Segmentation and recognition method and system for combined verification code

Abstract

The invention discloses a segmentation and recognition method for a combined verification code. The method comprises the steps that RGB three-channel separation is performed on an original image, anda G-channel image is selected; binarization processing and open operation with 3*3 being a core are performed on the selected G-channel image to obtain a binary character image; white connected domains are subjected to statistical analysis, a connected domain threshold is set, and whether the binary character image has a black background or a white background is judged; according to different ranges of connected domain statistical values, X-axis ordering of the connected domains is performed, X-axis positions in a maximum interval of the connected domains are calculated, and a whole image withwords being units is intercepted; and equalization processing and binarization processing are performed on the intercepted image, and pictures based on word group segmentation are output; and the output pictures are trained and recognized. Through the segmentation and recognition scheme for the combined verification code, by improving a basic framework of image segmentation, unit particles and atraining object of the image are converted from pure Chinese characters into word groups based on the Chinese characters, and segmentation and recognition of four types of word group pictures can be met by the adoption of the same algorithm.

Description

technical field [0001] The invention relates to the field of image processing, in particular to a method for segmenting and recognizing character combination verification codes and a system thereof. Background technique [0002] The identification of verification codes has been developed for a long time, and its identification methods can be divided into three categories: identification methods based on statistical features, identification methods based on structural features, and a combination of various machine learning methods in the field of artificial intelligence based identification method. [0003] Among the many forms of verification codes, verification codes of numbers, images, characters and combinations thereof are currently the most commonly used. Especially for the text combination, for the current processing of this type of verification code, the conventional graphics algorithm is generally used to preprocess the picture, and finally a valid information and a...

AI Technical Summary

Problems solved by technology

[0004] The main problem of the existing technology is to individually identify each Chinese character in the problem, and the concept of phrase is missing, but in this type of problem, a single Chinese character is meaningless, and in the existing mature algorithm model, for special The support effect of uncommon fonts and Chinese characters is not very good, so there are certain uncertain factors in the accuracy of the Chinese characters that will be recognized soon

Images