[Objective/Significance] With the rapid development of artificial intelligence, computer vision is one of the important branches of artificial intelligence, and its new hot technical topics are constantly emerging. By analyzing the evolution of cutting-edge hot topics in the field of computer vision, we can gain an in-depth understanding of the hot technologies, development directions and evolutionary paths in the field of computer vision, and provide guidance and reference for related research and practice. [Method/Process] Based on computer vision-related patent data in the Derwent patent database, the BERTopic model was used to extract topics, and the BERT-based DTM model was used to further analyze the evolution of the topics, and summarize the research frontiers and hot topics of computer vision in recent years. [Results/Conclusions] Hot research topics in the field of computer vision are mainly concentrated in the fields of autonomous driving, multi-modal artificial intelligence, human-computer interaction, and three-dimensional modeling. Computer vision as a whole shows huge development space and potential.
Wang Lanlan, Liu Yanli.
Hot Technology Topics and Evolution Analysis in the Field of Computer Vision Based on BERTopic[J].
Science Focus, 2024, 19(2): 46-60 DOI:10.15978/j.cnki.1673-5668.202402005
本研究选取的数据类型为专利数据,数据来源为德温特专利数据库(DII),德温特专利数据库覆盖涉及近50个国家/地区的、各种语言公布的专利(如德国、法国、日本、西班牙、葡萄牙等非英语国家专利),经过阅读后统一用英语编辑改写成DWPI专利数据。该数据库的优势在于,具有专利人员对原始专利信息进行二次改写的、独特的DWPI专利摘要。选择DII专利数据库的DI专利分析平台进行专利数据的收集,数据采集年份为2010-2022年。对计算机视觉相关内容进行调研,并对DI专利分析平台的检索策略进行适配。为检索出最为贴近的计算机视觉领域专利,经过多次检索调整,缩减调整检索式为:(CTB=(COMPUTER ADJ VIS*)) AND AY>=(2010) AND AY<=(2022);对检索得到的数据进行合并专利同族等预处理,获得数据15823条,去除空白和字段严重不全的数据,最终获得15651条数据。
(SongK, ZhuY J. Patent frontier technology topic identification and trend prediction: a case analysis of artificial intelligence[J]. Journal of Intelligence, 2021, 40(1): 33-38.)
(ChenL, LinP, DuanY Q. Technology topic mining and trend analysis from the perspective of industrial chain combined with K-means and LDA—taking virtual reality technology as an example[J]. Knowledge Management Forum, 2020, 5(3): 135-146.)
(ZhangJ Z, ZhangX L. Breakthrough innovation identification based on mutation of cited scientific knowledge topics[J]. Data Analysis and Knowledge Discovery, 2016(S1): 42-50.)
(ChenJ, LiH W, ZhouD X. Research on thematic maps based on technology of keywords clustering—taking GIS visualization as an example[J]. Geomatics & Spatial Information Technology, 2010, 33(5): 115-120, 123.)
(ZhouH W, WuC F. Emerging technology identification based on patent SAO structure and multi-index evaluation: mobile chip as an example[J]. Journal of Intelligence, 2022, 41(2): 86-94, 48.)
(FengL J, ZhouW, WangJ F, et al. Examining multi-dimensional technology evolution path and technology innovation opportunity identification based on SAO semantic analysis[J]. Journal of the China Society for Scientific and Technical Information, 2022, 41(11): 1149-1160.)
(LiuX L, TanZ Y. Research on technology topic division method based on patent multi-attribute fusion[J]. Data Analysis and Knowledge Discovery, 2022, 6(S1): 45-54.)
[34]
GrootendorstM. BERTopic: Neural topic modeling with a class-based TF-IDF procedure[J]. arXiv preprint arXiv: 220305794, 2022.
(HaoW K, YangJ L. Predicting emerging topics in social sciences based on semantic representation and dynamic topic model[J]. Information Studies: Theory & Application, 2023, 46(2): 184-193.)
[37]
EggerR, YuJ. A topic modeling comparison between LDA, NMF, Top2Vec, and BERTopic to demystify twitter posts[J]. Frontiers in Sociology, 2022, 7: 886498.
(TengG Q, JiangY, TuoR. Comparative study on domain knowledge evolution based on multiple data source dimensions: taking graphene field research in the United States as an example[J]. Information and Documentation Services, 2023, 44(6): 61-70.)
(XuH Q, TengG Q. Opportunities and challenges: library subject text mining in AI environment based on BERTopic[J]. Information Science: 1-20.)
[42]
ZankadiH, IdrissiA, DaoudiN, et al. Identifying learners’ topical interests from social media content to enrich their course preferences in MOOCs using topic modeling and NLP techniques[J]. Education and Information Technologies, 2023, 28(5): 5567-5584.
[43]
GuizzardiS, ColangeloM T, MirandolaP, et al. Modeling new trends in bone regeneration, using the BERTopic approach[J]. Regenerative Medicine, 2023, 18(9): 719-734.
[44]
ContrerasK, VerbelG, SanchezJ, et al. Using Topic Modelling for Analyzing Panamanian Parliamentary Proceedings with Neural and Statistical Methods; proceedings of the 2022 IEEE 40th Central America and Panama Convention (CONCAPAN), F, 2022[C]. IEEE.
(YaoY Z, LiJ L. Feature weight analysis and improvement of TF-IDF based on category information[J]. Computer Systems & Applications, 2021, 30(9): 237-241.)
(XuD D, WuS B. improved TF-IDF feature selection based on categorical description[J]. New Technology of Library and Information Service, 2015, (3): 39-48.)
(MengL G, MaJ F, ZhangX Y. Meeting summarization technology confused with SVM and MMR based on theme[J]. Computer Engineering and Design, 2016, 37(10): 2695-2699.)
(ZhuX X, SongJ X, MengJ F. Analysis of online public opinion information based on the dynamic theme-emotion evolution model[J]. Information Science, 2019, 37(7): 72-78.)
(ShanX H, HanS X, LiuX Y. Research on disruptive technology identification based on the evolution of technology themes[J]. Information Studies: Theory & Application, 2023, 46(8): 113-123.)
(LiuJ M. With the help of computer technology to promote the application of virtual reality technology effect[J]. Information and Computers (Theoretical Edition), 2023, 35(20): 54-56.)
(WuC S, TongH, FanX M. Overview of research on face generation and detection techniques in generation countermeasures networks[J]. Digital Communication World, 2023, (7): 28-33.)
(GuoW Y, RanH X. Patent technology analysis of scene understanding in computer vision[J]. China Science and Technology Information, 2023, (13): 49-52.)
(ZengZ Q, ChenY, ZouX J, et al. Research on interactive virtual assembly based on machine vision perception[J]. Machinery Design & Manufacture, 2022, (9): 261-266.)
(YangX N, WangS, NiuH W, et al. Research status and prospects of key technologies of eye movement interaction[J]. Computer Integrated Manufacturing System: 1-22.)