top of page

3-D Spatial Mapping of Sparse featured environments

This project explores using RTAB-Map (Real-Time Appearance-Based Mapping) to create 3D indoor maps for applications like calculating paint/wallpaper requirements. The system employs a ZED Mini RGB-D camera and ROS 2 to perform SLAM (Simultaneous Localization and Mapping). Testing in Northeastern University’s tunnels revealed that mounting the camera on a rolling chair improved stability and accuracy compared to handheld use, though challenges like motion blur on carpets and algorithmic sensitivity persisted. Results showed precise alignment with ground truth maps in structured environments but misalignment in complex layouts. The project highlights RTAB-Map’s potential for cost-effective, real-time indoor mapping despite limitations in data reprocessing and environmental adaptability.

Vision

To revolutionize indoor spatial measurement by automating 3D mapping, eliminating costly human estimation errors, and providing a scalable solution for home improvement and infrastructure management.

The Prompt

High-quality paint and wallpaper are expensive, and manual estimation often leads to over-purchasing. Hiring professionals is costly, and existing tools lack precision. This project addressed these gaps by developing a robotic system to generate accurate 3D maps for precise material calculations.

Design/Methods

  • Hardware:

    • Stereo Labs ZED Mini: RGB-D camera with IMU and motion tracking.

    • Acer Nitro Laptop: Ubuntu 20.04, ROS 2 Humble, CUDA Toolkit for GPU acceleration.

    • Mounting: Rolling chair/desk for stability in smooth environments.

  • Software:

    • RTAB-Map: Graph-based SLAM with loop closure detection (Bayesian bag-of-words).

    • ROS 2: Data fusion, odometry, and real-time visualization via RVIZ 2.

    • ZED SDK: Camera calibration and depth streaming.

  • Methodology:

    • Front-End: Odometry estimation, graph construction (nodes/edges), and loop closure.

    • Back-End: Graph optimization (G2O, TORO) to refine 2D/3D maps.

    • Testing: Compared handheld vs. stabilized setups in tunnels (smooth floors) vs. classrooms (carpeted).

Problems Solved

  1. Costly Over-Purchasing: Automated 3D mapping reduces material waste.

  2. Odometric Drift: RTAB-Map’s loop closure minimized cumulative errors.

  3. Stability Issues: Chair-mounted camera improved accuracy in structured environments.

  4. Real-Time Visualization: Enabled users to monitor mapping progress dynamically.

Technical Specifications

  • Sensors:

    • ZED Mini (Depth Resolution: 1280x720 @ 30fps, FOV: 90°).

    • IMU (6-axis accelerometer/gyroscope).

  • Software Versions:

    • ROS 2 Humble, RTAB-Map 0.20.22, CUDA 11.7.

  • Algorithms:

    • Feature Matching: SIFT/SURF for loop closure.

    • Graph Optimizers: G2O (sparse graphs), TORO (large datasets).

  • Storage: ROSbags (~40x100 logs, 10–20GB per session).

Applications

  1. Home Improvement: Precise wall area calculations for paint/wallpaper.

  2. Infrastructure Mapping: Campus tunnel documentation, facility management.

  3. Robotics: Navigation for service robots in indoor environments.

  4. VR/AR: 3D environment reconstruction for simulations.

Performance Metrics

  • Accuracy:

    • Chair-Mounted: Sub-meter alignment with ground truth in tunnels.

    • Handheld: Drift up to 1.5m in non-linear corridors.

  • Speed: Optimal motion at <0.5 m/s to prevent tracking loss.

  • Stability: Chair setup reduced drift by 60% compared to handheld.

  • Limitations:

    • Carpeted Areas: Motion blur degraded map quality.

    • Reprocessing: ROSbag inconsistencies due to large file sizes.

Innovation and Future

  • Innovations:

    • Hybrid sensor fusion (RGB-D + IMU) for robust odometry.

    • Cost-effective hardware setup leveraging consumer-grade devices.

  • Future Work:

    • Sensor Fusion: Integrate LiDAR for texture-less environments.

    • Edge Computing: Deploy NVIDIA Jetson for portable processing.

    • AI Enhancements: Semantic segmentation for dynamic obstacle handling.

    • User Interface: Develop a mobile app for DIY users.

Conclusion

The project successfully demonstrated RTAB-Map’s capability to generate accurate 3D indoor maps in structured environments like tunnels, offering a cost-effective alternative to professional services. While challenges remain in uneven or dynamic spaces (e.g., carpets, cluttered rooms), the system’s modular design allows for future improvements. By addressing algorithmic sensitivity and storage constraints, this approach holds promise for broader applications in robotics, AR/VR, and smart infrastructure management.

Project Gallery

bottom of page