Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Development of Ultra-Portable 3D Mapping Systems for Emergency Services

Charles Hamesse, Timothée Fréville, Juha Saarinen, Michiel Vlaminck, Hiep Luong, Rob Haelterman

TL;DR

The paper addresses the need for real-time, high-fidelity 3D mapping in hazardous emergency scenarios by developing and evaluating ultra-portable wearable systems. It implements LVIO- and ESIKF-based fusion across four configurations, ranging from helmet-mounted ToF cameras to dual body-worn LiDARs, and validates them through field-like trials. Key findings show that sensor choice critically affects map density, range, and robustness, with Azure Kinect offering improved geometry but form-factor and outdoor limitations, while dual non-rigid LiDAR setups provide broader coverage and better odometry at the cost of potential divergence in degenerate cases. The work highlights practical considerations for deployment, such as sensor placement and calibration, and points to future improvements in data fusion to enable reliable, real-time situational awareness for emergency responders.

Abstract

Miniaturization of cameras and LiDAR sensors has enabled the development of wearable 3D mapping systems for emergency responders. These systems have the potential to revolutionize response capabilities by providing real-time, high-fidelity maps of dynamic and hazardous environments. We present our recent efforts towards the development of such ultra-portable 3D mapping systems. We review four different sensor configurations, either helmet-mounted or body-worn, with two different mapping algorithms that were implemented and evaluated during field trials. The paper discusses the experimental results with the aim to stimulate further discussion within the portable 3D mapping research community.

Development of Ultra-Portable 3D Mapping Systems for Emergency Services

TL;DR

The paper addresses the need for real-time, high-fidelity 3D mapping in hazardous emergency scenarios by developing and evaluating ultra-portable wearable systems. It implements LVIO- and ESIKF-based fusion across four configurations, ranging from helmet-mounted ToF cameras to dual body-worn LiDARs, and validates them through field-like trials. Key findings show that sensor choice critically affects map density, range, and robustness, with Azure Kinect offering improved geometry but form-factor and outdoor limitations, while dual non-rigid LiDAR setups provide broader coverage and better odometry at the cost of potential divergence in degenerate cases. The work highlights practical considerations for deployment, such as sensor placement and calibration, and points to future improvements in data fusion to enable reliable, real-time situational awareness for emergency responders.

Abstract

Miniaturization of cameras and LiDAR sensors has enabled the development of wearable 3D mapping systems for emergency responders. These systems have the potential to revolutionize response capabilities by providing real-time, high-fidelity maps of dynamic and hazardous environments. We present our recent efforts towards the development of such ultra-portable 3D mapping systems. We review four different sensor configurations, either helmet-mounted or body-worn, with two different mapping algorithms that were implemented and evaluated during field trials. The paper discusses the experimental results with the aim to stimulate further discussion within the portable 3D mapping research community.
Paper Structure (11 sections, 5 figures)

This paper contains 11 sections, 5 figures.

Table of Contents

  1. INTRODUCTION
  2. RELATED WORK
  3. Portable sensors
  4. Algorithms
  5. PROPOSED SYSTEMS
  6. EXPERIMENTS
  7. LVIO with helmet-mounted Intel Realsense L515
  8. LVIO with helmet-mounted Azure Kinect DK
  9. LVIO with helmet-mounted LiDAR-visual-inertial system with Livox Mid-360
  10. Dual non-rigid LIO with body-worn sensors
  11. CONCLUSION

Figures (5)

  • Figure 1: Illustration of one of our mapping systems, showing example result point clouds and the prototypical hardware system.
  • Figure 2: Experiment A. Helmet-mounted Intel Realsense L515 ToF camera, including RGB camera and IMU (cable not shown). The action camera mounted on top of the helmet is not used for mapping.
  • Figure 3: Experiment B. Helmet-mounted Microsoft Azure Kinect ToF camera, including RGB camera and IMU, connected to a backpack PC.
  • Figure 4: Experiment C. Livox Mid-360 LiDAR-inertial device, with two Luxonis OAK-D Pro Wide cameras, facing backward and forward. The mapping test was realized in an office building. Colored point clouds use the RGB information of points captured by the LiDAR and visible by the cameras, which is why there are fewer points in the colored scans.
  • Figure 5: Experiment D: dual non-rigid LiDAR setup. LiDARs are denoted L1 and L2. The blue and red colors indicate which LiDAR has observed this point.