Handwritten Mongolian detection and recognition method based on segmentation and deformation LSTM

Abstract

The invention discloses a handwritten Mongolian detection and recognition method based on segmentation and deformation LSTM. The detection of handwritten Mongolian in a complex environment is realized by using a segmentation-based arbitrary shape text detector SAST. A RoI Rotate module is used for combining a text detection function and a text recognition function; The extracted text candidate box is taken as an input image, and the text recognition of the input image is realized by using a deformation LSTM in combination with a CTC module. According to the method, the SAST is utilized to more effectively extract the polygon representation of the text in any shape. Meanwhile, the long-range correlation of pixels can be captured, a more reliable segmentation result can be obtained at a time, the contents of the detection stage and the recognition stage of the handwritten Mongolian are connected through the application of the RoI Rotate module, and the training efficiency can be further improved. The recurrent neural network is combined with the deformation LSTM, so that the recognition accuracy can be further improved in the implementation of handwritten Mongolian recognition.

Description

technical field [0001] The invention belongs to the technical field of artificial intelligence and relates to character detection and recognition, in particular to a method for detecting and recognizing handwritten Mongolian characters based on segmentation and deformation LSTM. Background technique [0002] With the rapid development of the Internet and artificial intelligence, education informatization has begun to affect and change traditional education methods. Human-computer interaction scenarios such as online answers are becoming more and more common. The problem of handwriting detection and recognition has become a research direction in the field of computer vision. It is very simple for humans to recognize and recognize handwritten text, but it is very complicated for computers. In recent years, the development of deep convolutional neural networks has brought revolutionary changes to the field of computer vision. The combination of convolutional neural networks and...

AI Technical Summary

Problems solved by technology

[0003] However, the handwriting detection and recognition of minority languages ​​such as Mongolian started late, and there are few related researches, and Mongolian has the characteristics of huge vocabulary, free writing, serious character deformation, and long writing, which all give Mongolian handwriting Both detection and identification pose significant challenges

Images