A method, system and device for automatically recognizing and segmenting text images

Abstract

The invention discloses a method for automatically identifying and segmenting a text image, comprising the steps of: S1, acquiring an input text image and performing preprocessing; S2, segmenting the text image to determine a first target area; S3, in the first target area Run the label detector inside, if the label detector detects successfully, output the detection result to the discriminator for confirmation, and execute step S4, if the label detector fails to detect, then perform step S5; S4, based on the final judgment of the discriminator The position of the label and the information of the first target area are fitted to obtain the second target area and intercepted; S5, using image processing technology to process and fit the third target area and intercepted. Also disclosed are a system and device for automatically recognizing and segmenting text images. Users only need to align the center of the screen with the test question they want to search in the text image, press the camera button, and the program can automatically divide the designated test question area and upload it, so as to achieve a complete and smooth user experience.

Description

technical field [0001] The invention relates to the field of image recognition, in particular to a method, system and device for automatically recognizing and segmenting text images. Background technique [0002] The tutoring of exercises in traditional education often relies on the participation of teachers, which will cause great trouble to students' learning at night or on rest days. With the popularization of mobile devices, Internet search technology is widely welcomed by students. Generally speaking, its technology is to collect test question information through terminal devices (such as mobile phones, tablet computers, etc.), and upload it to the cloud server. The server analyzes and processes the input information, and returns the result with high matching degree to the terminal after comparing with the massive exercise data in the database. [0003] The existing question search system is mainly divided into two routes in terms of terminal collection: 1. Collect te...

AI Technical Summary

Problems solved by technology

Poor user experience due to factors such as screen size and narrow title text spacing

In addition, the completeness of the box questions directly affects the accuracy of the subsequent recognition and search steps, while there are too many differences in the manual selection method, it is difficult to deal with it uniformly, and corresponding preprocessing steps need to be designed

[0006] In the existing mainstream technology, the image area is roughly and automatically divided by the grid method. This method actually transfers the manual adjustment process after taking pictures to before taking pictures, and does not change the problem of poor user experience.

However, although the recognition rate has been improved to a certain extent through the server-side multi-window redundant search method, it has paid a huge cost of computing resource consumption and additional upload traffic.

[0007] In addition, due to accuracy reasons, none of these products completely remove the step of manual box selection by the user

Even if the automatic segmentation is correct, the user still needs to manually confirm before uploading, which does not fundamentally solve the problem of user experience

Images