面向GUI界面的轻量级文本识别算法  

A Lightweight Text Recognition Algorithm for GUI Interface

作　　者：马荧炜 MA Yingwei(Nanjing University of Aeronautics and Astronautics,Nanjing 210016,China)

机构地区：[1]南京航空航天大学,江苏南京210016

出　　处：《现代信息科技》2024年第14期94-97,共4页Modern Information Technology

摘　　要：图形用户界面成为用户与软件进行交互的必要方式之一。然而,随着软件规模和复杂性的增长,测试代价愈加高昂。基于OCR算法的文本控件识别方法引起了越来越多的关注,例如Sikuli通过OCR技术匹配窗口控件的文本,从屏幕上基于图像匹配来定位窗口控件。目前OCR引擎应用广泛,但是OCR算法通常运行缓慢,它们可能需要几秒钟的时间来识别GUI屏幕上的文本。文章介绍了一种快速的轻量级文本识别算法以加速OCR的文本识别过程。实验表明,在RICO数据集上,该算法降低了基于OCR的文本识别84%的平均时间。Graphical User Interface(GUI)has become an essential way for users to interact with software.As the scale and complexity of software increase,the cost of testing also rises.Text control recognition methods based on OCR algorithms have attracted increasing attention.For example,Sikuli uses OCR technology to match the text of window widget,enabling the localization of window widget based on image matching on the screen.Currently,OCR engines are widely used.However,OCR algorithms often run slowly,and they may cost several seconds to recognize text on the GUI screen.The paper introduces a lightweight text recognition algorithm to accelerate the text recognition process of OCR.Experiments show that on the Rico data set,this algorithm reduces the average time for OCR-based text recognition by 84%.

关 键 词：GUI测试 OCR 计算机视觉 

分 类 号：TP311.5[自动化与计算机技术—计算机软件与理论]