基于图像识别的绳结形成规律的研究
展宝彤 李欣妍 谢逸恒 刘襄南 史旭光（通讯作者）

北京林业大学 理学院，北京，100083；

摘要：本文通过设计实验获得了不同材质、长度绳子所形成的纽结图片。应用多模态大模型和图像识别基础对绳结中交叉点的个数进行了统计。由统计结果发现不同长度的三种材质的线绳随着盒子绕转的圈数的增加，交叉点的数目随之增加。但当圈数很大时，交叉点的数目趋于一个稳定的数值。相较于传统的Opencv图形识别算法，本算法结合多模态大模型，具有轻量化，准确度较高等优点。

关键词：纽结；多模态大模型；二值化；图像识别

参考文献

[1]Kleckner,D.;Kauffman,L.& Irvine,W.How superfluid vortex knots untie.Nature Phys.,2016,12:650–655.

[2]Bisset,R. N.Wang,W.& Ticknor, C.; Robust vortex lines, vortex rings,and hopfions in three-dimensional Bose-Einstein condensates. Phys. Rev. A, 2015, 92: 063611.

[3]Wan,C.Shen,Y.& Chong,A.Scalar optical hopfionseLight,2022,2:22.

[4]Zhong,J.Wan,C.& Zhan,Q.Optical twisted phase strips.ACS Photonics,2023,10:3384–3389.

[5]Zhao,H.Tai,J.S.B.;Wu,J.S.& Li,Y.Liquid crystal defect structures with Möbius strip topology.Nat.Phys.2023,19:451–459.

[6]Bauer,T.Banzer,P.& Karimi,E.Observation of optical polarization Möbius strips.Science,2015,347:964–966.

[7]Kobayashi,M.& Nitta,M.;Toroidal domain walls as hopfions.2013,arXiv:1304.4737.

[8]Kobayashi,M.& Nitta,M.;Torus knots as Hopfions.Phys.Lett.B,2014,728:314–318.

[9]Yang, H. M.;Shi,X.G.;The Free Energy of Nucleosomal DNA Based on the Landau Model and Topology, Biomolecules,2023,13:1686.

[10]Wang,Y.& Shi,X.G.;Knot soliton in DNA and geometric structure of its free-energy density.J.Biol.Phys.2018,44:81–91.

[11]王正汉;结绳与量子计算.数学通报,2010,49(5):1–5.

[12]鄢盛丰;朱皓莹;朱文燕&刘建波;纽结理论在绳结模型上的应用与分析.中国运筹学会,第十三届中国不确定系统年会暨第九届中国智能计算大会论文集,2015:430–434.

[13]Raymer,D.M.& Smith,D.E.Spontaneous knotting of an agitated string.PNAS,2007,104(42):16432–16437.

[14]Alexander,K.Taylor,A. J.& Dennis, M. R.Knotting and weak knotting in confined, open random walks using virtual knots.J.Phys.A: Math.Theor.2017,50:405001.

[15]Zhang,Y.Li,X.& Wang,Z.A survey on multimodal large language models.2023,arXiv:2306.13549.

[16]刘一闻;李彤;王宇琦&张磊;开源多模态大语言模型初探.中国传媒科技,2024,31(8):82–85.