自然场景图像中的文本检测现状研究和展望  

Research and Prospect of Text Recognition in Natural Scene Images

作　　者：刘洁芯 陈虹光 邓成梁 LIU Jie-xin;CHEN Hong-guang;DENG Cheng-liang(Department of Information Engineering,Southwest Jiaotong University Hope College,Chengdu 610000;College of Information Engineering,Chengdu University of Information Technology,Chengdu 610000)

机构地区：[1]西南交通大学希望学院信息工程系,成都610000 [2]成都信息工程大学软件工程学院,成都610000

出　　处：《现代计算机》2021年第5期81-87,共7页Modern Computer

摘　　要：对自然场景图像中的文本检测问题及其方法的研究进展进行综述:首先,论述自然场景图像中的文本检测方法的研究背景与意义,文本的基本特点和难点,其次,从传统文本检测以及深度学习文本检测的视角出发,梳理、分析、总结现阶段自然场景图像下的文本检测方法的基本原理、优势与不足。最后,总结自然场景图像下的文本检测领域面临的一些问题和发展方向。In this paper,the natural scene text detection problems in image method and its research progress were summarized:first,dis⁃cusses the method of detecting text in natural scene images of the research background and significance,the basic character⁃istics of the text and the difficulty,secondly,from the perspective of traditional text detection and deep learning text,review,analysis,summarize the present stage under natural scene image text detection method basic principle,advantages and disad⁃vantages.Finally,some problems and development directions of text detection in natural scene images are summarized.

关 键 词：证件识别 模式识别 图像特征 文本检测 深度学习 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]