文本检测算法的发展与挑战

对自然场景中的文字进行识别和理解是大量计算机视觉应用的基础。文本检测算法旨在识别出自然图像中的文字信息，目前已经成为计算机视觉和智能信息处理领域研究的一个热点。本文首先对文本检测算法的目标、技术路线及其所面对的挑战进行了分析与介绍。然后回顾了几种经典的文本检测算法，并介绍了两种代表最新研究趋势的深度学习型文本检测算法。进一步，本文阐述了几个主流的文本检测数据集并总结了一些代表性文本检测算法在这些数据集上的检测结果。最后，本文讨论了文本检测的研究现状、面临的挑战和发展的趋势。 

The Developments and Challenges of Text Detection Algorithms

Reading text in natural scene is fundamental to computer vision and intelligent information processing. In this way, text detection is a major part of recognizing text and thus has become very popular topic in recent years. In this paper, we first introduce the object, methods, and challenges of text detection. We then review some classic algorithms on text detection, and introduce two deep learning based algorithms that represent the trends of text detection research. Moreover, we summary typical text detection datasets available, as well as the results of representative algorithms on these datasets. Finally, we conclude the current researches on text detection and the challenges we face, and point out some prospective directions of text detection.