A detection and recognition method for sensitive text in natural scene images

Abstract

The invention discloses a detection and recognition method for sensitive characters in natural scene images, comprising the steps of: obtaining training samples, the training sample data including self-built natural scene sensitive character data sets and public character data sets; establishing sensitive characters based on improved DSSD network Directly recognize the model and perform training and learning; obtain a test sample image in a natural scene, input the test sample image into the sensitive text direct recognition model, detect and recognize the sensitive text in the test image, and realize multi-directional, Sensitive word recognition in text regions with complex changes. The invention solves the stability of recognition based on single characters, solves the problem of character samples, simplifies the recognition process, and greatly improves the detection and recognition speed and accuracy of sensitive characters in natural scenes; Sensitive words in the text area are identified.

Description

technical field [0001] The invention belongs to the technical field of image detection and recognition, in particular to a detection and recognition method for sensitive characters in natural scene images. Background technique [0002] In natural scenes, scene text, as a common visual object, widely exists in various occasions, such as on the street, on the wall, etc. Due to the complex background of text in natural scenes, the changeable direction, size and scale, and often affected by uncontrollable factors such as materials and lights, it has great diversity and obvious uncertainty. For example, the text contains multiple languages, and each language contains multiple characters, and each character can have different sizes, fonts, colors, brightness, contrast, etc.; the text usually exists in the form of text lines, but the text lines may have different Arrangement and alignment, such as: multiple lines, horizontal, vertical, curved, etc.; moreover, due to the randomness...

AI Technical Summary

Problems solved by technology

[0004] In order to solve the above problems, the present invention proposes a detection and recognition method for sensitive characters in natural scene images, which solves the stability of recognition based on single characters, solves the problem of character samples, simplifies the recognition process, and greatly improves the sensitivity of natural scenes. Text detection and recognition speed and accuracy; able to identify sensitive words in multi-directional, complex-changing text areas in natural scenes

Images