Real-Time Color Based Object Tracking

I have been trying to track moving objects for quite some time. I wasn't very successful in modifying my motion detection code to perform motion tracking because shadows were really making life difficult for my program. The program was having difficulties distinguishing the actual object from its shadow. So, I went for a different approach: "color based object tracking". Here's how it works. After the program captures an image from the webcam, it introduces it to a mean filter which causes the image to become blurred. This blurring is done to reduce noise from the image. After doing that, it uses an RGB filter to remove all objects from the image except the object to be tracked (I have to adjust the RGB filter manually for different objects). Then, to make life easier, the program thresholds the image. Thresholding turns the image into a 1-bit binary image. Check out the picture below:


Then, after performing some additional noise removal, it tries to determine the center of the object. See the picture below:


That's it! It's pretty simple, but I had to spend two days experimenting and testing this idea. Check out the video above to see this program tracking a yellow ball. I can't wait to hook this thing up with my floppy drive camera panner! It would be so much fun. :)