Research On Text Detection And Recognition In Natural Scene Based On Deep Neural Network

With the rapid development of the Internet,image has gradually become the main carrier of information,and the text information in the image can help the computer to understand the image content more accurately.However,in natural scenes,the background of text is complex and changeable,the style of text is different,and the embedding position is uncertain.It is difficult to obtain satisfactory results by directly applying OCR to natural scenes.Therefore,this paper studies the method of text detection and recognition in natural scenes based on deep neural network.The research contents of this paper are mainly embodied in the following aspects:1)Proposed a multi-directional text detection method based on deep neural network.In view of the large aspect ratio of text,a convolutional neural network is constructed to extract multi-scale features,and a text line construction algorithm is proposed to enable the network to effectively detect skewed text.2)Proposed a text recognition method based on attention mechanism decoding.The features of the image are first extracted using a deep convolutional neural network,and then the features are encoded and decoded using two circular neural networks.At the time of decoding,the attention mechanism is used to obtain the global features,and the CTC is used to achieve the attention constraint,so that the network can converge quickly.3)In this paper,the proposed detection and identification methods are comprehensively analyzed and verified on different data sets.Compared with the current algorithm,the method proposed in this paper has achieved good results.