Tracking objects outside the line of sight using 2D intensity images.

Area of Science:

• Computer Vision
• Optical Physics
• Robotics

Background:

• Observing objects in inaccessible or occluded regions presents significant challenges across various applications.
• Prior methods for non-line-of-sight object reconstruction often rely on complex and costly optical systems.
• Existing techniques for seeing around corners typically involve indirect diffuse light reflections.

Purpose of the Study:

• To demonstrate real-time tracking of occluded objects using a simplified setup.
• To introduce an analysis-by-synthesis approach for non-line-of-sight object tracking.
• To validate the method's capability with both known and unknown objects.

Main Methods:

• Utilized a standard 2D camera and a laser pointer for data acquisition.
• Employed an analysis-by-synthesis strategy involving repeated light transport simulations.
• Determined object parameters by matching simulated scene intensity distributions to measured data.

Main Results:

• Successfully tracked the translation of unknown objects in real time.
• Achieved real-time tracking of both translation and orientation for a known object.
• Experimental validation confirmed the efficacy of the simplified, simulation-based approach.

Conclusions:

• Real-time non-line-of-sight object tracking is feasible with basic equipment.
• The analysis-by-synthesis method offers a robust alternative to complex optical setups.
• This technique has potential applications in robotics, surveillance, and autonomous systems.