Character click verification code identification and filling method based on Chinese character structure

Abstract

The invention relates to a Chinese character structure-based character click verification code identification and filling method, which comprises the following steps of: pre-configuring a semantic word group library, collecting semantic Chinese word groups and adding the semantic Chinese word groups for retrieval; pre-configuring a structured Chinese character library, collecting Chinese characters and adding a structure label to a single Chinese character for retrieval; collecting character click verification code pictures, detecting areas with Chinese characters in the character click verification code pictures, cutting the areas into single Chinese character pictures, and recording area coordinates of the Chinese character pictures as filling values; pre-creating a recognition model forpredicting Chinese characters; according to the identification model, performing identification prediction on each Chinese character picture to obtain each predicted Chinese character; carrying out retrieval matching on each predicted Chinese character, an input semantic class phrase library and a structured Chinese character library, and carrying out weighted comprehensive evaluation to obtain atarget Chinese character filling sequence; and according to the obtained target Chinese character filling sequence and the regional coordinates of each Chinese character picture, clicking a verification code in a verification window of the verification code and submitting the verification code.

Description

technical field [0001] The invention relates to a method for identifying and filling in a text-selection verification code based on a Chinese character structure, and belongs to the technical field of verification code identification. Background technique [0002] There are generally two types of verification codes for text selection. One type prompts the text and order to be selected outside the verification code picture, which is called non-semantic type selection; the other type does not prompt and needs to be selected according to the semantic order. Clicks, usually Chinese phrases such as idioms, food, scenery, etc., are called semantic clicks. [0003] The traditional text-click verification code recognition process generally organizes and trains a classification model based on each Chinese character. The specific operation process is as figure 1 As shown, first, the collected text verification codes are preprocessed by denoising, binarization, etc., and then cut; af...

AI Technical Summary

Problems solved by technology

Therefore, if you need about 50 samples according to one category, you need to manually collect nearly 200,000 samples according to common Chinese characters. If there are more Chinese characters to collect samples, you need to collect more samples. Complex, simple recognition model combined with random algorithm is easy to lead to low recognition accuracy

Images