[1]蒋志鹏,潘坤榕,张国林,等.基于置信度融合的自然场景文本检测方法[J].计算机技术与发展,2021,31(08):39-44.[doi:10. 3969 / j. issn. 1673-629X. 2021. 08. 007]
　JIANG Zhi-peng,PAN Kun-rong,ZHANG Guo-lin,et al.Research on Scene Text Detection Based on Confidence Fusion[J].,2021,31(08):39-44.[doi:10. 3969 / j. issn. 1673-629X. 2021. 08. 007]
点击复制

基于置信度融合的自然场景文本检测方法()
分享到：

《计算机技术与发展》[ISSN:1006-6977/CN:61-1281/TN]

卷:
31
期数:
2021年08期
页码:
39-44
栏目:
图形与图像
出版日期:
2021-08-10

文章信息/Info

Title:
Research on Scene Text Detection Based on Confidence Fusion
文章编号:
1673-629X(2021)08-0039-06
作者:
蒋志鹏1; 潘坤榕1; 张国林1; 刘玉琪1; 张 瑛1; 孙科学1; 2*
1. 南京邮电大学 电子与光学工程学院,江苏 南京 210023;
2. 射频集成与微组装技术国家地方联合工程实验室,江苏 南京 210023
Author(s):
JIANG Zhi-peng1; PAN Kun-rong1; ZHANG Guo-lin1; LIU Yu-qi1; ZHANG Ying1; SUN Ke-xue1; 2*
1. School of Electronic and Optical Engineering,Nanjing University of Posts and Telecommunications,Nanjing 210023,China;
2. Nation-Local Joint Project Engineering Lab of RF Integration & Micropackage,Nanjing 210023,China
关键词:
自然场景; 文本检测; 卷积神经网络; 非极大抑制; 置信度融合
Keywords:
natural scene; text detection; convolution neural network; non-maximal suppression; confidence fusion
分类号:
TP391. 4
DOI:
10. 3969 / j. issn. 1673-629X. 2021. 08. 007
摘要:
自然场景文本检测技术已经成为计算机视觉领域重要的研究任务,在图像检索、辅助驾驶、工业检测等领域具有广泛应用。 在现有的基于深度学习的自然场景文本检测方法中,非极大抑制算法在对同一个真实文本框的重复检测进行合并和筛选时,将预测框的分类置信度作为排序依据,导致那些定位更精确而分类置信度略低的预测框被抑制, 从而影响检测准确率。 为了提高预测框的定位精确度,文中提出基于置信度融合的文本检测方法。 首先,设计了交并比网络,作为每个预测框的定位置信度; 其次,在非极大抑制算法中,将定位置信度与文本分类置信度融合作为预测框排序的依据;最后,在 ICDAR2011 和 ICDAR2013 数据集上对该方法进行了实验,结果表明,该方法检测的文本框更加紧致,包含的背景区域更少,可以提高文本检测的准确率。
Abstract:
Natural scene text detection technology has become an important research task in the field of computer vision,and has a wide range of applications in image retrieval, driver assistance,industrial detection and other fields. In the existing natural scene text detection methods based on deep learning,when the non-maximum suppression algorithm merges and filters the repeated detection of the same real text box, it uses the classification confidence of the prediction box as the sorting basis,resulting in? the suppression of those prediction boxes with more accurate positioning and slightly lower classification confidence, which affects the detection accuracy. In order to improve the positioning accuracy of the prediction frame,we propose a text detection method based on confidence fusion. Firstly,the intersection and comparison network is designed as the positional reliability of each prediction box; Secondly, in the non - maximum suppression algorithm,the fusion of the positional reliability and the text classification confidence is used as the basis for ranking the prediction boxes. Finally,experiments on the ICDAR2011 and ICDAR2013 data sets show that the text box detected by this method is more compact and contains less background area,which can improve the accuracy of text detection.

相似文献/References:

[1]许肖,顾磊. 复杂背景下文本检测研究[J].计算机技术与发展,2015,25(03):40.
　XU Xiao,GU Lei. Research on Text Detection under Complex Background[J].,2015,25(08):40.
[2]陈梓洋,王宇飞,钱侃,等. 自然场景下基于区域检测的文字识别算法[J].计算机技术与发展,2015,25(07):230.
　CHEN Zi-yang,WANG Yu-fei,QIAN Kan,et al. Character Recognition Algorithm Based on Region Detection in Natural Scene[J].,2015,25(08):230.
[3]李虎月,郝鹏飞,廖云霞,等.车牌检测与识别系统设计[J].计算机技术与发展,2020,30(07):150.[doi:10. 3969 / j. issn. 1673-629X. 2020. 07. 032]
　LI Hu-yue,HAO Peng-fei,LIAO Yun-xia,et al.Design of License Plate Detection and Recognition System[J].,2020,30(08):150.[doi:10. 3969 / j. issn. 1673-629X. 2020. 07. 032]
[4]周成伟. 基于卷积神经网络的自然场景中数字识别[J].计算机技术与发展,2017,27(11):101.
　ZHOU Cheng-wei. Recognition of Numbers in Natural Scene with Convolutional Neural Network[J].,2017,27(08):101.
[5]郭 闯,邱晓晖.基于 BLSTM 网络的改进 EAST 文本检测算法[J].计算机技术与发展,2020,30(07):21.[doi:10. 3969 / j. issn. 1673-629X. 2020. 07. 005]
　GUO Chuang,QIU Xiao-hui.Improved EAST Natural Scene Text Location Algorithm Based on BLSTM Network[J].,2020,30(08):21.[doi:10. 3969 / j. issn. 1673-629X. 2020. 07. 005]
[6]彭祥云,陈 黎.安防视频时间戳同步检测方法研究[J].计算机技术与发展,2021,31(11):195.[doi:10. 3969 / j. issn. 1673-629X. 2021. 11. 032]
　PENG Xiang-yun,CHEN Li.Research on Synchronous Detection Method of Security Video Time Stamp[J].,2021,31(08):195.[doi:10. 3969 / j. issn. 1673-629X. 2021. 11. 032]
[7]李 普,陈 黎.基于超像素随机游走的自然场景图像分割方法[J].计算机技术与发展,2021,31(12):61.[doi:10. 3969 / j. issn. 1673-629X. 2021. 12. 011]
　LI Pu,CHEN Li.Natural Scene Image Segmentation Method Based on Super-pixel Random Walk[J].,2021,31(08):61.[doi:10. 3969 / j. issn. 1673-629X. 2021. 12. 011]
[8]赵晓芹.融合局部特征与全局特征的场景文本检测算法[J].计算机技术与发展,2022,32(S2):25.[doi:10. 3969 / j. issn. 1673-629X. 2022. S2. 004]
　ZHAO Xiao-qin.Scene Text Detection Algorithm Combining Local and Global Features[J].,2022,32(08):25.[doi:10. 3969 / j. issn. 1673-629X. 2022. S2. 004]
[9]方承志,倪梦媛,唐 亮.基于残差网络及笔画宽度变换的场景文本检测[J].计算机技术与发展,2023,33(01):49.[doi:10. 3969 / j. issn. 1673-629X. 2023. 01. 008]
　FANG Cheng-zhi,NI Meng-yuan,TANG Liang.Scene Text Detection Based on Residual Network andStroke Width Transform[J].,2023,33(08):49.[doi:10. 3969 / j. issn. 1673-629X. 2023. 01. 008]
[10]丁 泽,程艳云.面向自然街景改进的文本检测[J].计算机技术与发展,2023,33(04):82.[doi:10. 3969 / j. issn. 1673-629X. 2023. 04. 012]
　DING Ze,CHENG Yan-yun.Improved Text Detection for Natural Streetscape[J].,2023,33(08):82.[doi:10. 3969 / j. issn. 1673-629X. 2023. 04. 012]

更新日期/Last Update: 2021-08-10