Projection Distortion-based Object Tracking in Shader Lamp Scenarios.

Area of Science:

• Computer Vision
• Augmented Reality
• Robotics

Background:

• Shader lamp systems enhance real environments by projecting virtual textures onto physical objects.
• Accurate object tracking is crucial for maintaining the illusion in dynamic augmented reality scenes.
• Traditional texture or contour-based trackers often fail due to interference from projected content.

Purpose of the Study:

• To develop a robust model-based tracking strategy for projector-camera systems that leverages projected content.
• To improve pose estimation accuracy in dynamic augmented reality environments.
• To enable real-time alignment and projection updates in closed-loop systems.

Main Methods:

• An iterative pose estimation algorithm analyzes visible distortions caused by object movement.
• Synthetic frames are rendered and compared to camera images using an optical flow-based edge matching method.
• Pixel-wise linear optimization implemented on GPUs facilitates efficient computation and avoids complex radiometric calibration.

Main Results:

• The proposed method achieves precise registration results in both real and synthetic image evaluations.
• It demonstrates robustness against distractions from projected content, outperforming traditional trackers.
• The approach allows for automatic threshold adaptation, simplifying the system setup.

Conclusions:

• The model-based tracking strategy effectively utilizes projected content for accurate pose estimation in projector-camera systems.
• This method enhances the stability and precision of augmented reality experiences in dynamic scenes.
• The technique is adaptable for other applications, including extrinsic pose estimation between cameras and projectors.