Hi! I wanted to know if there's any way by which we can detect the position of an object in an image using some programming language ?
For example : If i have an image of a ball that is updating itself say every 100 milliseconds,is it possible to get the coordinate of the ball through some program,using something like C++ or Java ?
Thanks!
If it is your own program, it may be possible. If it is someone else's (third party), it is still possible (but may be tougher).
This problem may actually have more to do with the platform environment (e.g. protection levels across processes) etc rather than the language.
Yes, it is.
One method that is often used when you want to detect some fixed predefined object in an image in reasonable time is boosting in combination with haar-like features. You can read up on it in the following papers:
An extended set of haar-like features for rapid object detection
Rapid object detection using a boosted cascade of simple features
It depends on the complexity of the scene. If the image contains one red ball, in front of a scene that contains no other red objects it is easy:
You select pixels with high values in the red channel, remove stray single pixels (e.g. with "opening"), and compute the center of gravity of the remaining pixels.
If the scene is only slightly more complex the necessary algorithm needs to be much more complex and you must start to read scientific papers, like the ones the TC mentioned. The already mentioned Hough Transform is a relatively simple algorithm that can recognize partially obscured shapes. Therefore, if you want to get away with a simple algorithm, set up the scene carefully and use good illumination with multiple lamps.
For libraries that do the basic image analysis operations, OpenCV and Camellia, that were mentioned already, are the kinds of libraries that you need. For simple algorithms and when you don't need real time, you could also also try the image analysis module from Scipy.