Related Experiment Video
Updated: Dec 30, 2025
06:03
Integrating Augmented Reality Tools in Breast Cancer Related Lymphedema Prognostication and Diagnosis
Published on: February 6, 2020
7.0K
Real-Time 3D Reconstruction of Thin Surface Based on Laser Line Scanner
Yuan He1, Shunyi Zheng1, Fengbo Zhu1
1School of Remote Sensing and Information Engineering, Wuhan University, Wuhan 430079, China.
Sensors (Basel, Switzerland)
|January 23, 2020
Summary
This study enhances 3D reconstruction using truncated signed distance fields (TSDF) for thin industrial products. The improved method achieves high accuracy and real-time performance with reduced memory usage.
Area of Science:
- Computer Vision
- 3D Reconstruction
- Geometric Fusion
Background:
- Truncated Signed Distance Fields (TSDF) offer fast, accurate 3D reconstruction but struggle with thin surfaces, causing topological errors and high GPU memory demands.
- Existing TSDF methods face challenges in accurately reconstructing thin objects, leading to mesh collapse, overlaps, intersections, and gaps.
- High graphics processing unit (GPU) memory consumption in current TSDF approaches limits the scale of 3D reconstruction scenes.
Purpose of the Study:
- To improve the accuracy and efficiency of 3D reconstruction for thin industrial products using TSDF.
- To address topological errors and memory limitations inherent in existing TSDF methods.
- To develop a real-time, high-precision 3D reconstruction solution for thin surfaces.
Main Methods:
- Introduced a real-time thin surface attribution judgment method using the angle between surface normals and the line of sight to differentiate opposite sides of thin surfaces.
- Implemented a post-processing technique for automatic detection and repair of topological errors, particularly in areas with potential misjudgments.
- Developed a hybrid Central Processing Unit (CPU) and GPU framework to optimize performance and reduce GPU memory footprint.
Main Results:
- Achieved 3D reconstruction accuracy comparable to state-of-the-art laser line scanners, with an error of 0.02 mm for thin surfaces.
- Guaranteed real-time performance exceeding 60 frames per second (FPS).
- Reduced GPU memory usage to under 500 MB, enabling larger-scale reconstruction.
Conclusions:
- The proposed TSDF enhancements effectively overcome limitations in reconstructing thin surfaces, providing accurate and topologically sound 3D models.
- The integrated CPU-GPU framework ensures real-time processing while significantly lowering memory requirements.
- This method offers a robust solution for high-precision, real-time 3D reconstruction of thin industrial products.
