基于R2CNN的自然场景图像中文本检测方法

Text detection method in natural scene image based on R2CNN

在互联网世界中,图片是传递信息的重要媒介。特别是电子商务、社交、搜索等领域,每天都有数以亿兆级别的图像在传播。自然场景就是我们所处的生活环境,自然场景图像中存在着大量的文本信息,例如路标信息、商店门店信息、商品包装信息等。随着深度学习的发展,基于深度学习的文本检测技术也逐渐流行起来。文章主要提出的是基于R2CNN的文本检测算法。在R2CNN算法的基础上对算法的结构进行改进,最终算法在ICDAR2015数据集上的召回率为87.2%,精确率为81.43%。

In the Internet world,pictures are an important medium for transmitting information.Especially in the fields of e-commerce,social networking,search,etc.,hundreds of millions of images are being transmitted every day.The natural scene is the living environment we are in.There are a lot of text information in the natural scene image,such as road sign information,store store information,product packaging information and so on.With the development of deep learning,text-based detection technology based on deep learning has also become popular.This paper mainly proposes a text detection algorithm based on R2CNN.Based on the R2CNN algorithm,the structure of the algorithm is improved.The final algorithm has a recall rate of 87.2%and an accuracy rate of 81.43%on the ICDAR2015 data set.