I’ll assume that you mean driver station vision feed and not automatic vision targeting.
As others have said, framerate and FPS don’t matter, the competition field will not support anything above what even the worst camera can handle. What does matter with your camera is the field-of-view (FOV). You can look around on amazon for cheap camera modules with high FOV lenses. Add the keyword “module” to get small, cheap, unpackaged boards. The important thing to look out for is that the camera supports Video4Linux2, often abbreviated V4L2. Any camera that does will work for FRC. Last year, my team wen’t for a 120deg FOV, which meant around 90deg horizontal FOV. Most of the top teams I’ve seen driving use high FOV cameras.
Other hardware can help reliability and FPS, but once you have basically any kind of coprocessor this is no longer a problem. The next big way to improve your video feed is the streaming method. That mostly means getting away from MJPEG-over-HTTP-over-TCP, and using something like H264-over-RTP-over-UDP. Some people will disagree that this helps, but in our experience it significantly reduced latency and increased on-field reliability. Basically, it allows a single dropped packet to not stall the network for a few hundred ms, and reduces the bytes transferred per frame. There’s been a bunch of posts about this kind of setup on chief already. Try google, or this or this for some good starting points.