基于深度图像与三角测量的高精度物体表面重建方法
High-Precision Object Surface Reconstruction Method Based on Depth Images and Triangulation
针对传统线激光三角测量在高曲率、尖锐边缘区域易出现点云塌陷、边缘圆化等精度衰减问题,提出一种融合深度图像与微分几何约束的高精度表面重建方法。该方法先构建工业相机与线激光三角测量架构,获取初始点云并通过海森矩阵特征分解识别高曲率敏感区;建立光学传播与表面几何耦合的偏置补偿模型,生成边缘畸变偏置映射表;对敏感区采样点执行亚像素坐标修正与非线性回归调整;最后采用带特征权重的NURBS算法完成曲面拟合。实验表明,该方法可将高曲率边缘轮廓偏差由64.2μm降至8.4μm以下,角度重建误差小于0.11°,显著提升尖锐特征重建精度与鲁棒性,适用于精密制造在线检测场景。
To address the accuracy degradation issues such as point cloud collapse and edge rounding in traditional line laser triangulation for high-curvature and sharp edge regions, a high-precision surface reconstruction method integrating depth images and differential geometry constraints is proposed. This method first constructs an industrial camera and line laser triangulation setup to obtain the initial point cloud and identifies high-curvature sensitive areas via Hessian matrix eigenvalue decomposition; then establishes a bias compensation model coupling optical propagation with surface geometry to generate an edge distortion bias map; sub-pixel coordinate correction and nonlinear regression adjustment are performed for sampling points in sensitive areas; finally, a NURBS algorithm with feature weights is used to complete surface fitting. Experiments show that this method can reduce the contour deviation of high-curvature edges from 64.2 μm to below 8.4 μm, with angle reconstruction error less than 0.11°, significantly improving the reconstruction accuracy and robustness of sharp features, and is suitable for precision manufacturing online inspection scenarios.
| [1] |
王骏, 吴强, 杨瑞平, |
| [2] |
林原, 王萌. 机器视觉高精度测量技术在农业工程中的应用[J]. 南方农机, 2026, 57(01):136-139. |
| [3] |
徐彦中, 李涛. 计算机视觉测量技术在数智化工程监理中的应用[J]. 建设监理, 2025,(08):99-102.DOI:10.15968/j.cnki.jsjl.2025.08.007. |
| [4] |
张爽. 基于深度学习的三维点云补全与上采样算法研究[D]. 石家庄铁道大学, 2025.DOI:10.27334/d.cnki.gstdy.2025.000719. |
| [5] |
彭帝, 朱英硕, 夏晓华, |
| [6] |
徐胜军, 蒲宇彤, 吕红强, |
| [7] |
宋宪奎, 高春凤, 许京业. 机器视觉技术在粒状种子休止角测量中的应用[J]. 电子技术, 2025, 54(01):232-233. |
| [8] |
舒小华, 王珺, 闫松. 基于计算机视觉的结构位移测量技术及应用进展综述[J]. 火箭推进, 2024, 50(06):40-51. |
| [9] |
罗捷, 李远智, 黄宇飞, |
| [10] |
刘永强. 基于深度学习的三维点云配准算法研究[D]. 辽宁科技大学, 2024.DOI:10.26923/d.cnki.gasgc.2024.000626. |
| [11] |
杨帆. 基于NURBS的无序点云曲面重建研究与应用[D]. 电子科技大学, 2024.DOI:10.27005/d.cnki.gdzku.2024.001574. |
| [12] |
魏烨, 秦少谦, 吕盼稂. 时域视差降低激光三角测量系统中的边缘效应误差[J]. 工业控制计算机, 2023, 36(04):93-95+98. |
| [13] |
陈念年, 张佳成, 范勇, |
/
| 〈 |
|
〉 |