基于改进TextBoxes++的多方向场景文字识别算法的研究

Deformation-Sensitive Word Filtering Algorithm Based on Improved Trie Tree

多方向自然场景文字识别是计算机视觉领域中最困难和最有价值的挑战之一。现存的大多数方法只针对水平方向的文字,或将文字检测和识别视为单独的任务。基于目前先进的多方向场景文字检测算法TextBoxes++提出一个统一的端到端可训练的多方向文字识别方法,用于文字的同时检测和识别。为了适应多方向文字的识别,在TextBox-es++文字检测分支添加对四边形文字框角度的预测;并且通过添加文字识别分支扩展TextBoxes++的网络结构,用于文字的识别;引入RoIRotate以在检测和识别之间共享卷积特征。在公开数据集ICDAR 2015和ICDAR 2017 MLT上的实验证明所提出的方法的有效性。

Multi-directional natural scene text recognition is one of the most difficult and valuable challenges in the field of computer vision. Most existing methods are only for horizontal text, or for text detection and recognition as separate tasks. Based on the current advanced multi-directional scene text detection algorithm TextBoxes++, proposes a unified end-to-end trainable multi-directional text recognition method for simultaneous detection and recognition of text. In order to adapt to the recognition of multi-directional text, the text box of TextBoxes++adds a prediction of the angle of the quadrilateral text box; and by adding text recognition branch to extend the network structure of TextBoxes++ for text recognition; finally, introduces RoIRotate for detection and recognition, the convolution feature is shared between. Experiments on the public dataset ICDAR 2015 and ICDAR 2017 MLT demonstrate the effectiveness of the proposed method.