DRIM: Depth Restoration with Interference Mitigation in Multiple LiDAR Depth Cameras
Seunghui Shin, Jaeyun Jang, Sundong Park, Hyoseok Hwang
AI summary
Problem
Simultaneous deployment of multiple LiDAR depth cameras causes mutual interference artifacts that degrade depth accuracy, and existing image restoration methods cannot distinguish these from natural sensor artifacts or run in real-time.
Approach
DRIM uses a shared encoder with a high-resolution attention path to classify depth data into background, sensor, and interference artifacts, then applies mask-guided correction to selectively restore only the interference-affected regions.
Key results
- First publicly available depth interference dataset with ground-truth artifact masks
- Superior depth restoration accuracy compared to existing image restoration baselines
- Real-time processing speed of approximately 33 FPS
- Effective restoration across diverse camera poses and mixed sensor environments
Why it matters
Enables reliable multi-camera LiDAR perception for robotics and autonomous navigation without requiring complex hardware synchronization or sacrificing real-time performance.
Abstract
LiDAR depth cameras are widely used for accurate depth measurement in various applications. However, when multi- ple cameras operate simultaneously, mutual interference causes ar- tifacts in the captured depth data, which existing image restoration methods struggle to handle. In this letter, we propose DRIM, a novel approach for real-time depth restoration under multi-device inter- ference. Our method begins by distinguishing interference-induced artifacts, then predicts and leverages these artifacts to guide the restorationprocess.Sincethereisnoexistingdatasetforlearningin- terference in multiple LiDAR depth cameras, we create and provide the first depth interference dataset. Our experiments demonstrate superior depth restoration performance compared to other image restoration methods, achieving real-time processing speeds (≈33 FPS) that are significantly faster than existing approaches while showing the capability to restore depth in challenging scenarios. These results demonstrate that our proposed method effectively restores interfered depth in multiple LiDAR depth cameras with practical real-time performance.