Text detection method, computing device and readable storage medium

Abstract

The invention discloses a text detection method which is suitable for being executed in computing equipment, and the method comprises the steps: obtaining a to-be-processed image which comprises textinformation, and the text information comprises one or more lines of texts; inputting the to-be-processed image into a first target detection model for detection to obtain a text image containing textinformation; inputting the text image into a second target detection model for detection, and obtaining an individual character detection box containing individual character information; and sortingall the individual character information according to the coordinates of the individual character detection box to obtain complete text information. The invention also discloses corresponding computing equipment and a readable storage medium.

Description

technical field [0001] The invention relates to the field of image processing, in particular to a text detection method, computing equipment and a readable storage medium. Background technique [0002] With the development of deep learning technology in detection and recognition, for text recognition, a typical text detection method is the combination of target detection and recognition (such as Faster R-CNN, YoloV3, etc.) and convolutional neural network (CRNN) for sequence recognition The method is to first use the target detection and recognition algorithm to detect and recognize the text area, and then use the CRNN algorithm to perform sequence recognition on the text area. Since the CRNN algorithm needs a large amount of data to train the model accurately, recently someone has proposed a multiple detection algorithm, that is, the first step is to use the detection and recognition algorithm to detect the text area, and the second step is to use the detection and recognit...

AI Technical Summary

Problems solved by technology

[0003] The current method for multi-line text detection is a multiple detection scheme, which marks different lines in the same recognition item of the text as different categories, and then performs single-character recognition for each area, and finally stitches the results together to identify Complete text information, because different lines of the same identification item are marked in different positions, which increases the number of label categories and labels in the detection and identification algorithm, increases the memory usage and calculation time complexity, and reduces the identification. Speed, on the other hand, because the method of multiple detection is proposed for small data sets, there may be missed detection for text recognition containing multi-line text information

[0004] Another method is to mark different lines of the same identification item of text in the same label box, and then mark the area a second time, identify different lines as different categories, detect and identify a single line, and then mark each line One line is used for single character recognition, and finally the results are spliced ​​together to recognize the complete text information. The first step of the scheme is to detect the overall area, and the second step is to detect a single line based on this, which increases the degree of detection and recognition. Therefore Make the process more cumbersome and slow down

Images