arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2405.05079cs.CV
收藏

面向无需初始化的大规模光束平差的幂次变量投影法

Power Variable Projection for Initialization-Free Large-Scale Bundle Adjustment

Munich Center for Machine Learning(慕尼黑机器学习中心);Technical University of Munich(慕尼黑工业大学);Hanyang University(汉阳大学)

Simon Weber, Je Hyeong Hong, Daniel Cremers

AI总结:

针对现有光束平差求解器依赖初始化的问题,提出幂次变量投影法,结合幂级数逆展开与黎曼流形优化,首次实现无需初始化的大规模可扩展光束平差,在BAL数据集上速度与精度达当前最优。

摘要:

大多数光束平差(Bundle Adjustment, BA)求解器(如Levenberg-Marquardt算法)都需要良好的初始化。与之相对,无需初始化的BA仍是一个基本未被探索的领域。研究尚不充分的变量投影算法(Variable Projection, VarPro)即使在无初始化的情况下也具有较宽的收敛域。结合物方空间误差公式,近期研究已证明其能够求解小规模无需初始化的光束平差问题。为使这类无需初始化的BA方法具备可扩展性,我们提出了幂次变量投影(Power Variable Projection, PoVar),对近期基于幂级数的逆展开方法进行了扩展。重要的是,我们将幂级数展开与黎曼流形优化关联了起来。这一投影框架对于求解无初始化的大规模光束平差问题至关重要。我们使用真实世界的BAL数据集开展实验,结果表明我们的求解器在速度和精度上均达到了当前最优水平。据我们所知,这项工作首次解决了无初始化BA的可扩展性问题,为无需初始化的运动恢复结构(structure-from-motion)开辟了新方向。

英文摘要:

Most Bundle Adjustment (BA) solvers like the Levenberg-Marquardt algorithm require a good initialization. Instead, initialization-free BA remains a largely uncharted territory. The under-explored Variable Projection algorithm (VarPro) exhibits a wide convergence basin even without initialization. Coupled with object space error formulation, recent works have shown its ability to solve small-scale initialization-free bundle adjustment problem. To make such initialization-free BA approaches scalable, we introduce Power Variable Projection (PoVar), extending a recent inverse expansion method based on power series. Importantly, we link the power series expansion to Riemannian manifold optimization. This projective framework is crucial to solve large-scale bundle adjustment problems without initialization. Using the real-world BAL dataset, we experimentally demonstrate that our solver achieves state-of-the-art results in terms of speed and accuracy. To our knowledge, this work is the first to address the scalability of BA without initialization opening new venues for initialization-free structure-from-motion.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑