Abstract: To resolve the problem of improving the three-dimensional surface reconstruction accuracy and completeness for scenes with small structures,a three-dimensional surface reconstruction algorithm based on the fusion of scale and confidence was proposed.A surface implicit function was constructed with Gaussian function as its basis function and third-order Bessel function as its weight function.The coverage of the basis function and the value of the weight function were adjusted adaptively according to the scale and confidence of point clouds.The data was stored and processed in the octree structure,which was adaptively divided according to the scale of point clouds,realizing adaptive three-dimensional sampling of the surface implicit function.The three-dimensional surface was obtained as the isosurface of the implicit function through the unconstrained isosurface extraction method.Compared with several typical surface reconstruction algorithms,the experimental results showed that the small structures reconstructed by the proposed algorithm were clear and the overall reconstruction accuracy and completeness were improved.

Key words: three-dimensional reconstruction, octree, continuous implicit function, scale, confidence index

摘要： 针对包含细小结构场景的三维表面重建精度和完整度难以提高的问题,提出一种融合三维点云尺度与信心指数的三维表面重建算法。构建了一个表面隐函数,采用高斯函数为其基函数,三阶贝塞尔函数为其权重函数,并根据点云尺度与信心指数大小自适应调整基函数覆盖范围与权重函数取值。采用八叉树存储与处理数据,根据点云尺度大小自适应划分八叉树结构,从而自适应调整表面隐函数三维采样密度。通过无约束等值面提取方法,提取隐函数等值面,从而得到三维表面。与几种典型表面重建算法对比实验,结果表明所提算法重建三维表面细节清晰,整体重建精度与完整度均有提高。

关键词: 三维重建, 八叉树, 连续隐函数, 尺度, 信心指数