JPH03204792A - Method and device for detecting marking - Google Patents

Abstract

PURPOSE:To automatize the detection of marking by performing character recognition by binarizing image pickup signals obtained with image pickup means in plural inspection areas, respectively, limiting a specific marking member based on the change point of a binarization signal, and performing the binarization processing of a marking image in the marking member. CONSTITUTION:When high change points h1-h4 are detected in all the set plural inspection areas out of plural marking surfaces in a photographic image, a part enclosed with each of the change points h1-h4 is specified as one marking surface. Information for character recognition can be obtained by applying the binarization processing and collateral processing at need are applied only to the specific marking surface. In such a way, one of the marking surfaces can be specified even when plural marking surfaces are included in the photographic image, and the detection of marking can be automatized.

Description

[Detailed Description of the Invention] [Industrial Application Field] The present invention provides a detection method for selecting one stamped material from a plurality of stamped materials on which stamped images are formed and detecting the stamped stamp. In particular, the present invention relates to a method and apparatus for detecting stamps, in which a marking stamped on a comparatively small primary rolled steel material is photographed with a television camera, and the content thereof is recognized by image processing.

[Prior art] For example, relatively small primary rolled steel materials (e.g. billet,
Bloom) is marked with numbers, letters, symbols, etc., and the markings are automatically read and recognized and used for manufacturing, processing, inventory management, etc.

This marking is formed by pressing a marking member (letter mold) on which numbers, letters, symbols, etc. are formed in a horizontally reversed manner onto the end face of the rolled steel material. is formed in the shape of a groove.

One method for this is to obtain a stamped image formed on a rolled steel material (hereinafter referred to as the stamping material) as a multi-value original image signal using an imaging device such as a television camera, and convert this into a binary signal. There is a method that recognizes characters using the final image. An example of image processing for obtaining a final image for single character recognition is disclosed in Japanese Patent Application Laid-Open No. 61-80372.

The technology shown here uses one threshold value for the entire image in the binarization process that converts a multivalued original image into a binary final image, and this threshold value is It is characterized by updating and determining the number of low brightness areas in the area that matches the set value.

[Problems to be Solved by the Invention] However, in the conventional stamp detection device as described above, when a plurality of stamp materials are conveyed to the imaging area at the same time,
Since it is not possible to specify one of them, marking detection cannot be performed.

Furthermore, since the entire image is binarized using only one threshold value, erroneous recognition may occur.

In other words, the engraving surface is not flat and often has rust or contamination, and the degree of occurrence and formation conditions are not constant. Therefore, dirt etc. can be detected along with the engraving, leading to misrecognition or recognition. It will lead to incompetence.

This invention was made in order to solve the problems of the prior art, and even if a plurality of stamping materials arrive at the image capturing area at the same time, it is possible to identify one of them and detect the stamping stamp. An object of the present invention is to provide a method and device for detecting a stamp.

[Means for Solving the Problems] In order to achieve the above object, the present invention obtains an imaging signal by simultaneously imaging the stamping surfaces of a plurality of stamping materials on which stamped images are formed, and A plurality of inspection areas are set radially from the center of the imaged area, each of the imaging signals in this inspection area is binarized, and the part where the change point appears most strongly is identified as the stamping surface of the stamping material to be inspected. , a method of performing binarization processing for character recognition only on the identified stamped surface, and an imaging means for capturing an image by illuminating the stamping surfaces of a plurality of stamping materials on which stamped images are formed from an oblique direction.

A plurality of inspection areas are set radially from the center of the imaging area obtained by the imaging means, and means for binarizing each imaging signal obtained by the imaging means in the inspection area; This device is equipped with a character recognition means that limits a specific stamped material based on the change point of the binary signal, and performs character recognition by binarizing the stamped image on the stamped material.

[Operation] By configuring as above, when a large change point is detected in any of the plurality of set inspection areas among the plurality of stamped surfaces in the photographed image, each change point is surrounded. Identify the part as one stamped surface. Information for character recognition can be obtained by performing binarization processing and additional processing as necessary on only the identified stamped surface. Therefore, even if a photographed image includes a plurality of marking surfaces, one of them can be identified, and marking detection can be automated.

[Example] FIG. 1 is a block diagram showing one embodiment of the present invention.

The imaging device 1 is for photographing the stamping surface of one of the plurality of stamping materials (in this embodiment, a billet is exemplified), and for example, a television camera is used. Imaging device
A stamp detection device 2 is connected to the stamp detection device 2, and character recognition processing units 3-1 to 3-N are connected to the stamp detection device 2.

As shown in FIG. 2, the imaging device l is arranged so that at least one of the stamping surfaces of the plurality of stamping materials 4 is taken as a photographing area, and also has a plurality of A lamp 5 is provided.

In order to reliably extract the stamped image, it is desirable that the density difference (contrast) between the background and the characters be large.

This density difference is determined by ■the irradiation direction of the illumination light, and ■the effective linear length of the character determined by the character shape (here, when considering a straight line parallel to each illumination direction, the length that the straight line and the character overlap) The longest length is defined as the effective straight line length), ■It is related to the stamping method of characters, etc. In order to improve the extraction, it is better to make the characters deeply engraved and to shorten the effective straight line length. It's good to do it.

The stamp detection device 2 includes an end face mask image reproduction unit 6 that identifies only one of the plurality of stamp materials 4 included in the original image signal 11 from the imaging device 1, and a stamp image reproduction unit 6 that performs processing to emphasize only the stamp image. Emphasis processing unit 7, AND circuit 8, marking material orientation detection processing unit 9. Engraving position detection section 10, engraving image separation processing section 11
, and a binarization processing unit 11. ~12-. It consists of

Next, the general operation of the configuration shown in FIG. 1 will be explained.

First, an image of the stamped surface of the stamping material 4 is captured from the direction shown in FIG. 2 by the imaging device 1, and the stamped image is obtained as a multivalued original image signal 13. On the other hand, only the stamp material 4 whose entire stamp surface is within the imaging area is extracted by the end face mask image generation unit 6, and an end face mask image 14 is output.

On the other hand, the original image signal 13 is subjected to a process of emphasizing only the stamped image (images of letters, numbers, symbols, etc.) by the stamped image enhancement processing section 7, and a stamped image enhancement signal 15 is output. Then, the AND circuit 8 generates the end face mask image generator 6.
Coordinates Z (x, y) of the end face mask image 14 specified by
Only the engraved image emphasis signal 16 of the engraved image (letters, numbers, symbols, etc.) included in the image is allowed to pass through the gate.

Next, the direction of the stamp material is detected by the stamp material orientation detection processing section 9, and stamp material orientation information 17 is output.

Further, a stamp image enhancement signal 12 and stamp material orientation information 13
Based on this, the marking position is detected by the marking position detection processing section 8. Based on the stamp position information 18 and the original image signal 13 obtained by the stamp position detection processing section 10, the stamp image separation processing section 11 separates the original image signal 13 for each stamp into N image signals for each stamp. Get 19. This stamp-specific image signal 19 includes only one stamp image among N images. Therefore, the final images 20-□ to 20-H are obtained by performing binarization processing independently for each marking by the binarization processing units 12-1 to 12-8. For these final images 20-8 to 20-H, the character recognition processing unit 3-1
~3-. 4, character recognition processing is executed.

Next, the processing of the end face mask image generation section 6 will be explained with reference to FIGS. 3 to 6.

Here, as shown in Fig. 3, there are three wooden engravings in the photographing area, one in the center, and two wooden engravings on both sides with a predetermined gap. This will be explained using an example. Then, from the center of the imaging area, there is a narrow frame w10 extending vertically and horizontally from the center of the imaging area as an inspection area. who, w3
0. Set w40.

Further, the resolution of the stamp image is, for example, 0.305 x O,
3t) Set the level to 5□/1 pixel, and set the width of the frame W to 1
00 pixels (p i x), and the length of the frame is 256 pixels.

The three stamping materials are bundled in close contact with each other, but since the peripheral edges at both ends are rounded, grooves deeper than the stamps are formed between the adjacent edges and on the outside of the peripheral edges. ing. Therefore, when the inside of the frame is binarized, a change point of a binary level that is clearly different from the marking occurs, and the position of this change point can be regarded as the periphery of the marking surface.

Specifically, first, the above frame wlo, w20゜w3G
, w40 is subjected to binarization processing using discrimination threshold bias (S31). Next, use the binarization results to create a frame w10. w20. When the profile is sequentially measured in the length direction from the center outward for each of w30 and w40, as shown in Figure 4, the periphery of the stamped surface of the stamping material 4 to be extracted and the stamped The profile changes greatly at the position where the gap between the materials (or the open space) comes into contact, and the change point hl (exists in the frame wlO), the change point h2 (exists in the frame who), the change point h3 ( The transition point h4 (existing in the frame w30) is obtained (S32), and the part surrounded by the transition point t- is determined to be the end face of the stamping material 4 (533), and this is determined by image processing. Then, as shown in FIG. 6, it is set to white, and this is output as the end face mask image 14.

Next, details of the stamped image enhancement processing section 7 will be explained.

The stamp formed on the stamp material 4 is captured as an image by photographing a shadow caused by illumination (hereinafter, an image caused by this shadow will be referred to as a stamp image). Therefore, for the imaging device l, as long as the stamped image is clear enough to perform stamping material direction detection processing and stamping position detection processing, stamped image enhancement processing is not necessary. However, in general, stamped images can be clearly obtained. This is difficult, and good image detection cannot be obtained without engraving image enhancement processing.

The processing in the stamped image enhancement processing section 6 can be expressed mathematically as follows.

That is, the density value of the coordinates (x, y) of the image signal 13 is
(x, y), the coordinates of the engraving enhanced image (x, y)
The density value g (x, y) of is given by the following equation.

g (x, y) = Σaixf (x-xi, y yi) where,
xi, yi are x when the original image signal is subjected to image movement processing
, the amount of movement in the y direction, and ai indicates a weighting constant. Furthermore, the constants (N, xi, yi, ai) in the above equation can be determined from the characteristics of the target image signal. By selecting an appropriate constant, only the stamped image can be emphasized without emphasizing the noise.

Next, details of the marking material orientation detection processing section 9 will be explained with reference to FIG.

The engraving on the engraving material 4 (in this embodiment, it consists of a combination of numbers in three stages, the first stage consists of five digits, the second stage consists of two digits, and the third stage consists of one digit) is optimal for the imaging device 1. It doesn't necessarily mean that the location will be sent to you. Then, the vertical and horizontal directions are determined from the image of the engraving image emphasis signal 16, and the image is converted into an image in the correct position.

Specifically, the vertical and horizontal density projections of the stamped image are obtained, and the arrangement of the stamps is point-symmetric with the center of the stamping material 4 as the rotation center, and the arrangement of point pairs is point-symmetric with the center of the stamping material 4 as the rotation center. If not, the maximum deviations (iic, d) or magnitudes of the vertical and horizontal integrated values a, b will differ depending on the orientation of the marking material 4.

By performing analysis based on this information, the orientation of the marking material 4 can be detected. For example, depending on whether the maximum value position revolves at the upper left, upper right, lower right, or lower left of the stamping material 4, it is determined that the rotation angle is at 0°90°, 180°, or 270°. .

Next, details of the processing of the marking position detection processing section 8 will be explained with reference to FIG.

In this process, a specific range of the stamping material 4 (in this embodiment,
The position of the marking is detected by determining the density projection for the density projection calculation range 1 in FIG. 8 and detecting the unevenness of the integrated value.

In FIG. 8, the five numbers at the top of the markings are detected, but by repeating the same process, it is possible to detect the positions of all markings.

Next, details of the operation of the binarization processing units 10-1 to 10- will be described.

The outline of the operation in this configuration is as follows:
The density histogram of the image is calculated, and the image for each marking is binarized based on the binarization threshold determined based on the density histogram. Since the image as it is contains various noises, a noise removal process is performed on the binarized image by a noise removal processing unit (not shown).

This binarization process will be explained using each billet shown in FIG. 7 as an example, with reference to FIGS. 9, 10A, 10B, and 11.

As shown in FIG. 9, the input image is divided so that each character is surrounded by a frame, and frames wl to w8 are assigned. For each of the frames wl to W8, the area within the frame is binarized using discrimination threshold bias (threshold: thl, separation degree bl) to obtain a binarized image wal (S41). Then, the area a of the remaining region is
rl (542), as shown in Figure 11, for points within the frame that have a density equal to or higher than the threshold value th1 (number of pixels)
Binarization processing is performed using the discrimination threshold bias (threshold: th2. degree of separation: b2) to obtain a binarized image wa2 (S43).

Next, the area ar2 of the remaining region is measured (544) and the degree of separation bl and b2 are compared (S45).

In S45, if bl>b2, the process moves to flow ○ in FIG. 10B, and if b2>bl, the process moves to flow -2 in FIG. 1OB.

In flow (2), the area arl is compared with the area threshold arth, and if ar 1 < ar th, the image to be processed thereafter is set as wal (S47). Also, arl
>arth, the image to be processed thereafter is set to wa2 (348). This process of S4B is also executed when flow ○ is selected in 346.

When the processing in S46 or S48 is completed, the feature amount in the image to be processed is measured (S49), and those with an area of less than 30 pixels (pix: pixels) are removed.
50), and further remove those with an area ratio of 0.5 or more (
351). S49 to S51 correspond to noise removal processing.

Next, specific examples of the present invention will be described.

Here, it is assumed that each stamp material 4 is a square billet of approximately 1200□ x tzoo□, and a product number is stamped on the end face.This product number is detected by the stamp detection device 2, and this is detected by the character recognition processing section Characters are recognized by -I to 3-. Further, each constant in the stamped image enhancement processing section 7 was set as follows.

N=5 (xi, yi, ai) = (0,0,4), (4,O,-1).

(-4,O,-1), (0,4,-1).

(0, -4, -1) As a result, when detected using the conventional technology, partial rust and dirt remained in the final image, and the recognition rate was about 76%. However, with the detection according to the present invention, it was confirmed that a clear final image of only the markings could be obtained, and the recognition rate reached 95%.

[Effects of the Invention] As is clear from the above, according to the present invention, the stamped surfaces of a plurality of stamped materials on which stamped images are formed are simultaneously imaged to obtain an imaging signal, and the imaged areas are A plurality of inspection areas are set radially from the center, each of the imaging signals in these inspection areas is binarized, the part where the change point appears most strongly is identified as the stamping surface of the stamping material to be inspected, and the identified stamping material is Binarization processing for character recognition is applied only to faces, so even if a photographed image contains multiple stamped faces, one of them can be identified, and stamp detection can be automated. becomes possible.

[Brief explanation of drawings]

FIG. 1 is a block diagram showing an embodiment of the present invention, FIG. 2 is an explanatory diagram showing the installation and illumination means of the imaging device shown in FIG. 1, and FIG. An explanatory diagram showing the process of identifying one piece, Figure 4 is a change point detection characteristic diagram showing the profile process to identify the stamping material, Figure 5 is a flowchart showing the process of determining the change point, and Figure 6 is a change point diagram. 7 is an explanatory diagram showing the processing of the engraving material orientation detection processing section 9; FIG. 8 is an explanatory diagram showing the processing of the engraving position detection processing section IO; FIG. 10A and 10B are flowcharts showing the processing of the binarization processing section, and FIG. 11 is an explanatory drawing showing threshold setting. In the figure. l: Imaging device 2: Embossing detection device 1, to 3-. : Character recognition processing unit 4: Engraving material 5: Lamp 6: End face mask image generation unit 7: Engraving image enhancement processing unit 8: AND circuit 9: Engraving material orientation detection processing unit 10: Engraving position detection processing unit 11: Engraving image separation Processing units 12-8 to 12-. 1: Binarization processing unit 13: ji (image signal q 14: end face mask image 15° 16: engraving image enhancement signal 17: engraving material orientation information 18: engraving position information 19: image signal by engraving 20-□~2( L: Final image

Claims (2)

[Claims] (1) Imaging surfaces of a plurality of stamped materials on which stamped images are formed are simultaneously imaged to obtain imaging signals, a plurality of inspection areas are set radially from the center of the imaged area, and the inspection is performed. Binarize each of the imaging signals in the region, identify the part where the change point appears most strongly as the stamped surface of the stamped material to be inspected, and perform the binarization process for character recognition only on the identified stamped surface. A method for detecting stamps characterized by: (2) an imaging means for capturing images by illuminating the stamping surfaces of a plurality of stamping materials on which stamped images have been formed from an oblique direction; means for setting an inspection area and binarizing each imaging signal obtained by the imaging means in this inspection area;
Character recognition means is provided, which limits a specific stamped material based on the change point of the binary signal obtained by this means, and performs character recognition by binarizing the stamped print image on the stamped material. A stamp inspection device characterized by: