The title of my research project is "Design of an image processing algorithm for detecting a soccer ball".
The ultimate aim is to simulate an outdoor robot soccer game using the
iRobot Create's, Asus Eee PC's and a soccer ball.
My part of the project deals with Computer Vision.
The skills that I use are C++; a working knowledge of ROS (Robot Operating System)and OpenCV
(Open Source Computer Vision); and the will to learn more about robotics and computer vision.
Being my first research experience in this field, I have achieved some things.
I have written a program to save images of a ball (as seen by the robot) while the robot is driving towards the ball.
Another program lets a user to "manually" locate the ball in these images and record, in a text file,
the position of ball’s center and its radius in the image.
This database of images and ball information will be used in training the robot.
My third program was used to find a ball in an image using Hough Transform to find circles.
Obtaining many false positives, a criterion was added to check for shadows underneath the ball.
This method worked fine for some images in the database but not for others.
Future work will utilize other methods of object recognition. Also, using machine learning,
the robot can be trained on how to recognize a ball in its view using the image database.
The project was a good way for me to get my hands on Computer Vision and Robotics.
|
|