-
I've tested the software on the velodrome only a handful of times, but I haven't noticed it being particularly sensitive to different lighting conditions. Setting the motion detection threshold seems pretty easy. I haven't seen an abrupt change from constant random triggering to missing meaningful frames.
Quick changes (like a passing cloud) could probably trigger the motion detection accidentally, but that would be more of an inconvenience because the key frames are picked manually.
Too dark might increase noise and cause motion blur. I've used a simple method to reduce the effect of sensor noise, which looks to work pretty effectively. And if it's dark enough I guess using a torch to illuminate the passing bike could work.
I guess I could mention this here, because you'd need a Linux device to try it out (a laptop might be most suitable at the moment). I've made stopwatch software that uses a webcam, suitable for single person use on a velodrome or other similar tracks.
The program works by using motion detection to capture images of a rider passing the finish line, after the ride matching key frames are picked out by the user. The software calculates lap times and speeds as well as cumulative total for the whole ride. Rides are saved to the disk.
I've made the software mostly for my own personal amusement, so it well might not be suitable for anything resembling serious use, or a solution to a problem that doesn't exist. I did try it out on the local velodrome a couple of times last autumn, and I though it worked pretty nice.
A couple of caveats: timing comes strictly from image capture times so full laps only, and estimating standing start time can be a bit fiddly. I've also tested only a single webcam, there might be some compatibility issues.
You can get the source code from here. I'd be interested in hearing if anyone tries it out.
An image of what you'll end up with. The layout might be slightly lfgss inspired, sorry about that.