Text feature extraction method and device, electronic equipment and readable storage medium

Abstract

The embodiment of the invention provides a text feature extraction method and device, electronic equipment and a readable storage medium. The text feature extraction method comprises the steps of obtaining a to-be-processed image containing a first text area; and inputting the to-be-processed image into a preset convolutional neural network to obtain first text feature information of the first text region. Wherein the preset convolutional neural network is obtained by training a plurality of groups of sample images containing sample text regions; wherein each group of sample images comprises afirst sample image containing the same text content information in a sample text area and a second sample image containing the text content information different from the text content information ofthe first sample image in the sample text area. Through the method provided by the embodiment of the invention, the accuracy of text feature information extraction can be improved in the text featureextraction process, so that the accuracy of text region image matching in the video data is improved.

Description

technical field [0001] The invention relates to the technical field of streaming media video, in particular to a feature extraction method, device, electronic equipment and readable storage medium. Background technique [0002] In recent years, there have been many text recognition methods based on static images, but there are relatively few text recognition methods based on video data. Because video data is more susceptible to interference from various noises in the external environment than static images, for example, changes in lighting, motion blur, changes in text font size, occlusion, etc. will affect video data. Therefore, in order to ensure that the image quality of all frames in the video data is maintained at a high level, text tracking technology can be used to match and associate text region images belonging to the same text between different frames of video data, and obtain text image sequences of the same text. Therefore, according to the result of the matchin...

AI Technical Summary

Problems solved by technology

[0004] In the above text feature extraction process, because the low-level image appearance information is easily affected by external noise, image distortion will occur, and, in the same background environment, the appearance of text image regions containing different texts may be relatively similar. The extracted low-level image appearance information will be biased, and the text area images of different texts will be matched and associated, which will affect the text area image matching results

Images