Scene character recognition method based on scene classification and super-resolution

Abstract

The invention relates to a scene character recognition method based on scene classification and super-resolution. The method comprises the steps of By constructing the c-ssd and sr-rcnn networks, thecorresponding convolution layer parameters are initialized respectively, the data set is expanded, and the c-ssd network and the sr-crnn network are trained to obtain the corresponding c-ssd model andsr-crnn; evaluating the trained model with the evaluation data set; if the target is reached, the gestured image is input into the trained c-ssd model for processing, returning the coordinate position and scene information of the gesture, and designing the error detector with the text feature; returning to the final scene information, select the corresponding text recognition model for text recognition, and get the final recognition result. According to the invention, classification and detection are realized by using an independent network, an end-to-end algorithm is realized, different operations do not need to be completed by using a plurality of networks respectively, and the identification precision of low-resolution characters can be improved.

Description

technical field [0001] The invention belongs to the technical field of general image data processing or generation, and in particular relates to a scene text recognition method based on scene classification and super-resolution to help amblyopia group to indicate the position and understand the text in the front position scene. Background technique [0002] Scene text is one of the most common visual objects in natural scenes, often appearing on road signs, license plates, product packages, etc. Reading scene text facilitates many useful applications, such as image-based geolocation. [0003] Recently, the use of photos in social networks has been growing, photo posts generally generate more engagement than text-only posts, and interestingly, most of the images circulated through the Web have embedded text. First, text embedded in online photos can accompany important information about the photo such as author, location, and time, second, if the image is a video frame captur...

AI Technical Summary

Problems solved by technology

[0006] The invention solves the problem that in the prior art, the algorithms for classification and detection are separated, and two algorithms are needed for the recognition of scene characters, which cannot achieve end-to-end realization. At the same time, the existing recognition algorithm depends on the resolution of the original image. When the resolution of the Chinese text area is low, errors often occur in the recognition, resulting in recognition failure or errors. An optimized scene text recognition method based on scene classification and super-resolution is provided.

Images