Description
A real-time markerless subject tracking system for projecting 2D imagery into an augmented reality space is proposed using shape and color features to accomplish high computational efficiency and robust detection. The system is implemented on Android devices and is able to run consistently at a speed of around 30FPS. Results suggest that the algorithm is reasonably robust to variation in viewing distance and angle.
The system performed well in environments where there was a strong color contrast between the subject of interest and its immediate surrounding area. We were able to track the rectangular subject in various environments and successfully project a user image with correct perspective. Using a Samsung Galaxy S6 phone for testing, the Android application, developed using the OpenCV mobile library, was remarkably quick when running on x64 hardware and was able to maintain a frame rate of around 30FPS (frames per second) during all processing.
Select examples of system output in varying environments.
Robust mean-shift tracking with corrected background-weighted histogram
Reviews
There are no reviews yet.