down to some basic image manipulation. He firsts gathers the images necessary to make the calculations by using a motion-detecting webcam program called YawCam. The images are analyzed to establish which parts have changed between frames; this gets rid of all the stationary objects. Now the frames can be compared to establish the distance in pixels. By calibrating the shot through measurements of the target area, this data can be directly converted into actual distance. It is then compared with the timestamps from each frame to arrive at speed. This can be used for vehicles on the street like we see above, or more whimsical measurements like pet turtle progress.