Using OpenCV and two cameras to detect 3D position of object.
Requirements: OpenCV (http://opencv.org/)
Build & Run:
mkdir build
cd build
cmake .. && make
./twocams
Required configurations files:
config.xml: Define camera indexes, calibration files and parameters for automatic goal detection (like goal size and offsets)- Intrinsic calibration files: One file per camera required, can be created using the program provided by this OpenCV-Tutorial: http://docs.opencv.org/master/d4/d94/tutorial_camera_calibration.html
How to use:
- Start program
- Place chessboard as your "goal" (both cameras must see it completely)
- Press SPACE
- If chessboard has not been detected properly there will be an error message and you have to try it a again (go to 2.)
- Extrinsic calibration has been sucessfully done, projection matrices of cameras have been calculated
Now you have several options:
- Mark a position in both views (by clickin in them)
- Press SPACE to get 3D position of this point
OR
- Press 'g' to start goal detection mode
- Now the ball will be detected automatically and marked in the views by a red circle
- When it crosses the Z-coordinate of the goal line (see
config.xml) a goal or a miss will be detected (depending of the X and Y position of the ball)
OR
- Press 't' to start deviation test
- Chessboard corners will be detected automatically, its 3D points will be calculated and compared to the actual positions of this edges
- Output is the deviation in cm for each point and in average