Abstract
Depth sensors like LIDARs and Kinect use a fixed depth acquisition strategy that is independent of the scene of interest. Due to the low spatial and temporal resolution of these sensors, this strategy can undersample parts of the scene that are important (small or fast moving objects), or oversample areas that are not informative for the task at hand (a fixed planar wall). In this paper, we present an approach and system to dynamically and adaptively sample the depths of a scene using the principle of triangulation light curtains. The approach directly detects the presence or absence of objects at specified 3D lines. These 3D lines can be sampled sparsely, non-uniformly, or densely only at specified regions. The depth sampling can be varied in real-time, enabling quick object discovery or detailed exploration of areas of interest. These results are achieved using a novel prototype light curtain system that is based on a 2D rolling shutter camera with higher light efficiency, working range, and faster adaptation than previous work, making it useful broadly for autonomous navigation and exploration.
Introduction
Traditional depth sensors, such as LiDAR and RGB-D cameras, operate with a fixed acquisition strategy, scanning the entire field of view at a set resolution regardless of the scene's content. This leads to significant inefficiencies: undersampling small or fast-moving objects (like wires or projectiles) and oversampling uninformative areas (like flat walls). This paper introduces an "agile" sensor that dynamically adjusts its depth sampling based on the task, allowing for real-time "zooming" into regions of interest while ignoring empty space.
System Design and Methodology
The core of the proposed system is the Triangulation Light Curtain. A light curtain is formed by the intersection of a plane of light from a laser and an imaging plane from a camera. When an object occupies this specific 3D intersection, the camera detects the reflected laser light.
Key Innovations
- 2D Rolling Shutter Integration: Unlike previous 1D line sensor prototypes, this system utilizes a standard 2D CMOS rolling shutter camera. By synchronizing a steerable laser mirror with the camera's sequential row exposure, the system can sweep out complex, arbitrary 3D "ruled surfaces" at 60 fps.
- Ambient Subtraction: To ensure outdoor functionality, a column-interleaving technique captures laser-on and laser-off data within a single frame, allowing for robust detection even in 50klx of sunlight.
- Adaptive Sampling: The system can transition from sparse discovery curtains to dense, object-wrapped curtains in real-time to refine shape and track movement.
Experimental Evaluation
The prototype was evaluated through various real-world scenarios, demonstrating significant performance gains over traditional sensors:
- High-speed detection: The system successfully detected small balls (30-70mm) and frisbees at high velocity.
- Thin structure detection: It reconstructed thin wire fences that traditional LiDAR (e.g., Velodyne VLP-16) almost entirely missed.
- Mobile Mapping: In navigating cluttered industrial environments, the adaptive curtains covered relevant scene parts in less than 0.25 seconds.
The system achieves a 60 fps refresh rate (a 10x improvement over prior work) and a range of up to 20m outdoors and 50m indoors.
Discussion and Conclusion
The primary contribution of this work is the concept of Software-Defined Sensing. By exploiting rolling shutter characteristics—traditionally viewed as a distortion—the authors enable high-speed 3D triangulation with minimal post-processing. While limited by mechanical galvomirror inertia and occlusion, the system offers a powerful tool for active exploration in robotics. Future work includes multi-camera setups to mitigate occlusions and deeper integration with SLAM algorithms for autonomous navigation.
Summary updated