Smart Football Table - Detection
Summary
1) Detect the ball
- with shape search (OpenCV)
- with machine learning (OpenCV+YOLO)
2) Use ball positions to calculate different things
- goal detection
- ball velocity
- trace
Detection in detail
We are using OpenCV to detect the ball on gamefield. At the moment, there exist two solutions:
- single OpenCV ball detection, with image preprocessing and shape search:
- get input image (file or webcam)
- convert to hsv color range
- mask image to color range
- erode and dilate the picture
- find contours
- YOLO ball detection:
- get input image (file or webcam)
- find objects based on training data, which is the ball in this case
- for more, see README in yolo3 folder
Arguments for python scripts
| Argument | Description | Sample Input | default |
|---|---|---|---|
| -v | path to an (optional) video file, overwrites camera | “-v path/to/video.avi” | empty |
| -b | length of lightning trace | 64 | 200 |
| -i | index of camera | 0 | 0 |
| -c | color values (hsvmin,hsvmax) for object you want to detect (unneccessary for yolo) | 20,100,100,30,255,255 | 0,0,0,0,0,0 |
| -r | recording output into given file name | “fileName” | empty |