KR100555883B1 - An Effective Bar Code Detection Method using Mathematical Algorithm - Google Patents

Abstract

The present invention relates to a method of efficiently extracting image images of a barcode by obtaining and labeling geometric features of the barcode using a mathematical algorithm. More specifically, the present invention provides a method for determining a direction of an image pixel of a barcode using four line operators, and b) obtaining a cumulative histogram of pixels of each direction for each block. And c) determining a block of the barcode region as a block having the largest difference between the maximum value and the minimum value in the cumulative histogram, and d) a threshold using the largest difference between the maximum value and the minimum value in the cumulative histogram. Obtaining a value, e) classifying a block whose difference between the maximum value and the minimum value of the histogram for each block is larger than a threshold value as a block having a geometrical feature of a barcode, and f) a barcode determined as a block having a geometrical feature Binarizing the image of the barcode; and g) labeling the image of the binary barcode to determine a candidate block of the barcode; and h) A method for quickly and accurately extracting a video image of a barcode consisting of detecting a barcode by obtaining a slope and a center point of the barcode using pixels of the candidate blocks and drawing a straight line perpendicular to the barcode and passing through the center of the barcode. There is this.

Barcode, Line Operator, Binarization, Labeling

Description

An Effective Bar Code Detection Method Using Mathematical Algorithm

1. Cumulative histogram of pixels along the 0 °, 45 °, 90 °, and 135 ° directions of (a) a block in the barcode region and (b) a block outside the barcode region according to the present invention.

2. Line operators in (a) 0 ° direction, (b) 45 ° direction, (c) 90 ° direction, and (d) 135 ° direction according to the present invention.

Fig. 3. (a) Experimental image according to the present invention and the resulting image by applying the line operator in (b) 0 °, (c) 45 °, and (d) 90 ° to this image

4. A block of a barcode area having a maximum difference between a maximum histogram value and a minimum histogram value according to the present invention.

5. (a) Binarized image obtained by using a threshold value and (b) Candidate blocks of a barcode region according to the present invention

6. A straight line passing through a bar code obtained using (a) a block including a bar code and (b) candidate blocks in the bar code area according to the present invention.

Fig. 7. Barcode detected by the proposed algorithm according to the present invention

8. (a) Experimental image 1 according to the present invention and (b) a straight line passing over the barcode by the proposed method and (c) the detected barcode

9. (a) Experimental image 2 according to the present invention and (b) the straight line passing over the barcode by the proposed method and (c) the detected barcode

The present invention relates to a method of efficiently extracting image images of a barcode by obtaining and labeling geometric features of the barcode using a mathematical algorithm.

In the prior art, a method of finding a barcode as an information storage medium having a geometric characteristic in which a straight line of a certain size is arranged in a certain area is used as a method of using a magnitude of a slope of a pixel or a method of using neural networks. And a method of using a HT (hough transform), but the method of using the magnitude of the gradient of the pixel is difficult to determine the threshold value of the magnitude of the gradient determined as the pixel of the barcode region, and the magnitude of the gradient is larger than the threshold value. The neural network or clustering is used to detect only the pixels in the barcode region among the pixels, which is complicated and takes a long time to perform the algorithm. Detecting barcodes requires a learning process. Therefore, it is difficult to implement in real time, and the method of detecting a barcode using HT is difficult in real time due to the process of performing HT, and also distinguishes between an edge represented by a barcode and another edge among the edge components found in HT. There was a problem with this difficulty.

The technical problem to be achieved by the present invention is to recognize the problems of the prior art as described above, and to extract the image image of the barcode more quickly and accurately by using four line operators (line operator) has the image pixels of the barcode Obtain a direction and obtain a cumulative histogram of the pixels for each block for each block, and determine the block of the bar code area that has the largest difference between the maximum value and the minimum value in the cumulative histogram and use the largest difference value. By determining a threshold value, the block whose difference between the maximum value and the minimum value of the histogram is larger than the threshold value for each block is determined / classified as the object of the block having the geometrical feature of the barcode, and the image of the barcode determined as the object of the block is binarized. Candidates for Barcodes by Labeling Images of Binary Barcodes It is to determine the lock and provide a method of the using the pixels of the candidate block, obtain the slope and center point of the bar code the bar code and the vertical and drawing a straight line passing through the center of the bar code by detecting the bar code efficiently extract a video image of a bar code.

The present invention relates to a method of efficiently extracting image images of a barcode by obtaining and labeling geometric features of the barcode using a mathematical algorithm. First, the algorithm for quickly and accurately reading an image of a barcode according to the present invention will be described in detail. The histogram of the direction of an image pixel is obtained by using a line operator considering four directions in an image of a barcode read with a laser beam. . The difference between the maximum histogram value and the minimum histogram value for each direction is determined for each block, and the block having the largest difference is determined as the block of the barcode region, and the threshold value is obtained using the difference. After binarizing the image using a threshold value, labeling is performed to determine blocks having the largest number of connected components as candidate blocks of the barcode region. After obtaining the slope and the center point of the barcode using all the pixels of the candidate blocks, a straight line is detected by connecting points perpendicular to the barcode and passing through the center of the barcode. When the vertical line passes by, pixel values are sequentially obtained to grasp the information of the barcode. In more detail, the directionality of the pixels and the block determination of the barcode region are described in detail. The barcode is formed by combining a bar and a space in a specific form and codes the letters, numbers, and symbols optically for easy reading. In order to clearly distinguish the beginning and end of, there is a space before and after the first character. The image of the barcode has more edge components and edges than any other part because the region with the barcode in the image containing the barcode has a geometric characteristic in which a straight line of uniform size due to the combination of the bar and the space of the barcode is arranged inside the region. The component has a constant direction, and the image at the edge is concentrated with pixels having the same direction as other parts. Accordingly, the barcode region can be found by finding blocks having the most uniform pixel direction in blocks. To determine the direction of each pixel, calculate the size of the direction using four line operators that consider four directions (0 °, 45 °, 90 °, and 135 °), and then select the direction with the largest value. Determine the direction of the pixel. Therefore, in any pixel, the direction θ (x, y) that the pixel has

(1)

(One)

는 라인 연산자,

는 고려된 네 개의 방향을 나타내며

에 대해 각각

을 나타낸다. Is determined. From here

Is a line operator,

Represents the four directions considered

For each

Indicates.

After finding the direction of the pixel, the barcode area is searched in blocks. FIG. 1 illustrates a cumulative histogram of pixels in each direction of a specific block after obtaining directions of the pixels. As shown in FIG. 1, when the image is divided into blocks, a block including a background or a character, that is, a block outside the barcode area is composed of pixels having various directions, and thus the cumulative histogram value of the number of pixels in each direction. This is similar. However, since the block including the barcode includes many edges in the same direction, the histogram value in a specific direction is relatively larger than the histogram value in another direction according to the straight line direction of the barcode in the histogram in which the number of pixels in each direction is accumulated. Therefore, the block of the bar code area has a large difference between the maximum and minimum histogram values for each direction. Therefore, after obtaining a histogram obtained by accumulating the number of pixels in four directions, the maximum and minimum histogram values for each direction are determined. Find the difference. The difference between the maximum and minimum histogram values for each direction

와

는 수행되고 있는 블록

에서 방향

와

를 가지는 화소의 누적 히스토그램 값을 나타낸다. 그리고

에 대해

및

이며

와

는 동일하지 않다. 이렇게 구해진 값이 가장 큰 블록은 같은 방향의 화소를 가장 많이 포함하고 있으므로 동일한 방향의 직선을 가지는 동일한 방향의 에지들로 구성된 바코드의 기하학적인 특징을 가장 잘 나타내는 바코드 영역의 블록이다. 레이블링에 의한 바코드 영역 검출에 대하여 구체적으로 살펴보면, 바코드의 정보 해석을 위해서는 바코드의 중심을 지나가는 직선이 필요하다. 각 블록에 대해 네 개의 방향에 대한 화소의 개수를 누적한 히스토그램에서 각 방향에 대한 최대 히스토그램 값과 최소 히스토그램 값과의 차이가 가장 큰 블록을 이용하여 바코드의 기울기와 중심좌표를 구할 수 있으나, 바코드의 정확한 기울기와 중심점을 찾기 위해서는 바코드 영역의 더 많은 후보 블록들이 필요하다. 따라서 바코드 영역의 블록들을 찾기 위해 최대 히스토그램 값과 최소 히스토그램 값의 차 중에서 가장 큰 값을 이용하여 문턱 값을 결정한다. 이때 문턱 값은 Same as From here

Wow

Is the block being executed

Direction

Wow

A cumulative histogram value of a pixel having And

About

And

And

Wow

Is not the same. The block having the largest value is the block of the barcode area that best represents the geometrical characteristics of the barcode composed of the edges of the same direction having straight lines of the same direction since the block contains the largest number of pixels in the same direction. Looking at the bar code area detection by labeling in detail, a straight line passing through the center of the bar code is required to interpret the bar code information. In the histogram that accumulates the number of pixels in four directions for each block, the slope and the center coordinate of the barcode can be obtained by using the block having the largest difference between the maximum and minimum histogram values for each direction. More candidate blocks of the barcode area are needed to find the exact slope and center point of. Therefore, the threshold value is determined using the largest value among the difference between the maximum histogram value and the minimum histogram value to find the blocks of the barcode region. The threshold is

는 임의의 상수이다. 문턱 값을 이용하여 각 블록의 각 방향에 대한 최대 히스토그램 값과 최소 히스토그램 값의 차가 문턱 값보다 큰 블록 을 바코드의 기하학적 특징을 갖는 바코드를 포함하는 블록으로 확정/분류하고 확정된 블록을 대상으로 바코드 이미지영상을 이진화 한다. 바코드의 기하학적 특징을 갖는 블록들은 바코드 영역 밖에도 존재하지만 주로 바코드 영역의 다른 블록들과 인접하여 많이 분포한다. 그러므로 바코드 영역의 후보 블록들을 추출하기 위하여 이진화 한 이미지영상에 대하여 레이블링을 행하여 연결 성분이 가장 많은 블록들을 바코드 영역의 후보 블록으로 결정한다. 그리고 바코드를 포함하는 바코드 영역의 블록으로 결정된 블록들이 최대 히스토그램 값과 최소 히스토그램 값의 차이가 가장 큰 블록을 포함하고 있는지를 확인함으로써 검출된 후보 블록들이 바코드 영역의 블록들임을 확인한다. 바코드 기울기와 중심점 계산 및 바코드 검출에 대하여 구체적으로 살펴보면, 바코드 영역의 후보 블록들을 이용함으로써 각 방향에 대한 최대 히스토그램 값과 최소 히스토그램 값의 차이가 가장 큰 블록 하나만을 이용할 때보다 바코드의 기울기와 중심점을 좀더 정확하게 구할 수 있다. 바코드 영역의 후보 블록들의 모든 화소에 대해 수평방향으로의 변화량과 수직방향으로의 변화량을 구하여 바코드의 기울기를 구하고 후보 블록들의 중심점들의 평균을 구하여 바코드의 중심점을 구한다. 이때 바코드의 중심점을 지나는 직선의 기울기는 바코드를 구성하는 직선의 기울기와 수직이므로 직선의 기울기 는 식 (4), 식 (5), 및 식 (6)을 이용하여 구한다., Where

Is any constant. By using the threshold value, the block whose difference between the maximum histogram value and the minimum histogram value in each direction of each block is larger than the threshold value is determined / classified as a block including the barcode having the geometrical characteristics of the barcode, and the barcode is determined based on the determined block. Binarize the image. Blocks having the geometrical characteristics of a barcode exist outside the barcode region but are mainly distributed adjacent to other blocks of the barcode region. Therefore, labeling is performed on the binarized image image to extract candidate blocks of the barcode region, and the blocks having the most connection components are determined as candidate blocks of the barcode region. In addition, it is confirmed that the detected candidate blocks are blocks of the barcode region by checking whether the blocks determined as the blocks of the barcode region including the barcode include the blocks having the largest difference between the maximum histogram value and the minimum histogram value. Looking at the barcode slope and center point calculation and barcode detection in detail, the candidate blocks of the barcode area are used to determine the slope and the center point of the barcode compared to using only one block having the largest difference between the maximum and minimum histogram values in each direction. You can get it more accurately. The slope of the barcode is obtained by calculating the change in the horizontal direction and the change in the vertical direction for all the pixels of the candidate blocks in the barcode area, and the center point of the barcode is obtained by calculating the average of the center points of the candidate blocks. At this time, since the slope of the straight line passing through the center point of the barcode is perpendicular to the slope of the straight line constituting the barcode, the slope of the straight line is obtained using equations (4), (5), and (6).

는 추출된 바코드 영역의 후보블록들을,

는 수평과 수직 방향에 대한 필터를 각각 나타낸다. 이렇게 구한 기울기와 중심점을 이용하여 바코드와 수직이고 바코드의 중심을 지나가는 직선을 그을 수 있다. 또한 바코드 영역의 후보블록들을 이용하여 4개의 최외각 점을 구하여 바코드가 포함된 직사각형의 영역을 만든 후, 기울기와 중심점을 이용하여 직사각형을 회전함으로써 바코드를 검출한다. 수직선이 지나갈 때의 화소 값을 순차적으로 획득함으로써 바코드가 가지고 있는 정보를 파악한다. 즉, 본 발명은 a) 화소의 방향성과 바코드 영역의 블록 결정 단계와, b) 레이블링에 의한 바코드 영역 검출 단계와, c) 바코드 기울기와 중심점 계산 및 바코드 검출 단계를 거치면서 바코드의 이미지영상이 가지고 있는 정보를 신속하고 정확하게 읽어낸다. From here

Candidate blocks of the extracted barcode region,

Denotes filters for the horizontal and vertical directions, respectively. Using the obtained slope and the center point, a straight line passing perpendicular to the barcode and passing through the center of the barcode can be drawn. In addition, four outermost points are obtained using candidate blocks of the barcode area to make a rectangular area including a barcode, and then the barcode is detected by rotating the rectangle using a slope and a center point. By acquiring the pixel values when the vertical line passes, the information possessed by the barcode is identified. That is, the present invention includes the image image of the barcode through a) the pixel orientation and the block determination of the barcode region, b) the barcode region detection by labeling, and c) the barcode slope and the center point calculation and the barcode detection stage. Read the information quickly and accurately.

[ Example ]

Looking at the embodiment according to the present invention in detail based on the drawings, CODE39 widely used in various fields, KAN used for domestic and foreign product management, UPC used for product management in the United States and Canada, regions other than the United States and Canada Used images of 1024 × 1024 including various barcodes such as EAN used for product management and ISBN used for managing serials including books. It is a video. FIG. 2 illustrates four line operators used to obtain a histogram that accumulates the number of pixels in four directions, such as 45, 90, and 135 °, for each block. In FIG. 3 is a direction of pixels having the largest value, and the direction of each pixel obtained using four line operators is shown in FIG. 3. It can be seen that pixels having the same direction are concentrated in the barcode region as can be seen in the histogram in which the number of pixels in each direction of the block including the barcode of FIG. As shown in FIG. 4, a block having a maximum difference between the maximum histogram value and the minimum histogram value for the detected direction using the geometric features of the barcode area may be included in the barcode area in the image including the barcode. . The image is binarized using the threshold value to detect more candidate blocks including the barcode to find the exact slope and the center coordinate of the straight line passing through the center of the barcode. 5 (a) shows a binarized image, in which blocks having a difference between a maximum histogram value and a minimum histogram value larger than a threshold value are distributed in a barcode region. Labeling is performed on the binarized image to determine the labeled blocks having the largest number of connected components as candidate blocks including barcodes. 5B shows candidate blocks of the barcode area after labeling. It can be seen that the candidate blocks are in the barcode region. FIG. 6 shows a straight line passing through the center of the barcode obtained using only the block having the largest difference between the maximum histogram value and the minimum histogram value in each direction, and a straight line passing through the center of the barcode obtained using candidate blocks of the barcode area. As shown in FIG. 6, a straight line may be obtained using only blocks having a maximum difference between the maximum histogram value and the minimum histogram value in each direction, but the straight line obtained using candidate blocks in the barcode area passes through the center of the barcode more accurately. You can see the sense. In addition, it can be seen from FIG. 7 that barcode detection is possible by the proposed algorithm according to the present invention. Table 1 shows the execution time and the result of barcode detection for 80 images including various barcodes and backgrounds such as CODE39, KAN, UPC, and ISPN. It was run on a personal PC with a Pentium 3 CPU 866GHz and 256MB of DRAM. The average detection time was 0.89 seconds and the barcode detection rate was 95%. However, since many blocks have a value larger than a threshold value in an image having a complex background, a barcode detection time may be lengthened in some block-by-block labeling process, and a barcode may be difficult to detect.

Table 1. Barcode Detection Time and Detection Rate

 Barcode detection time   Barcode Detection Rate [%]  Detection time [s]  Detection time [s]  Average detection time [s]  0.622  1.42  0.89  95

In order to confirm the reliability of the proposed algorithm according to the present invention, various bar code detection experiments were performed with various bar code angles and background patterns. 9 is shown. As a result of applying the proposed method to the image including the barcode rotated at an arbitrary background and angle, we could draw a straight line passing through the barcode and detect the barcode.

The present invention obtains the direction of the image pixels of the barcode using a mathematical algorithm and obtains a cumulative histogram of the pixels in each direction for each block to determine the barcode having the largest difference between the maximum value and the minimum value in the cumulative histogram. Determining the block of the area and obtaining a threshold value using the largest difference value, classifies the block having the geometrical feature of the barcode as the object of the block with the difference between the maximum value and the minimum value of the histogram greater than the threshold value for each block. Binary image of the determined barcode by labeling the image of the binary barcode to determine the candidate block of the barcode and using the pixels of the candidate blocks to obtain the slope and the center point of the barcode to perpendicular to the barcode and to pass through the center of the barcode Draw a straight line to detect the bar code It has the effect of extracting video images quickly and accurately.

Claims (2)

In the method of extracting the image image of the barcode by obtaining the geometric features of the barcode using a mathematical algorithm and labeling, a) using the line operator to determine the direction of the image pixels of the barcode; b) obtaining a cumulative histogram of pixels for each direction; c) determining a block of the barcode region as a block having a largest difference between a maximum value and a minimum value in the cumulative histogram; d) obtaining a threshold value using the largest and smallest values in the cumulative histogram; e) classifying a block having a geometric characteristic of a barcode as a block having a difference between a maximum value and a minimum value of a histogram greater than the threshold value for each block; f) binarizing an image of a barcode determined for the block having the geometrical feature; g) labeling an image of the binary barcode to determine a candidate block of the barcode; h) extracting the image image of the barcode quickly and accurately by obtaining a slope and a center point of the barcode using the pixels of the candidate blocks and drawing a straight line perpendicular to the barcode and passing through the center of the barcode. How to. The method according to claim 1, Method of extracting a video image of a barcode quickly and accurately, characterized in that the four lines operator is detected by the direction of the image pixels of the barcode in step a) 0 °, 45 °, 90 ° and 135 °.

Images