CN111626299A - Outline-based digital character recognition method - Google Patents

Abstract

The invention provides a contour-based digital character recognition method, which is used for preprocessing a target digital image of a character to be recognized, and comprises graying processing and binarization processing; finding out the maximum external closed contour, and selecting the interested area to remove the interference of the disordered edge background to obtain a digital interested area; automatically vertically and horizontally dividing the selected sensitive digital area, and acquiring and recording a minimum area containing the number and the initial coordinate and the end coordinate position of the horizontal and vertical axes of each digital character; segmenting ideal feeling digital areas and single digital areas according to the acquired data coordinate information; and performing digital recognition and calculation combination by utilizing the contour features with different numbers. The invention does not need the acquisition of a large amount of template data training sets or the learning training of a neural network algorithm, and improves the characteristics of complex data processing, low efficiency, low operation speed and the like of a common template matching method.

Description

Outline-based digital character recognition method

Technical Field

The invention belongs to the technical field of metering detection, and particularly relates to a digital character recognition method based on a contour.

Background

At present, computer vision technology is rapidly developed, manual reading in many scenes is time-consuming and labor-consuming, and automatic processing of a machine, such as license plate recognition, speed real-time recording, vehicle speed detection and the like, needs to be realized. The most common recognition methods for numbers at present are template matching based methods and neural network based methods. And (3) developing a basic template matching method by template matching, and a method such as an SVM training set derived from the basic template matching method. Most of the template matching methods count the characteristics of the outline, projection, grid and the like of the character, and traverse the image data set for matching, so that the identification efficiency is low, the time consumption is long and the identification capability of similar characters is poor due to too large dimension of the characteristic data; the neural network has higher hardware requirement for designing a network structure and selecting a network, cannot be developed by using a common ARM board, does not use development and migration, and has poorer applicability; the method for the SVM training set is high in accuracy, a large number of training subsets are needed for training, operation is complex, operation and learning are not easy to conduct on some specific occasions, timeliness is achieved, and the occasions which require high response speed are not used for processing.

Disclosure of Invention

The invention aims to provide a digital character recognition method based on outlines.

The technical solution for realizing the purpose of the invention is as follows: a digital character recognition method based on contour includes the following steps:

step 1: preprocessing a target digital image of a character to be recognized; step 2: determining a maximum outer closure profile;

and step 3: carrying out sensing area segmentation on the maximum external closed contour image obtained in the step (2) to obtain digital character coordinates;

and 4, step 4: dividing each individual digit according to the acquired coordinate information;

and 5: and identifying according to the digital contour characteristics to obtain a real number.

Preferably, the specific method for preprocessing the target digital image of the character to be recognized is as follows:

and carrying out graying processing, self-adaptive median filtering, graying processing and binarization processing to obtain a binarization image.

Preferably, the method for obtaining the coordinates of the digital characters by segmenting the sensitive region of the maximum external closed contour image obtained in step 2 comprises the following steps:

using the pixel distribution of the binary image as a projection distribution image;

and respectively carrying out vertical segmentation and horizontal segmentation to obtain an OCR image by carrying out horizontal and vertical segmentation by taking the horizontal and vertical axis coordinates of the first number and the horizontal and vertical axis critical point coordinates of the last number as boundaries, and storing the initial coordinates and the end coordinates of the two-dimensional image of each digital character.

Preferably, the specific method for dividing each individual number according to the acquired coordinate information is as follows:

and performing vertical cutting and horizontal cutting on the image according to the two-dimensional image starting coordinate and the two-dimensional image end coordinate of each digital character to obtain each single number.

Compared with the prior art, the invention has the following advantages:

(1) the method does not need to collect and train a training data set, is simple to operate, has simple and quick operation processing and low requirement on environment, and is convenient for measuring digital characters in various occasions;

(2) the invention does not need normalization and complex processing of traversing the template library, and has simple processing, smaller data dimension and higher identification efficiency.

The present invention is described in further detail below with reference to the attached drawing figures.

Drawings

Fig. 1 is an image after the gradation binarization processing according to the present invention.

Fig. 2 is a processing effect diagram of a layer a of a sensing region in an image obtained by contour processing and filled with white pixels according to the invention.

Fig. 3 is a diagram illustrating the effect of the process of copying the a layer of the present invention into the original image to obtain the perceived area.

FIG. 4 is a diagram illustrating the processing effect of the present invention after horizontal and vertical segmentation.

FIG. 5 is an image of the present invention after each character is independently segmented and outlined.

Fig. 6 is a flow chart of the present invention.

Detailed Description

A digital character recognition method based on contour includes the following steps:

step 1: the method for preprocessing the target digital image of the character to be recognized comprises the following steps:

and respectively carrying out conventional image processing on the target digital image of the character to be recognized, and respectively carrying out graying processing, adaptive median filtering, graying processing and binarization processing to obtain a binarized image.

Step 2: determining the maximum external closed contour by the specific method:

and finding the maximum outer closed contour of the image by using the contour characteristic of the image. Probably due to some special shooting conditions, the digital display screen cannot be completely imaged in the image, and white pixel points of 4 pixels are respectively amplified and expanded on the upper, lower, left, right and periphery of the image to obtain the maximum closed external contour information; firstly, a black A layer with the same size as the original image is newly built, and then the space inside the maximum outline is filled with white pixels on the A layer. And then, the original image is superposed on the layer A by using an image copying function, so that all pixel point information in the maximum outline can be obtained, and the interference of the surrounding disordered point image information is removed.

And step 3: and (3) carrying out region-of-interest segmentation on the maximum external closed contour image obtained in the step (2). I.e. the extraction of the smallest area containing all the regions of interest and a single number. The specific method comprises the following steps: generating a black-white pixel distribution characteristic histogram by using a binary image, obtaining the coordinates of the critical points of the horizontal and vertical axes of the first digit and the last digit according to the identification conditions such as interval transformation characteristics and change rate in the pixel distribution histogram, and then respectively performing vertical segmentation and horizontal segmentation on the original image by using the coordinates to obtain a large area of the interested digital image containing all the digits and the initial coordinates and the end coordinates of the two-dimensional image of each digital character.

And 4, step 4: and dividing each individual digit according to the coordinate information obtained by dividing. The specific method comprises the following steps:

and (3) performing vertical cutting and horizontal cutting on the image by mainly utilizing the four point coordinates of the enclosing moment of the single digital character obtained in the step (3) and sequentially storing the images in a function container for standby. The method is characterized in that some numbers with decimal points are displayed under specific conditions, right small-angle area scanning of each number needs to be added, and accurate judgment of the decimal points is accurately realized by combining circle detection, small outline detection and the like.

And 5: and identifying according to the digital contour characteristics to obtain a real number. The specific method comprises the following steps:

the first small right angle between the numbers may have an effect on the determination of the profile, which is first filtered. And after screening, judging the number according to the characteristics of the contour. The number of contours is 1, 2, 3, 5, 7, the number of contours is 2 is 0, 4, 6, 9, and the number of contours is only 8, which is 3. For the figure 1 of the outline, the figure 1 is determined by the aspect ratio of the digital pixel, the figures 2, 3 and 7 are distinguished by the distribution characteristics (calculating the number of white pixel points in a specific row and column range) of the pixel and the gravity center distribution of the outline, and the figure 3 is judged by approximate symmetrical distribution and the gravity center characteristic; for a figure with a contour of 2: firstly, detecting whether a triangular area containing angular points is included to judge a number 4, and then judging numbers 0, 6 and 9 by utilizing the gravity center of the contour and the distribution characteristics of pixels; the number of contours can be directly judged to be 3. Finally, the identified digit combination and the decimal point which may be contained are processed and output.

Claims (4)

1. A digital character recognition method based on contour is characterized by comprising the following specific steps:

step 1: preprocessing a target digital image of a character to be recognized;

step 2: determining a maximum outer closure profile;

and step 3: carrying out sensing area segmentation on the maximum external closed contour image obtained in the step (2) to obtain digital character coordinates;

and 4, step 4: dividing each individual digit according to the acquired coordinate information;

and 5: and identifying according to the digital contour characteristics to obtain a real number.

2. The contour-based digital character recognition method of claim 1, wherein the specific method for preprocessing the target digital image of the character to be recognized is as follows:

and carrying out graying processing, self-adaptive median filtering, graying processing and binarization processing to obtain a binarization image.

3. The contour-based digital character recognition method according to claim 1, wherein the method for obtaining coordinates of digital characters by segmenting the sensitive region of the maximum external closed contour image obtained in step 2 comprises:

using the pixel distribution of the binary image as a projection distribution image;

and respectively carrying out vertical segmentation and horizontal segmentation on the projection distribution image, acquiring an OCR image taking the horizontal and vertical axis coordinates of the first digit and the horizontal and vertical axis coordinates of the last digit as a critical point boundary, and storing the initial coordinates and the end coordinates of the two-dimensional image of each digital character.

4. The contour-based numeric character recognition method of claim 1, wherein the specific method of segmenting each individual digit according to the acquired coordinate information is:

and performing vertical cutting and horizontal cutting on the image according to the two-dimensional image starting coordinate and the two-dimensional image end coordinate of each digital character to obtain each single number.

Images