JPH03218447A - Picture data converting method - Google Patents

Abstract

PURPOSE:To obtain picture data emphasizing a linear fault part by comparing luminance distribution pattern data in a direction to cross the reference linear fault part with source picture data, calculating the degree of coincidence and defining it as new picture data. CONSTITUTION:The source picture data obtained by picking the image of a work W surface are extracted by a cross filter 21 and the degree of coincident with the pattern data is calculated. Based on this degree of coincidence, the picture data are calculated. Since these pattern data are prepared from the luminance data obtained by picking up the image of the linear reference fault part, when the linear fault such as cruck, etc., is generated on the surface of the work W, the part is brightly projected and displayed on a CRT 13 while emphasizing the linear fault part. From the shape of the cross filter 21, the fault part is emphasized and displayed even when it is formed in any direction such as longitudinal, lateral and oblique directions. Therefore, by such a picture data converter, the linear fault part on the work surface is easily inspected.

Description

[Detailed Description of the Invention] A Dangerous Leopard in a Costume [Industrial Application Field] The present invention relates to an image data conversion method, and more specifically, to create image data that emphasizes linear defects formed on the surface of a workpiece. This invention relates to an image data conversion method.

[Prior Art] Conventionally, there is a method of inspecting defective portions formed on the surface of a workpiece based on image data of the surface of the workpiece obtained by imaging with a television camera. A method of conducting an inspection based on such image data is described, for example, in Japanese Patent Application Laid-Open No. 60-1
It is disclosed in Japanese Patent No. 14750. Since the image data is based on the difference in intensity of light reflected from the surface of the workpiece, that is, the brightness, there are cases where defective parts cannot be accurately captured, and therefore, sharpening of the image is required. This sharpening usually involves
This is done using processes such as differentiation and Laplacian.

[Problems to be Solved by the Invention] However, when such differential processing and the like are performed, the following problems occur.

For example, as shown in FIG. 12, when a linear defect a with a certain thickness is imaged and differential processing is performed, the edge method of the defect a has a high rate of change in brightness.Two line segments b An image is formed as follows. Therefore, in order to correctly recognize the shape from this line segment b as a single defect, various methods such as filling in the interior surrounded by line segment b and integrating separated line segments b by expanding line segment b are performed. Furthermore, if these processes are performed, images other than linear defects will remain as shown in FIG. 13. The figure (A) is an image after the interior has been filled in, and the figure (B) is an image in which a plurality of point-like images are combined by dilation processing. Therefore, if you try to detect only linear defects such as cracks that occur in a workpiece, you will need to perform further complicated processing such as thinning processing and shrinkage processing in order to distinguish them from non-linear defects. The processing time becomes long and the equipment becomes large-scale. In addition, the inspection accuracy is not very good, and it is difficult to inspect products on the actual production line (
, 1. The purpose of the image data conversion method of the present invention, which was not suitable, is to solve the above-mentioned problems and to create image data that can accurately detect linear defects occurring on the surface of a workpiece.

The structure of the present invention that achieves the above object will be described below.

[Means for Solving the Problems 1] The image data conversion method of the present invention is a method of converting original image data of 12 workpieces into image data in which linear defects are emphasized, as illustrated in FIG. The first line is set linearly.
While scanning the original image data with a region and a second region set in a linear shape substantially perpendicular to the first region, original image data corresponding to both regions is extracted (P1), and a reference line is obtained. The brightness distribution pattern data in the direction across the defective portion is compared with the original image data in the first and second regions extracted above, and the degree of coincidence is calculated for each (P
2) Based on the degree of coincidence between the first area and the second area calculated above, brightness data is created for each scanning position (P3), and the data is created as new image data.

[Operation] The image data conversion method of the present invention having the above configuration scans the original image data of the workpiece with a first area set linearly and a second area substantially perpendicular thereto and set linearly. At the same time, the original image data corresponding to both regions are extracted (P]), and the original image data and pattern data in both extracted regions are compared to calculate the degree of matching for each (P2). This pattern data {is the brightness distribution in the direction that crosses the linear defect part based on the table standard, and if the degree of coincidence between this pattern data and the original image data extracted from the first and second areas is high, the standard linear defect The existence of a defective part similar to the defective part can be seen. Therefore, if brightness data is created for each scanning position based on the degree of coincidence in the first and second areas, and this brightness data is used as new image data (P3), linear defects on the workpiece surface will be emphasized. image data can be obtained.

[Example] In order to further clarify the configuration and operation of the present invention described above, preferred embodiments of the image data conversion method of the present invention will be described below.

FIG. 2 is a schematic configuration diagram of an image data conversion apparatus in which the method of the embodiment is carried out. This image data conversion device includes a television camera 1 that images a workpiece W placed on a conveyor C to be inspected, an A/D converter 3 that converts a video signal output from the television camera 1 into a digital signal, and an image data a video RAM 5 storing
, a CRT 13 as a display means, and a CR for image data.
It is composed of a CRTC 15 that outputs to T13, and an external device interface 17 that exchanges information with external devices. In addition, A/D converter 3, video RAM 5, CP
IJ7, ROM9, RAMI1, CRTC15, and external interface 17 are interconnected by bus 19.

Next, the image data conversion process executed by the image data conversion device will be described. FIG. 3 is a flowchart showing a graphic image data conversion routine, and this routine is started after the workpiece W is conveyed to a predetermined position by the conveyor C.

First, image data output from the television camera 1 is loaded into the video RAM 5 via the A/D converter 3 (step 00). Therefore, the original image of the workpiece W (for example, 640×400 pixels) is stored in a predetermined area of the video RAM 5.
) (hereinafter referred to as original image data since this image data is before conversion) is stored.

Next, from the original image of the work W, as shown in FIG. 4, original image data at a position corresponding to the cross filter 21 having a cross-shaped area set is extracted (step 110).

This cross filter 21 has a region set so that it can extract data at nine pixels each in the vertical and horizontal directions (the pixel at the center of intersection is common).

The process of step 110 is a process of extracting the original image data (luminance data) in the cross filter 21 from the upper left position of the screen A as shown in FIG. FIG. 6 shows the original image data D1 extracted by this process.

The data at the intersection 23 of the cross filter 21 is
, n, and the data above the intersection 23 are in order:
X m, n-1, Xm, n-2, X m,
n-3, X m, n-4, and the data below the intersection are, in order, X m, n+1, X m
, n+2, X m, n+3, Xm, n+4. Similarly, data in the horizontal direction is from left to
m-4,n, Xm-3,n, Xm-2,n,
Xm-1, n, Xm, n (data of intersection 23) F Xm+1, n, Xm+2, n, Xm+3
, n, Xm+4, n.

When the process of step 110 is completed, the degree of coincidence between the extracted original image data D1 and pattern data Do (shown in FIG. 8, which will be described later) is calculated (step 120). First, the pattern data Do will be explained.

The pattern data DO is created by capturing an image of a predetermined linear defect as a reference (hereinafter referred to as a reference defect) in advance and from a luminance distribution captured in a direction perpendicularly crossing the reference defect. Signal level (indicating brightness) at each sampling point obtained by scanning in a direction perpendicular to the reference defect area
is shown in Figure 7. Note that the sampling point corresponds to one pixel on the imaging screen. The pattern data DO is created by positioning the nine sampled luminance data vertically and horizontally of the cross filter 21, as shown in FIG. The brightness data of the sampling point where the brightness data takes the maximum value is taken as the data at the intersection 23 of the cross filter 21, and the brightness data of adjacent sampling points are sequentially positioned vertically and horizontally to create the data. Pattern data E created in this way: Y-4.0, Y-3.0, horizontally from the left
Y-2.0, Y-1.0, Y0,0, YO,
1, Y0,2, Y0,3, Y0,4, and similarly in the vertical direction from the top Y0,4,
Y0,3, Y0,2, Y0,1, YO,O,
YO, -1, YO, -2, YO, -3, YO
, -4. In this example, each luminance data constituting the pattern data DO has the same value in the horizontal and vertical directions, but the imaging position of the reference defect portion is set to 90.
Similar sampling may be performed at a position shifted by degrees, and luminance data in the vertical direction and horizontal direction may be obtained separately, and in the second case, the pattern data DO will be more accurate.

Next, the extracted original image data D1 and pattern data D
Calculation of the degree of coincidence with O will be explained.

If the degree of coincidence in the horizontal direction between the original image data D1 and the pattern data DO is gl (m, n), then the degree of coincidence in the vertical direction between the original image data and the pattern data is g2
If (m, n), then That is, the original image data D1 and pattern data D in the extracted cross filter 2]
The degree of coincidence is calculated by dividing the total value obtained by multiplying O by the coordinate position corresponding to the coordinate position by the total value of the squares of the data at each coordinate position of the pattern data DO, and this degree of coincidence gl (
When the value of g2(m, n) or g2(m, n) is close to 1 (upper matching is high, value]), the matching becomes lower as it moves away from the value.

When the process of step 120 is completed, luminance data is calculated from the calculated degrees of coincidence gl (m, n), g2 (m, n) (step) 3
0). This process is a process in which the closer the matching degree is to the value 1, the larger the luminance (brightness luminance data), and is performed as follows.In the following explanation, the calculated matching degree gl (m , n), g2 (m,
n) will be explained simply as the degree of coincidence x.

This conversion is performed, for example, assuming that the luminance resolution per pixel of the video RAM 5 is 8 bits, that is, 256 gradations.
This is performed using a linear function conversion table T1 as shown in FIG. When the degree of coincidence X of the brightness G is less than or equal to the value 1, it is calculated as G=255x, and when the degree of coincidence X exceeds the value 1, it is calculated as G=-255x+510. Furthermore, when the degree of coincidence X exceeds the value 2, G=O. Note that the brightness G is calculated in step 120 for each of the calculated vertical and horizontal matching degrees x. Then, the brightness G on the side having a larger value is set as image data G(m,n) for one pixel at the coordinate position of the intersection 23 of the cross filter 21.

In addition, in this embodiment, the conversion table T1 using a linear function is adopted, but a conversion table T2 using a common logarithmic function as shown in FIG. 10 may also be adopted. is calculated as G = 255 log (10x) when the degree of coincidence x is less than or equal to the value 1, and is calculated as G = -255 log (10x) + 512 when the degree of coincidence X exceeds the value 1 ( If ×〉2, G=0).

When the process of step 130 is completed, the obtained image data G (m, n) is written into the video RAM 5 (step 140). Note that this image data G(m,n) is written in a storage area provided separately from the storage area of the video RAM 5 into which the original image data was written in step 100. Next, it is determined whether writing of one screen worth of image data G (m, n) in which the surface of the workpiece W is imaged has been completed (step 150).
. If the determination is "NO", the cross filter 21 is scanned by one pixel and the process returns to step 120 (step 160). That is, the extraction position of the original image data D1 is changed by scanning the cross filter 21 until the writing of one screen worth of image data G (m, n) in which the surface of the workpiece W is imaged is completed, and then step] ]0~1
40 processes are repeated.

If the determination in step 150 is rYESJ, the video R
One screen worth of image data G (m,
n) is output to the CRT 13 via the CRTC 15 (step 170), and the main image data conversion routine ends.

The image data conversion device described above (above) extracts the original image data obtained by imaging the surface of the work W using the cross filter 21 and calculates the degree of coincidence with the pattern data Do,
Image data G (m, n) is obtained based on this degree of matching. Since this pattern data DO is created from the brightness data obtained by imaging the linear reference defect, if a linear defect such as a crack occurs on the surface of the workpiece W, CRT1 3
The image is displayed brightly, highlighting linear defects.

Furthermore, due to the shape of the cross filter 21, the defect is highlighted and displayed regardless of whether it is formed vertically or diagonally. Therefore, with this image data conversion device, linear defects on the surface of the workpiece W can be easily inspected.

In addition, since non-linear irregularities formed on the surface of the workpiece W and adhered dirt are not recognized as defects, only the targeted linear defects are highlighted. There is no need to carry out inspection of products on the production line.
It can be done in a short time and with high precision.

Although the embodiments of the present invention have been described above, the present invention is not limited to these embodiments in any way. For example, in order to increase the calculation speed, the matching degree gl (m, n
), g2 (m, n) may be calculated as follows.

4 In this case, a conversion table T3 as shown in FIG. 11 may be used. It is also possible to increase or decrease the number of pixels of the cross filter 21 (optimally 9 to 15 in each of the vertical and horizontal directions), without departing from the gist of the present invention.
Of course, it can be implemented in various ways.

Effects of the Invention As detailed above, according to the image data conversion method of the present invention, the original image data of the workpiece is extracted from the first and second regions, and the degree of matching with the pattern data is calculated. Image data is obtained based on the degree of matching. Since this pattern data is created from the linear defect part of the above standard, if a linear defect such as a crack occurs on the workpiece surface, that part will be highlighted and displayed due to the difference in brightness. It has an excellent effect of being In addition, since the first region and the second region are set to be approximately perpendicular to each other, the linear defect formed in any diagonal direction can be emphasized and captured. It will be done. Moreover, since non-linear irregularities formed on the surface of the workpiece or adhered dirt are not recognized as defects, only the targeted linear defects are highlighted, and complicated image processing like conventional tags is performed. There's no need.

[Brief explanation of drawings]

FIG. 1 is a flowchart illustrating the basic configuration of the present invention, FIG. 2 is a schematic configuration diagram of an image data conversion device, FIG. 3 is a flow chart showing an image data conversion routine, and FIG. 4 is a configuration diagram of a cross filter. Fig. 5 is an explanatory diagram showing the scanning of the cross filter, Fig. 6 is a configuration diagram of the extracted original image data, Fig. 7 is a graph showing the brightness obtained by imaging the reference defect part, and Fig. 8 is Configuration diagram of pattern data, Section 9.10
, 11 are graphs showing a table for converting the degree of coincidence into brightness, and FIGS. 12 and 13 are explanatory diagrams showing images obtained by conventional processing. 1...TV camera 5...Video RAM7...
・CPU 9...ROM11...RAM
13...CRT2]...Cross filter D1...Original image data D2...Pattern data T
l, T2, T3... Conversion table W... Work

Claims (1)

[Claims] 1. A method for converting original image data of a workpiece into image data that emphasizes a linear defect, the method comprising: a first region set linearly; and a method substantially perpendicular to the first region; While scanning the original image data with the second area set in a linear manner, extracting the original image data corresponding to both areas, and extracting the original image data corresponding to the two areas, and generating the brightness distribution pattern data in the direction across the reference linear defect part. , compare the original image data in the first region and the second region extracted above to calculate the degree of matching, and calculate the degree of matching for each of the extracted first and second regions for each scanning position based on the degree of matching of the first region and the second region calculated above. An image data conversion method that creates brightness data and uses it as new image data.