JPS62192883A - Character string extracting system - Google Patents

Abstract

PURPOSE:To extract one row of characters even when only about two rows of the characters are printed by obtaining the center of the gravity of respective characters in an image, examining the linearity between these mutual centers of the gravity and extracting a character string. CONSTITUTION:A printed character, etc., are photographed by a photographing camera 1 and a photographing image is inputted to an image processing device 4. By starting a program for estimation stored at an FD in an FD driving 6 with the operation executed by a CRT 5 and a keyboard, the recognition of the character is executed. In the image processing device 4 an image input part 12, and a segmentation part 13 to designate respectively the number to each black and white area of a binary image are made into a hardware circuit, and is operated by a CPU 14. The center of gravity of respective characters in the image is obtained by the image processing device 4, the linearity between the mutual centers of the gravity is examined and the character string is extracted.

Description

[Detailed Description of the Invention] [Industrial Application Field] The present invention relates to an image processing method in a mounted board appearance inspection device, and particularly relates to a character string suitable for recognizing characters printed on ICs, etc. This concerns the extraction method.

[Conventional technology]

As a conventional character string extraction method, Japanese Patent Application Laid-Open No. 57-18287
As described in Japanese Patent No. 3, it is known that a virtual frame is first provided and a rough character range is determined using signals from the lower line. This method is effective when there are several lines of text, such as written documents, but the lower line signal cannot be used for documents where each line of text is about two lines, such as IC marks. It has become a thing.

[Problem that the invention seeks to solve]

As described above, in conventional character string extraction, a rough character range is determined by a method of removing signals from other lines. However, since this method uses signals from other lines to extract one line of text, it is inappropriate as a method for extracting text from ICs that have only about two lines printed. It has become.

An object of the present invention is to provide a character string extraction method that can extract one line of characters without using information on other lines and even when only about two lines of characters are printed.

[Means for solving problems]

It is difficult to extract character strings on the screen using only information from images. Therefore, the idea is to extract a character string by finding the center of gravity of each character in an image and examining the linearity between these centers of gravity.

[Effect]

This method takes advantage of the fact that the centers of gravity of each character in a character string are approximately on a straight line. That is, for example, the inclinations between the centroid of the character located at the left end of the screen and the centroids of all other characters are determined in order, and then the inclinations are classified according to the magnitude of these inclinations, and the groups having approximately the same inclination are determined. If the number of elements is greater than or equal to a certain value, this set of elements can be recognized as one character string. In addition, by creating a straight line connecting the maximum coordinates of the outline of the characters located at the left and right ends of the character string, and creating a straight line connecting the minimum coordinates of the outline in the same way, you can create the upper and lower frames for that character string. Not only can the character string be generated, but also the slope of the character string can be known by linearly approximating the center of gravity using the most/I) square method.

〔Example〕

The present invention will be explained below with reference to FIGS. 1 to 7.

First, the configuration of an example of the mounted board appearance inspection apparatus according to the present invention will be described. FIG. 2 shows the configuration. This device includes an imaging camera 1, a magnifying lens 9
, XY table 10. an illumination light source 2 that uniformly illuminates the IC 8 on the mounting board 11 placed on the XY table 10;
A monitor 3 that displays the captured original image or a binarized image, an image processing device 4 that processes the image based on data from the imaging camera 1, and a CRT 5 that displays operation details.
, a floppy disk (FD) that stores and stores data.
It consists of a drive 6 and a keyboard 7 for inputting various commands and data.

With the above configuration, the characters printed on the IC 8 can be recognized. 6 That is, the characters printed on the IC 8 are imaged by the imaging camera 1 through the magnifying lens 9 and then imaged by the image processing device 4. Input the image and display it on the CRT5
The characters on the IC 8 are recognized by activating an evaluation program stored in the FD in the FD drive 6 by operating the keyboard 7. During this time, the original image or the binarized image is displayed on the monitor 3.

Figure 3 shows the hardware configuration of the image processing device.
This is shown together with 10. According to this, the image processing device 4 has an image input section 12 and a segmentation section 13 that numbers each black and white area of a binary image as a hardware circuit, and is operated by a 16-bit general-purpose CPU (central processing unit) 14. It is supposed to be done.

FIG. 4 shows a binary image of the characters printed on the IC 8 displayed on the monitor 3. In the display screen 15, there is a binary image corresponding to the characters in the background 18. Binary image pattern 17 corresponding to pattern 16 and IC pin
It is scattered.

Now, the processing procedure for character string extraction according to the present invention will be explained. FIG. 1 shows the processing procedure. In this case, the input image is first binarized as shown in Figure 4, and segmentation (labeling) processing is performed for each black and white area to number the appearance class in the scanning direction. In the process, the top coordinates and area value of each area are calculated at the same time. This allows the size and position of each area to be known. More specifically, for each black and white area, attention is paid to the continuity of pixels, and as a result, a number is assigned to the isolated area in the scanning direction appearance class. ■,■ in the example shown in FIG.

The numbers such as ■... indicate the appearance of the isolated area in the scanning direction. ■, 'ノ due to the continuity of pixels and the first pixel that first changes from white to black or from black to white in the isolated area.

■... It is possible to attach numbers such as... Note that the area value in this case is the number of pixels forming the isolated area.

Now, next, a determination is made based on the area value in order to extract only isolated areas that have an area within a certain range (equivalent to the character area). This removes backgrounds with large areas, small noises, and the like.

The center of gravity is calculated for the isolated area that is tentatively determined to correspond to a character in this area determination process, and the result of calculating the center of gravity is shown in FIG. This mark indicates the center of gravity of each area. For example, for an isolated area corresponding to the character rRJ, the center of gravity as the centroid is calculated from the pixels composing this area and their coordinates. Next, since the linearity of the centroids obtained in this way is examined in order from the leftmost one, numbers are assigned in descending order of the X coordinate. The numbered example is
As shown in the figure. The center of gravity of each area is 01-014 from the left end.
numbers are attached as shown in the figure. After this, the inclinations from the leftmost center of gravity, that is, G1, to each of the other centers of gravity are determined and Hough transform is performed. The Hough transform here is a general Hough transform with the addition of passing through one point Ge. Ge
The points will be substituted in order starting from G1. Therefore, first, as shown in Fig. 6, set Ge=01, and then
Find the slope al of the straight line connecting 2. Similarly, G1 and G
3. Gl and G4. ..., Gl and 014 (after calculating the slopes of the straight lines connecting all centers of gravity except 31 and 01, the slopes are classified according to the predetermined number of divisions, and a frequency distribution table is created. , the slope with the maximum degree is calculated.The slope amaχ with the maximum degree and the slopes amaχ before and after it.By selecting the straight line group with ±α, 1
A linear point sequence Ll (Gl, G
3. G4. G7゜GIO, G12. G14) is extracted. At this time, the center of gravity Gn selected as the Ge point
If it is not on the desired straight line, the number of extracted point sequences is not sufficient, so a determination is made based on the number of point sequences, that is, the length. If the length is less than a certain standard value, the next point, that is, G
The point ``Ge1'' is set as a new Ge, and the Hough transform is performed again.

If a straight line exceeding the standard value is extracted in this way,
The point sequence L1 is linearly approximated by the least squares method (the slope of the 2° character string is calculated. Furthermore, as shown in Fig. 7, the right end with the maximum and minimum points G14 and Gl in the point sequence L1 ,
Left end)] U' character 21. Y of the outline of the character string with 20 nitsui
Point 22.23 with maximum coordinates and point 24.2 with minimum coordinates
5 is found l/ By calculating the straight line connecting 6-2 points 22.23, the lower frame 26 of the character string is found, and in the same way, from the point 24.25, the upper frame 27 is found. This makes it possible to extract character strings.

〔Effect of the invention〕

As explained above, according to the present invention, the target character string can be extracted by using the center of gravity of the character. This has the effect of automatically extracting character strings even if there are variations.

[Brief explanation of drawings]

FIG. 1 is a diagram showing a flow of a processing procedure for character string extraction according to the present invention, FIG. 2 is a diagram showing an external configuration of an example of a mounted board appearance inspection apparatus according to the present invention, and FIG. is a diagram showing the hardware configuration of the image processing device in this configuration, FIG. 4 is a diagram showing an example of a binary image of characters printed on an IC, and FIG. 5 is a diagram showing the result of calculating the center of gravity for that image. FIG. 6 is a diagram for explaining the Hough transform for the centroid calculation result, and FIG. 7 is a diagram for explaining the character string extraction process according to the present invention. 1... Imaging camera, 4... Image processing device, 8...
IC112... Image input unit, 13... Segmentation unit. Agent buoy” buried soil Masami Akimoto Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7

Claims (1)

[Claims] 1. Using a character string extraction method that images characters and then extracts a string of characters through image processing, the center of gravity as the centroid of each imaged character is determined, and based on the inclination between the centers of gravity, the character string is roughly drawn on a straight line. By extracting the existing center of gravity,
A character string extraction method characterized by extracting character strings.