摘要：

Abstract:

【Purpose/significance】Analyze and refine the latest research hotspots and research methods based on deep learning image

retrieval locally and internationally, and reveal future research trends in this field.【Method/process】The paper uses International

top-level artificial intelligence conference set (CVPR/AAAI/IJCAI/NIPS/SIGIR), CNKI and Web of Science core collection, conducts

word frequency statistics on related documents, and generates word cloud maps to analyze the research hotspots in the field; uses the

Citespace visualization tool to quantify the evolution of research topics. It is reviewed in detail from four dimensions: research topic, al⁃

gorithm, model and application practice.【Result/conclusion】The existing image retrieval mainly focuses on the application of convo⁃

lutional neural networks in image representation and classification. In the future, further research is needed on the real-time of image

classification, feature fusion of deep hash algorithm, and intelligent application of image retrieval.【Innovation/limitation】The innova⁃

tion of this article is to systematically summarize the application status and future development trends of image retrieval based on deep

learning. In addition, in future research, the research data should be further expanded and enriched, so as to conduct in-depth discus⁃

sion and analysis of this research topic more comprehensively and accurately.