A method and system for automatically detecting character types in bills

Abstract

The invention discloses a method and system for automatically detecting character types in bills. The method includes: A. acquiring infrared images and color images of bills; B. binarizing the infrared images and color images, and outputting corresponding binary images ; C. Locate the character area in the binary image; D. Detect the outline of the character and generate a circumscribing rectangle, and judge the character type according to the parameters of the circumscribing rectangle. A system for correspondingly performing the method. The present invention extracts the characters through a series of processing, and performs characteristic analysis according to the parameters of the outline of the characters to distinguish between handwritten and printed. Improve identification efficiency.

Description

technical field [0001] The invention relates to a method and system for automatically detecting character types in bills, belonging to the field of image recognition. Background technique [0002] Banks or units can fill in some elements of bills manually or type them by machine. In character recognition, the recognition methods of printed and handwritten characters are very different. If the recognition method of printed characters is used for handwritten characters, it will be more accurate. The rate is very poor, and the handwritten character recognition method is not suitable for printed character recognition, and the time complexity is high and wastes resources. At the same time, banks and units also have requirements for the material and color of the pen written on the bill. What is filled in with an irregular pen must be handwritten, so the distinction between handwritten characters and printed characters is also of great significance for the detection of element fill...

AI Technical Summary

Problems solved by technology

This method has the following deficiencies: (1) The scope of application is narrow, and the method is only suitable for judging whether the capitalized amount is printed or handwritten; (2) After the capitalized amount is segmented, the method of judging whether the capitalized amount is printed or handwritten involves features Extraction and training are relatively complicated, time-complexity is high, and resources are wasted; (3) In terms of feature extraction, when the overall color of the image is lighter in this method, the calculated feature values ​​of horizontal and vertical strokes will be smaller; when the color of the check image is When it is very deep, the eigenvalues ​​of the number of connected domains obtained will be small; it is easy to cause misclassification

Images