Research On Natural Scene Text Detection Based On Deep Neural Network

The Internet,as an important way to acquire information,contains a wide range of content,and has a fast transmission speed.There is an increasing problem in network information security.In order to get away from the network filtering system which can block certain kind of information.some people try to disguise this kind of information in many different ways.For example some web pages embed texts into image files.Now we can see that the study of image text detection and recognition technology has special significance for network information security and network filtering.For text detection of natural scenes,the background is complex and it is difficult for traditional algorithms to accurately detect and locate.This poses a huge problem for subsequent text recognition.However,with the re-emergence of machine learning in recent years,deep neural networks seem to bring new opportunities to solve this problem.Accuracy is no longer the sole criterion for measurement.The speed of detection is also crucial.Because only fairly high speed can make detection for practical use.Detection of text in natural scenes is to detect and locate the text in the natural scene picture to prepare for subsequent text recognition work.In this paper,the text detection method based on SSD-CTPN is proposed,and a text detection system based on deep neural network is developed.The SSD and CTPN method are combined in series,and the SSD is used to rapidly detect the vertical local position of the text,and the CTPN is finally used.Determining the exact horizontal position of the text line enhances the accuracy and effectiveness of multi-text line detection while improving the speed of text detection.After evaluation of different methods in open data test dataset.The results shows that the proposed algorithm has better text detection performance than some existing algorithms.