基于差异化点云处理的双目结构光三维重建
Binocular Structured Light 3D Reconstruction Based on Differentiated Point Cloud Processing
DOI: 10.12677/app.2026.166059, PDF,   
作者: 赵 欣, 张 艺:天津工业大学物理科学与技术学院,天津;天津普达软件技术有限公司,天津;尚乾康, 张海明*:天津工业大学物理科学与技术学院,天津;段智魁:天津普达软件技术有限公司,天津
关键词: 太阳光干扰双目结构光三维重建点云处理小型工件Sunlight Interference Binocular Structured Light 3D Reconstruction Point Cloud Processing Small Workpieces
摘要: 针对自然光工业场景,太阳光干扰导致消费级双目结构光相机重建稳定性差、点云质量退化的问题,本文开展了面向小型工业工件的三维重建与优化研究。以奥比中光335L消费级双目结构光相机为硬件平台,搭建“硬件采集–软件处理–物体检测”三层架构实验系统。通过设置0.3 W/m2、22.5 W/m2、51.3 W/m2三种典型太阳光光照条件,对金属连接片、塑料定位盘、方形金属垫片三类工件开展原始点云采集与量化分析,明确了不同光照强度对点云的影响规律。在此基础上,提出“预处理–去噪–法线估计–连通域分割”全流程差异化参数点云处理方案,并结合泊松重建算法完成三维模型重构与精度验证。实验结果表明,优化后系统测量精度可达0.5 mm级别,低光照下点云噪声与边缘畸变得到抑制明显,中等光照下重建精度与稳定性进一步提升,强光环境下工件特征偏差降低约12%,可有效抑制太阳光干扰,满足中小企业低精度工业检测与自动化抓取的实际需求。
Abstract: Aiming at the problems of poor reconstruction stability and degraded point cloud quality of consumer-grade binocular structured-light cameras caused by solar light interference in natural light industrial scenarios, this paper conducts a research on 3D reconstruction and optimization for small industrial workpieces. Taking the Orbbec 335L consumer-grade binocular structured-light camera as the hardware platform, a three-layer experimental system including hardware acquisition, software processing and object detection is constructed. Under three typical solar irradiance conditions of 0.3 W/m2, 22.5 W/m2 and 51.3 W/m2, original point cloud acquisition and quantitative analysis are carried out for three types of workpieces, namely metal connecting sheets, plastic positioning discs and square metal gaskets, and the influence law of different light intensities on point clouds is clarified. On this basis, a full-process differentiated parameter point cloud processing scheme integrating preprocessing, denoising, normal estimation and connected domain segmentation is proposed, and Poisson reconstruction algorithm is adopted to complete 3D model reconstruction and accuracy verification. The experimental results show that the measurement accuracy of the optimized system reaches the 0.5 mm level. Point cloud noise and edge distortion are significantly suppressed under low light, and the reconstruction accuracy and stability are further improved under medium light. In strong light environments, the feature deviation of workpieces is reduced by approximately 12%. The proposed method can effectively suppress solar interference and meet the practical demands of small and medium-sized enterprises for low-precision industrial inspection and automatic grasping.
文章引用:赵欣, 尚乾康, 张艺, 段智魁, 张海明. 基于差异化点云处理的双目结构光三维重建[J]. 应用物理, 2026, 16(6): 647-656. https://doi.org/10.12677/app.2026.166059

参考文献

[1] Zhang, X., Wang, Z., Cai, Z., Liu, X. and Peng, X. (2024) Phase-Aided Online Self-Correction Method for High-Accuracy Three-Dimensional Measurement. Optics Express, 32, 2293-2305. [Google Scholar] [CrossRef] [PubMed]
[2] Liu, X.L., Yang, L.T., Chu, X.X. and Zhou, L. (2023) A Novel Phase Unwrapping Method for Binocular Structured Light 3D Reconstruction Based on Deep Learning. Optik, 279, Article 170727. [Google Scholar] [CrossRef
[3] 丁佰勇. 基于深度学习的RGB-D图像分析与重建模型研究与实现[D]: [硕士学位论文]. 北京: 北京服装学院, 2021.
[4] Zhao, S., Wu, J. and Zhang, H. (2025) A Binocular Structured Light Reconstruction Method Based on Disparity Fusion. Optics and Lasers in Engineering, 194, Article 109208. [Google Scholar] [CrossRef
[5] 陈登红, 庞宁, 邱宇, 等. 巷道形变轻量化点云扫描装备研制与应用分析[J]. 煤炭科学技术, 2025, 53(8): 141-155.
[6] 胡立楷, 黄云, 张坤球, 等. 基于RANSAC-LS的拱结构的三维激光点云线形重构技术[J/OL]. 建筑钢结构进展: 1-13. 2026-03-19.[CrossRef
[7] Hai, Z., Li, C., Guo, B., Wei, X., Yang, Z. and Zheng, J. (2026) Airborne Lidar Point Cloud Building Reconstruction Based on Planar Optimal Combination and Feature Line Constraints. ISPRS International Journal of Geo-Information, 15, Article 92. [Google Scholar] [CrossRef
[8] Huang, F., Shen, C., Fang, D. and Xiao, J. (2026) A Multi-Constraint Point Cloud Registration Method for Machining Error Measurement of Thin-Walled Parts. Applied Sciences, 16, Article 1003. [Google Scholar] [CrossRef
[9] Chu, H., Cheng, Y., Li, Y., Fan, J., Chen, N. and Wang, Y. (2025) Fine Extraction of Geometric Contour Feature Based on Spherical Thinning. Optics & Laser Technology, 192, Article 113784. [Google Scholar] [CrossRef