基于增强特征金字塔网络的场景文本检测算法  

Scene Text Detection Algorithm Based on Enhanced Feature Pyramid Network

作　　者：邵海琳 季怡[1] 刘纯平[1] 徐云龙[2] SHAO Hai-lin;JI Yi;LIU Chun-ping;XU Yun-long(School of Computer Science and Technology,Soochow University,Suzhou,Jiangsu 215006,China;Applied Technology College of Soochow University,Suzhou,Jiangsu 215300,China)

机构地区：[1]苏州大学计算机科学与技术学院,江苏苏州215006 [2]苏州大学应用技术学院,江苏苏州215300

出　　处：《计算机科学》2022年第2期248-255,共8页Computer Science

基　　金：国家自然科学基金(61972059,61773272,61602332);江苏省高校自然科学基金重点项目(19KJA230001);吉林大学符号计算与知识工程教育部重点实验室项目(93K172016K08);江苏高校优势学科建设工程资助项目。

摘　　要：场景文本检测有助于机器理解图像内容,在智能交通、场景理解和智能导航等领域应用广泛。现有的场景文本检测算法未充分利用高层语义信息和空间信息,限制了模型对复杂背景像素的分类能力和对不同尺度的文本实例的检测和定位能力。为解决上述问题,提出了一种基于增强特征金字塔网络的场景文本检测算法。该算法包括比率不变特征增强(Ratio Invariant Feature Enhanced,RIFE)模块和重建空间分辨率(Rebuild Spatial Resolution,RSR)模块。RIFE模块作为残差分支,增强了网络的高层语义信息传递,提高了分类能力,降低了误报率和漏捡率。RSR模块重建多层特征分辨率,利用丰富的空间信息改进边界位置。实验结果表明,所提算法提升了在多方向文本数据集ICDAR2015、弯曲文本数据集Totaltext以及长文本数据集MSRA-TD500上的检测能力。Scene text detection helps machines understand image content,and is widely used in the fields such as intelligent transportation,scene understanding,and intelligent navigation.Existing scene text detection algorithms do not make full use of high-level semantic information and spatial information,which limits the model’s ability to classify complex background pixels and the ability to detect and locate text instances of different scales.In order to solve the above problems,a scene text detection algorithm based on enhanced feature pyramid network is proposed.The algorithm includes a RIFE(ratio invariant feature enhanced)mo-dule and a RSR(rebuild spatial resolution)module.As the residual branch,the RIFE module enhances the high-level semantic information transmission of the network,improves the classification ability,and reduces the false positive rate and the false negative rate.The RSR module reconstructs multi-layer feature resolution and uses rich spatial information to improve the boundary location.Experimental results show that the proposed algorithm improves the detection capabilities on the multi-directional text dataset ICDAR2015,the curved text dataset Totaltext,and the long text dataset MSRA-TD500.

关 键 词：场景文本检测 特征金字塔网络 语义信息 空间信息 边界位置 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]