A multi-threshold fused crown size extraction method based on integrals

Abstract

The invention discloses a multi-threshold fused crown size extraction method based on an integral map, which comprises the following steps: (1) collecting a white light image of an object to be identified including one side of the crown size image and normalizing to obtain a gray-scale image including the crown size image region; 2) carrying out low threshold binarization on that first image to obtain a first image, and projecting the first image to a row and a column to obtain a second image; (3) calculating an integrating map for the second image and calculating Haar in the integrating map;Like H characteristic; (4) performing low threshold inverse binarization processing on the second image to obtain a third image containing only crown characters, and calculating a confidence matrix according to the third image; (5) performing high threshold inverse binarization processing on the second image to obtain a fourth image; using Haar-Removing the background texture between the crown characters in the fourth image from the Like H characteristic coordinates to obtain a fifth image; (6) using confidence matrix to mask the fifth image, removing the texture in the crown character, and obtaining the clear crown image. The method can be adapted to various conditions to achieve accurate extraction of crown size.

Description

technical field [0001] The invention belongs to the technical field of image processing, and more specifically relates to a multi-threshold fusion serial number extraction method based on an integral graph under a complex background. Background technique [0002] With the rapid and healthy development of the economy, it is necessary to effectively supervise the tickets with government credit (including banknotes, licenses, checks, property rights certificates, etc.). Financial and economic security is of great significance. At present, most of the automatic ticket recognition machines on the market do not have the ability to accurately identify the serial number; most of the machines with the ability to extract and record the serial number are oriented to the extraction of the serial number in simple backgrounds such as solid colors, for complex textures and obvious stains The extraction rate of serial numbers is extremely low, which cannot meet the extraction and recogniti...

AI Technical Summary

Problems solved by technology

At present, most of the automatic ticket recognition machines on the market do not have the ability to accurately identify the serial number; most of the machines with the ability to extract and record the serial number are oriented to the extraction of the serial number in simple backgrounds such as solid colors, for complex textures and obvious stains The extraction rate of serial numbers is extremely low, which cannot meet the extraction and recognition requirements under various conditions

[0003] In the prior art method of extracting the serial number from the banknote image, the banknote image is first horizontally corrected to reduce the error caused by the random appearance of the banknote image position in the source image, then the feature value is extracted, the face value is determined, and the feature value is calculated. Gray value matching, this method of extracting the serial number from the scanned banknote image can give strong support for the identification of the authenticity of the banknote; however, it cannot be applied to the extraction of the serial number under the condition of complex texture or obvious stains

[0004] Most of the existing technologies for processing dirty serial numbers are classification techniques, by extracting the precise area containing the serial number in the horizontal direction and the precise area containing the serial number in the vertical direction in the banknote image, and comparing the black spots of the characters in the accurate area containing the serial number in the horizontal direction The number and the number of black dots of the corresponding position characters in the precise area containing the serial number in the vertical direction, according to the comparison result, the serial number is classified, and the dirty serial number and normal serial number of banknotes can be accurately classified, but for the classified dirty serial number The image of the serial number is no longer recognized

Images