Efficient non-line-of-sight tracking with computational neuromorphic imaging

Area of Science:

• Robotics and Computer Vision
• Neuromorphic Engineering
• Sensing and Imaging Technologies

Background:

• Non-line-of-sight (NLOS) sensing enables object detection around obstacles.
• NLOS tracking of moving objects faces challenges like signal redundancy and background noise.
• Existing methods struggle with efficiency and accuracy in complex NLOS scenarios.

Purpose of the Study:

• To demonstrate computational neuromorphic imaging using an event camera for robust NLOS tracking.
• To overcome limitations of traditional NLOS sensing techniques.
• To achieve efficient and accurate tracking of moving objects in NLOS conditions.

Main Methods:

• Utilized an event camera sensitive to luminance changes in dynamic speckle fields.
• Employed computational neuromorphic imaging for processing sensor data.
• Focused on capturing non-redundant information for direct motion estimation.

Main Results:

• The proposed method demonstrated superior efficiency and accuracy in NLOS object tracking.
• The event camera approach proved unaffected by the relay surface.
• Successfully obtained non-redundant information crucial for motion estimation.

Conclusions:

• Computational neuromorphic imaging with event cameras offers a powerful solution for NLOS tracking.
• The method significantly enhances tracking performance by focusing on relevant events.
• This technique holds promise for applications requiring robust object tracking in occluded environments.