弗雷歇距离的释放:用青蛙近似一只狗
The Fréchet Distance Unleashed: Approximating a Dog with a Frog
AI总结:
本文提出了一种基于迪杰斯特拉/普里姆算法的弗雷歇距离近似方法,通过局部最优的变形实现高效计算,能更自然地处理噪声,提升实际应用价值。
摘要:
我们证明,一种连续弗雷歇距离的变体可以通过与解决离散版本所用的算法本质上相同的算法来计算。新的变体不一定是单调的,但这一缺点可以通过细化来轻松处理。结合迪杰斯特拉/普里姆类型的算法,这导致了一种弗雷歇距离(即变形)的实现,该变形是局部最优(即局部正确)的,既容易计算,而且在实践中对许多输入来说接近线性时间。新的变形具有 leash 始终尽可能短的性质。这些匹配/变形比标准算法计算的更自然,并且在处理噪声方面更优雅。这种方法应使弗雷歇距离在实际应用中更有用。我们实现了新的算法以及各种策略,以获得合理快速的实用性能。我们在新的算法上进行了广泛的实验,并发布了公开可用(且易于安装和使用)的Julia和Python包。我们的算法可以用来计算多边形曲线之间几乎精确的弗雷歇距离。实施和众多示例可在此处获得:https://frechet.xyz。然而,我们强调,现有的最先进算法/实现(C++)比我们的当前算法/实现快几个数量级。
英文摘要:
We show that a variant of the continuous Frechet distance between polygonal curves can be computed using essentially the same algorithm used to solve the discrete version. The new variant is not necessarily monotone, but this shortcoming can be easily handled via refinement. Combined with a Dijkstra/Prim type algorithm, this leads to a realization of the Frechet distance (i.e., a morphing) that is locally optimal (aka locally correct), that is both easy to compute, and in practice, takes near linear time on many inputs. The new morphing has the property that the leash is always as short as possible. These matchings/morphings are more natural and are better than the ones computed by standard algorithms -- in particular, they handle noise more graciously. This approach should make the Frechet distance more useful for real-world applications. We implemented the new algorithm and various strategies to obtain reasonably fast practical performance. We performed extensive experiments on our new algorithm, and released publicly available (and easily installable and usable) Julia and Python packages. Our algorithms can be used to compute the almost-exact Frechet distance between polygonal curves. Implementations and numerous examples are available here: https://frechet.xyz. We emphasize, however, that the existing state-of-the-art algorithm/implementation in C++ is faster, by several orders of magnitude, than our current algorithm/implementation.
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。