paper: Vision targeting with Contour Detection and Recognition

Thread created automatically to discuss a document in CD-Media.

Vision targeting with Contour Detection and Recognition
by: Jackflef

This paper was written about Team 3641’s vision targeting system from last year. It discusses how we detected and measured vision targets.

This white paper describes the uses of vision targeting for FRC games, as well as how vision targeting works.
In this paper, I will be using the example of programming a vision coprocessor (Like a Raspberry Pi) with Python and OpenCV installed. This is the system that the Flying Toasters used successfully in the 2017 season.

Robotics White Paper 2017.pdf (374 KB)

Great paper!! We will use it to help explain the topic to our programming team. All new/young programming team this year, The senior programmer graduated. :frowning:

This is a great paper. It explains very well what my team did last year. Obviously it’s different in quite a few ways as each team figures out their own way. I’d like to make a couple suggestions from my experience to new teams that look at this thread.

Something that I found is WAY more important than your thresholding is your camera settings. For instance if you have a bright light to reflect the tape you can drop your exposure which in turns lowers noise and keeps your image sharper.

I’d also like to just say that the official opencv tutorials are amazing. Some of them might go over your head the first few times you read them, but you can pick up a lot of tips and tricks. They have one about tracking the lanes in a car which is helpful because they explain only checking the parts of the picture you know the object is.