Object Tracking with Computer Vision Using the Histogram Method!

This is a project that can track an object in a video. I made this using information from my CS 639: Intro to Computer Vision class's lectures.

Demo

Here's a demonstration video of it tracking a rolling soccer ball!

Here's another demonstration video where the program tracks a person that's walking around!

Here's how it works:

• Pick a video with an object that's at least a little distinct from its surroundings.
• Run the program.
• On the first frame of the video you can draw a square over the section containing the object.
• Then, a histogram with an adjustable bin size is created that represents the pixels contained within that square.
• For each frame of the video, sections are checked for similarity with the target square until the right one is found.
• A yellow box is drawn over this area, which should contain the object!

Future Improvements:

I would like to try a different method of comparing the histograms to each other to test for a performance improvement; Currently the method is using correlation, but another method like vector norms could potentially work better. If I had access to better computing resources or made a more efficient algorithm, then I could try on a higher-quality video with a higher framerate for smoother tracking box transitions.