JP4184511B2 - Method and apparatus for defect inspection of metal sample surface - Google Patents

Description

[0001]
[Technical field to which the invention belongs]
The present invention relates to a defect inspection method and apparatus for a metal sample surface for detecting wrinkle defects generated on the surface of a metal product such as a lead frame.
[0002]
[Prior art]
In recent years, along with higher integration and higher functionality of semiconductor devices, lead frames used for electrical connection have been miniaturized, and it has become difficult to visually inspect defects on the surface of metal products. And the inspection is being automated.
[0003]
For example, as shown in FIG. 17A, an automatic inspection technique for the surface of such a metal product is provided by an area sensor camera 10 disposed above a metal sample W to be inspected and connected to a control unit. In addition, a method is known in which image data obtained by photographing the surface of the metal sample W is processed by an image processing unit to detect a defective portion and, if necessary, the defective portion is displayed on a visual monitor for inspection. ing.
[0004]
In addition, as the illumination used at that time, for example, as shown in FIG. 5B, the illumination light I from the light source 12 is irradiated by the half mirror of the beam splitter 14 and the camera 10 perpendicular to the surface of the metal sample W is used. There is a vertical epi-illumination method in which the camera 10 emits the reflected light R while irradiating in the same direction as the optical axis.
[0005]
By the way, as a kind of surface defect of the metal sample W, as shown in a cross-sectional view in FIG. 18, the defect (b) which is clearly recessed with respect to the non-defective product (a) whose front is flat, and some unevenness. There is a pseudo defect which is displayed in (c), which may not be a defect.
[0006]
[Problems to be solved by the invention]
However, when the above-described three kinds of surface state metal samples are photographed with the camera 10 shown in FIG. 17A under the vertical epi-illumination shown in FIG. 17B, there are the following problems. .
[0007]
That is, FIG. 19 (I) conceptually shows the relationship between the cross section of the metal sample W in each surface state shown in FIG. 18 and the illumination light I and the reflected light R. While the non-defective part is regular reflection, the reflected light R is irregularly reflected in the defect part (b) and the pseudo defect part (c). As shown in correspondence with (), the luminance of both the defective part (b) and the pseudo defective part (c) is lowered. Therefore, in order to detect the defective portion in (b), when the threshold value SL is set so that the image data is binarized and the defective portion can be automatically detected, the luminance profile of the binarized image is shown in FIG. As shown, the pseudo-defect portion (c) is also detected as a defect.
[0008]
Therefore, as a technology that can solve such problems, the present applicant illuminates the metal sample surface from three different elevation angles to photograph the same region, and image data obtained by illumination at each elevation angle obtained by the photographing. In addition, an appearance inspection technique for synthesizing an RGB color image from three image data obtained by assigning one of R, G, and B colors to each other, and discriminating a defect and a pseudo defect from the hue of the synthesized image, This is proposed in Japanese Patent Application No. 9-195238.
[0009]
In addition to using this technology, the applicant further calculates the hue value from the light intensity of the R color component, the light intensity of the G color component, and the light intensity of the B color component for each pixel of the synthesized image. Obtain and classify the hue value for each pixel of the synthesized image according to the database in which the surface state of the object to be inspected and the range of the hue value are associated with each other based on the obtained hue value of each pixel Japanese Patent Application No. 9-251430 has already proposed an appearance inspection technique for discriminating defects and pseudo defects based on the results.
[0010]
Hereinafter, the latter technique will be described in detail with a specific example.
[0011]
FIG. 1 is a block diagram showing a schematic configuration of an appearance inspection apparatus proposed by the applicant of the latter patent application.
[0012]
This inspection apparatus illuminates the surface of a metal sample from three different elevation angles, images the same region, and assigns each of R, G, and B3 colors to image data obtained by illumination at each elevation angle. A color image capturing unit 20 for capturing one image data, an input unit 22 corresponding to an interface for inputting these image data, and an image for performing various image processing on the image data input via the input unit 22 A processing unit 24, a determination unit 26 that determines a defect and a pseudo defect based on a processing result by the image processing unit 24, and an image on which the three image data are combined based on a determination result by the determination unit 26 And a display unit 28 for displaying a defective portion.
[0013]
The image processing unit 24 also includes R, G, and B for each pixel of color image data synthesized from the three image data or image data obtained by image processing of the three image data. A hue conversion unit 30 that obtains a hue value from the light intensity of each color component, and a classification assigning unit 32 that assigns a classification of the hue value for each pixel of the image synthesized based on the processing result by the conversion unit 30 are incorporated. In addition, data is input from the database 34 in which the surface state to be inspected and the range of hue values are associated with each other to the classification assigning unit 32. Further, the image processing unit 24 performs (1) filtering processing, (2) binarization processing, peripheral image removal processing from the image to be inspected for smoothing processing and isolated small area removal, and (3). It also has various processing functions such as composition processing and (4) labeling processing.
[0014]
This appearance inspection apparatus will be further described in detail with reference to FIG. 2 which specifically shows an example of the color image capturing unit 20. The take-in unit 20 includes an area sensor camera 10 that is a monochrome camera that photographs the surface of the metal sample W, and an illumination device 40 that illuminates the surface.
[0015]
In this illumination device 40, as shown in FIG. 2B, circular light sources 40A and 40B are arranged on the surface of the metal sample W at three different maximum, intermediate, and minimum elevation angles θ1, θ2, and θ3. , 40C. Here, the elevation angle is an angle formed between the photographing direction (optical axis) L0 of the camera 10 and the intersection P0 orthogonal to the surface S of the metal sample W and the circular light sources 40A, 40B, and 40C. As the circular light source, a circle-shaped fluorescent lamp, an optical fiber in which a tip portion irradiated with illumination light is arranged in a circle shape, an LED arranged in a circle shape, or the like can be used.
[0016]
The color image capturing unit 20 controls lighting of the transmission light source 42 that illuminates the metal sample W from the back side and the light sources 40A to 40C of the illumination device 40 when capturing a masking image to be described later. The lighting unit 44, the control unit 46 that controls the operation of the entire capture unit 20, and the reflected illumination from the circular light sources 40 </ b> A to 40 </ b> C of the lighting device 40 obtained by photographing with the camera 10, And an image color assigning unit 48 for assigning any one of R, G, and B colors by software.
[0017]
And in this taking-in part 20, when the surface of the metal sample W is image | photographed with the monochrome camera 10, each circular light source 40A-40C currently fixed to the said elevation angle position by the said control part 46 and the lighting part 44 is used. The lights are lit for a predetermined period of time at different times.
[0018]
The principle that the defect portion and the pseudo defect portion can be identified by using the image data captured by the color image capturing unit 20 described above will be described here for convenience. As shown in FIG. 3A, the amount of light incident on the camera 10 of specularly reflected illumination light from each elevation angle θ1 to θ3 varies depending on the inclination of the surface of the metal sample W. The obtained images have different light intensities depending on the inclination of the surface of the metal sample to be photographed.
[0019]
That is, in illumination from a position with a large elevation angle θ1, the closer the surface of the metal sample W is to be flat (the inclination θ21 is 0 °), the greater the amount of specular reflection, and the smaller the elevation angle θ2 and the elevation angle θ3. In this illumination, the closer the metal sample surface is to the predetermined inclinations θ22 and θ23, the greater the amount of specular reflection light.Therefore, on each of these surfaces, the images of the images by illumination from the corresponding elevation angles θ1, θ2, and θ3 Luminance (light intensity) increases.
[0020]
Therefore, the individual images obtained by illuminating from the above elevation angles have regions with high luminance in different surface ranges, as shown in FIGS. Then, for example, colors B, G, and R are assigned to images obtained by illumination from the respective elevation positions of θ1, θ2, and θ3, and when these are displayed, the images shown in FIGS. 3B, 3A, and 3C are displayed. Since the range where each luminance (light intensity) is large becomes the shades of B, G, and R, when these color images are combined, as shown in FIG. An image will be obtained.
[0021]
When this principle is applied to the surface inspection of the metal sample W shown in (a) to (c) of FIG. 4 (I) corresponding to FIGS. 18 (a) to (c), a composite image of R, G and B is obtained. Are shown in B color as a whole, and defective parts are displayed in B, G, and R colors in order from the center to the outside as shown in FIGS. The pseudo defect portion has a color in which R, G, and B are irregularly mixed. Usually, the pseudo defect portion is often achromatic or yellow on the composite image. In this way, in the composite image, since the hues of R, G, and B differ depending on the surface state of the metal sample, the hues of the defective portion and the pseudo-defect portion are also different. Become.
[0022]
In the appearance inspection apparatus shown in FIG. 1, in order to detect defects other than the pseudo defects based on the detection principle described above, the color image capturing unit 20 shown in FIG. Each image data of three inspection images to which each color of B is assigned and one masking image is fetched and input to the image processing unit 24 via the input unit 22, and the following processes are executed. The This is an example in which the object to be inspected is a metal sample manufactured by forming a through-hole in a metal plate such as a lead frame, and the flowchart shown in FIG. 5 and the image of the corresponding process are shown in FIG. This will be described with reference to specific examples.
[0023]
First, the color image capturing unit 20 illuminates the surface of the metal sample W from three different elevation angles to photograph the same region, and the obtained image data obtained by illumination at each elevation angle is R, G, B3 colors, respectively. The three image data obtained by assigning each of the colors are taken in (step 1).
[0024]
In the specific example of FIG. 6, (a) corresponds to a part of a color image obtained by synthesizing the three image data captured as the image to be inspected in Step 1. 6A, A is an inspection area, B is a peripheral portion (background portion), S is a surface side of a metal sample W, E is an edge portion thereof, Sa is a defective portion, and Sb is a pseudo defect. Part, Sc represents a noise part, and Sd represents a minute area having a different hue from the surroundings.
[0025]
Next, noise removal processing is performed on the color image data for inspection (inspected image) input to the image processing unit 24 (step 2). As this noise removal processing, for example, for each pixel of the processed image data, for example, filtering processing, smoothing is performed by newly setting the average value of nine pixels including the surrounding pixels (average value of light intensity) as the pixel value. Processing. In this filtering process, when each pixel value of the processed image data is f (i, j) and each pixel value of the processed image data is g (i, j), the expression shown in FIG. This corresponds to the process of applying the 3 × 3 filter h (t, l) shown in FIG.
[0026]
This noise removal is performed on the combined image data in FIG. 6A obtained by combining the three images input by illumination with the circular light sources 40A to 40C, and then the combined image is pixel by pixel. Hue conversion is performed (step 3). However, this noise removal may be performed on each of the three image data before synthesis, and then these three images may be synthesized.
[0027]
For example, the hue conversion uses the Haydn hue conversion formula given by the following equations (1) to (3) to obtain the hue value H and to convert the image into an image based on the hue value. The hue value H obtained as a real number between 0 and 3 according to the equation is further applied to the following equation (4) to calculate a new hue value H0. Executed. However, when it is calculated by the equation (4), the decimal point is rounded down.
[0028]
(When R, G ≧ B)
H = (GB) / (R + G-2B) (1)
(When G, B ≧ R)
H = (BR) / (G + B-2R) (2)
(When R, B ≧ G)
H = (RG) / (B + R-2G) (3)
H0 = 255 / 3H (4)
[0029]
FIG. 6B shows an image after the above-described noise removal processing and hue conversion processing are performed on the inspection image shown in FIG.
[0030]
In parallel with the processes in steps 1 to 3, masking image data for masking areas other than the inspection area of the inspection color image is captured (step 4). This masking image data is also created by photographing the metal sample W by the color image capturing unit 20 using the transmission light source (transmission illumination means) 42. FIG. 6C shows an image of this masking image.
[0031]
The masking image data captured in step 4 is subjected to a binarization process for binarizing with a predetermined threshold in pixel units (step 5). In this process, since the periphery of the inspection area is penetrated by the lead frame, when the inspection area and its periphery are photographed by illumination from the transmissive light source 42, transmitted light is transmitted in the inspection area and the periphery. Since there is a large difference in intensity, by binarizing the captured image data at a predetermined slice level, it is possible to divide the inspection region and its peripheral portion in units of pixels.
[0032]
In this way, with respect to image data obtained by imaging with transmitted illumination, a binary image is obtained by setting the value (light intensity) of each pixel to 1 when the value is greater than the slice level (threshold) and 0 when it is smaller. In general, the pixel value of the region corresponding to the peripheral portion is 1 and the inspection region is 0.
[0033]
However, in general, since the edge portion of the metal sample is tapered, it is necessary to exclude the edge portion from the inspection region in order to perform automatic inspection. Therefore, a process for expanding the mask area by several pixels in the peripheral area is performed as necessary for the binarized image data obtained by the binarization process (step 6).
[0034]
This will be specifically described with reference to FIG. FIG. 8 (a) shows a partial cross-sectional view of a metal sample having a tapered edge portion E. FIG. 8 (b) shows a reflection image when the metal sample W is photographed under vertical epi-illumination. FIG. 6C shows a binary image obtained by transmitting a transmission image obtained by transmission illumination, and shows the positional relationship between the two. In the figure, S and B are areas of values 0 and 1, respectively.
[0035]
FIG. 8D shows an image after the region B corresponding to the peripheral portion of the binarized image shown in FIG. 8B is expanded and expanded so as to include the region corresponding to the edge portion E. In the figure, S ′ and B ′ in the figure are an inspection area and a masking area that are actually used, respectively. By using such an actual masking image, it is possible to limit only an area that should originally be flat except for an edge portion of the metal sample as an inspection area. FIG. 6C shows binary image data obtained by performing the binarization processing in step 5 on the masking image corresponding to the image to be inspected shown in FIG. Shows the state in which the expansion processing of step 6 has been performed on the binarized image data, and corresponds to FIGS. 8 (c) and 8 (d), respectively.
[0036]
Next, the masking area B (peripheral area and background area) in FIG. 8D obtained by the expansion process in step 6 is excluded from the subsequent processing objects from the area of the image data subjected to hue conversion in step 3. Part (background part) removal processing is performed (step 7). In the specific example of FIG. 6, this processing is performed by the image data of FIG. 6B, the inspection data A shown in FIG. 5E from the image data shown in FIG. This corresponds to obtaining an image composed of the non-inspection area (B). The subsequent processing is performed only for the inspection area A.
[0037]
Next, with reference to the database 34 in which the range of the defect and the hue value is associated with the classification in advance, the classification corresponding to the hue value is given to the image data subjected to the hue conversion in units of pixels (step 8). For example, as shown in FIG. 9, this is a process of assigning A to E classifications corresponding to a predetermined range of the hue value H0 calculated by the equation (4). The normal surface, C is set as a pseudo defect portion, and D and E are set as a defect portion. In this classification assignment, for convenience, each classification of A to E is identified by assigning a numerical value representative of the hue. In other words, each pixel has a numerical value representing the hue of each classification, for example, the maximum value of each classification value. In the specific example of FIG. 6, the processing of step 8 is the image shown in FIG. 6F in which the classifications A to E are assigned to the masked image data shown in FIG. It corresponds to data.
[0038]
Next, in order to remove an area (pixel area) that is classified as a defect in the classified image data but is smaller than necessary, an isolated minute area removal process is performed (step 9). This process is performed using, for example, a numerical value representing the hue of each classification given to each pixel. This will be described with reference to FIG.
[0039]
When the pixel is represented by one rectangle in FIG. 10A and each central pixel is (f (i, j)), the same figure ( As shown in equation (1) of b), for each central pixel (f (i, j)), if the numerical value is maximum for a total of five pixels including the upper, lower, left, and right pixels of the pixel, As shown in the maximum filtering process in which the numerical value is newly set to the numerical value of the central pixel, or the expression (2) in FIG. For the pixel, a minimum filtering process is performed in which the numerical value of the pixel having the smallest numerical value is newly set as the numerical value of the central pixel, changing the order of both processes as necessary.
[0040]
For example, when the classification and numerical value are as shown in FIG. 10D, the maximum filtering process is performed after the minimum filtering process. Note that the maximum and minimum filtering processes are not limited to a total of 5 pixels including the central pixel and 4 neighboring pixels, but may be performed for a total of 9 pixels including the 8 neighboring pixels.
[0041]
Next, as described above, the locations classified as defects are labeled from the multilevel data classified into the respective categories (step 10). This labeling process is used when the final pass / fail judgment is made based on the area and the number in the subsequent processes. Next, a final pass / fail determination is performed based on the labeling result (step 11), and the determination result is further displayed (step 12). In the specific example shown in FIG. 6, FIG. 6G is image data corresponding to the result of performing the isolated small region removal processing in step 9 on the image data shown in FIG. h) corresponds to the image after the defective portion in step 11 is determined.
[0042]
According to the already proposed metal surface inspection technology described in detail above, when detecting defects on the surface of a metal product (metal sample) such as a lead frame, the defect portion can be reliably detected. There is an excellent advantage that the pseudo defect portion can be classified based on a certain standard, and the division can be automatically performed.
[0043]
However, as a result of further detailed examination of the appearance inspection technique by the present inventor, the feature is that the entire surface is almost flat and the surface state is closer to a mirror surface than a normal metal surface, depending on the inspection technique. It became clear that there was a new problem that it was not possible to detect defects with.
[0044]
That is, the normal defect as shown in FIG. 18B has a small elevation angle, that is, it can be detected by capturing the reflected light from the low-angle illumination with the upper camera (photographing means), while being handled. The contact flaws that are partially formed on the metal surface due to contact with a foreign object, such as when it is struck from above with a thin object, or scratched, etc., have a shallow recess and a mirror surface. Therefore, the reflected light from the low-angle illumination is regularly reflected and does not go in the direction of the camera. As a result, the contact rod and the product surface cannot be distinguished and cannot be detected.
[0045]
The present invention has been made to solve the above-mentioned new problems. The same area is photographed by illuminating the surface of the metal sample from three different elevation angles, and the obtained image data by illumination at each elevation angle is R, It is an object of the present invention to provide a technique capable of reliably detecting contact wrinkles when inspecting the surface of the metal sample based on three image data obtained by assigning G and B.
[0046]
[Means for Solving the Problems]
The present invention illuminates the surface of a metal sample from three different elevation angles and photographs from above, and synthesizes an RGB color image from the three image data obtained by assigning R, G, and B to the obtained image data of each elevation angle. , A defect inspection method on the surface of a metal sample for inspecting defects from the shades of R, G, B of the composite image, Of the three different elevation angles An image taken by illuminating from the maximum elevation angle is extracted, the image is binarized with a threshold value higher than the brightness of the normal metal sample surface, and the defect is determined based on the binarized image, thereby solving the above-mentioned problem. Is.
[0047]
The present invention is also based on illumination means for illuminating the surface of a metal sample from three different elevation positions, photographing means for photographing an image of reflected illumination from each position of the illumination means from above, and reflected illumination from each position. R, G, and B are assigned to each of the captured images, and an input unit that inputs each image and an image processing unit that detects a defect based on a composite image obtained by synthesizing the input images. A defect inspection device, Of the three different elevation positions Illumination by illumination means at the maximum elevation angle position, means for extracting the image captured by the imaging means, means for binarizing the extracted image with a threshold higher than the brightness of the normal metal sample surface, and the generated binary The above problem is similarly solved by providing a means for determining a defect based on the digitized image.
[0048]
That is, in the present invention, the photographing means arranged above the metal sample surface, Of three different elevation angles Since the image obtained by illuminating from the maximum elevation angle and photographing the surface is binarized with a threshold value higher than the brightness of the normal surface, almost only specular reflection light can be imaged, so regular reflection from the normal surface is possible. A large amount of contact wrinkles can be detected.
[0049]
DETAILED DESCRIPTION OF THE INVENTION
Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.
[0050]
The defect inspecting apparatus according to the first embodiment of the present invention has already been proposed by the present applicant, except that the image processing unit 24 shown in FIG. 1 has functions (means) described below. The visual inspection apparatus shown in FIGS. 1 and 2 is substantially the same. Therefore, the same reference numerals are used for the same parts, and detailed description thereof is omitted.
[0051]
In the present embodiment, the image processing unit (image processing unit) 24 illuminates with the illumination unit at the maximum elevation angle position, extracts the image captured by the imaging unit, and the extracted image is higher than the brightness of the normal metal sample surface. A means for binarizing with a threshold value, a means for determining a defect based on the created binarized image, a means for removing isolated points as necessary, and a defect part are extracted and necessary Accordingly, a means for calculating the area and the like and making a final pass / fail judgment from the size is provided. Each of these means (functions) is configured by software.
[0052]
In the present embodiment, contact wrinkles are inspected according to the flowchart shown in FIG. In this flowchart, each process of inputting an image to be inspected in step 1 and noise removal in step 2 are the same as those in the flowchart shown in FIG. 5, and during the hue conversion in step 2 and step 3, This corresponds to the addition of each processing step of steps 21 to 24 surrounded by a broken line. Therefore, step 4 and subsequent steps are omitted.
[0053]
First, as in the case of FIG. 5, an image to be inspected is input in step 1, and noise is removed from the image in step 2. Next, high-angle image extraction is performed to extract an image corresponding to photographing under illumination from the maximum elevation angle θ1 from the color image obtained in step 1. In this case, the camera 10 (imaging means) is arranged so that its optical axis is substantially perpendicular to the surface of the metal sample W, and the elevation angle θ1 is 70 ° or more, preferably 70 ° to 90 °. Has been.
[0054]
Next, the extracted high-angle image is binarized (step 22). When the surface of the metal sample W is illuminated from the maximum elevation angle θ1, the reflection on the normal surface A is compared with the reflection on the contact rod B in FIG. Although regular reflection is present, irregular reflection components are also present, whereas the contact rod is almost regular reflection. Therefore, when the pixel value of the captured image in the vicinity of the contact rod is displayed with a gradation value of 0 to 255, a high-angle illumination image in which the contact rod B is included in the normal surface as shown in FIG. As shown in FIG. 5B, the profile on the line L in FIG. 5B is obtained by adjusting the sensitivity so that the pixel value of the normal surface A becomes an appropriate value. May be saturated.
[0055]
Therefore, in the present embodiment, as shown in FIG. 13B, the threshold SL is set between the brightness of the normal metal surface A and the brightness (maximum value) of the contact defect (defect portion) B. Set and binarize.
[0056]
After the above-described contact wrinkle detection process is completed, the process may be terminated as it is. However, as shown in FIG. 11, each process after step 3 is performed as in the case of the flowchart of FIG. A defect and a pseudo defect may be inspected.
[0057]
According to this embodiment described above in detail, it is possible to reliably detect contact defects, and to detect true defects by distinguishing defects from pseudo defects as required.
[0058]
Next, a second embodiment according to the present invention will be described. This embodiment is substantially the same as the first embodiment except that the color image capturing unit 20 uses the apparatus shown in FIG. 15 instead of the apparatus shown in FIG. .
[0059]
In the image capturing section 20 shown in FIG. 15, a monochrome camera is similarly employed as the photographing means 10, and a single circular light source 40A as the illuminating device 40 is sequentially moved to three different elevation angles θ1 to θ3. Except for the provision of the portion 50, it is the same as that shown in FIG.
[0060]
Next, a third embodiment according to the present invention will be described. The present embodiment is substantially the same as the first embodiment, except that the apparatus shown in FIG. 16 is used as the image capturing apparatus 20.
[0061]
In this color image capturing unit 20, a color camera is adopted as the photographing means 10, and circular light sources 40A to 40C fixed at three different elevation angles θ1 to θ3 as illumination means are respectively R, G, and B. One color is used for illumination.
[0062]
Although the present invention has been specifically described above, the present invention is not limited to that shown in the above embodiment, and various modifications can be made without departing from the scope of the invention.
[0063]
For example, in the above-described embodiment, the case where the photographing unit is an area sensor camera is shown, but a line sensor camera may be used.
[0064]
Moreover, in the said embodiment, although the example provided with the circular light source as an illumination means was shown, it is not limited to this.
[0065]
【The invention's effect】
As described above, according to the present invention, the same area is photographed by illuminating the surface of the metal sample from three different elevation angles, and each of the obtained image data obtained by illumination at each elevation angle is 1 each of R, G, and B3 colors. When the surface of the metal sample is inspected based on the three image data obtained by assigning colors, it is possible to reliably detect contact wrinkles.
[Brief description of the drawings]
FIG. 1 is a block diagram showing a schematic configuration of a proposed appearance inspection apparatus.
FIG. 2 is an explanatory diagram showing a main part of a color image capturing unit provided in the appearance inspection apparatus.
FIG. 3 is an explanatory diagram showing the relationship between the surface state of a metal sample and a photographed image.
FIG. 4 is an explanatory diagram showing the relationship between the surface state of the metal sample and the synthesized image.
FIG. 5 is a flowchart showing a processing procedure by the proposed appearance inspection apparatus.
FIG. 6 is an explanatory diagram showing a processing procedure by giving a specific example of processed image data.
FIG. 7 is an explanatory diagram showing the principle of noise removal processing.
FIG. 8 is an explanatory diagram showing image data for masking and a creation procedure thereof.
FIG. 9 is an explanatory diagram showing the meaning of class assignment
FIG. 10 is an explanatory diagram showing the principle of isolated small region removal.
FIG. 11 is a flowchart showing characteristics of a processing procedure according to the first embodiment of the present invention.
FIG. 12 is an explanatory view showing optical characteristics of the contact rod.
FIG. 13 is an explanatory diagram showing the principle for binarizing a captured image and detecting contact wrinkles
FIG. 14 is an explanatory diagram showing an image of a binarized image after labeling a contact rod
FIG. 15 is an explanatory diagram showing features of the second embodiment.
FIG. 16 is an explanatory diagram showing features of the third embodiment.
FIG. 17 is an explanatory view showing a conventional inspection apparatus.
FIG. 18 is an explanatory diagram showing defects on the surface of a metal sample.
FIG. 19 is an explanatory view showing a conventional inspection method.
[Explanation of symbols]
W ... Metal sample
10 ... Camera
12 ... Light source
14 ... Beam splitter
20 ... Color image capture unit
22 ... Input section
24. Image processing unit
26: Determination unit
28 ... Display section
30 ... Hue conversion section
32 ... Classification part
34 ... Database
40 ... Lighting device
40A-40C ... Circular light source
42: Transmitted light source
44 ... Lighting part
46. Control unit
48. Image allocation unit
50 ... Moving part

Claims (15)

The surface of the metal sample is illuminated from three different elevation angles and photographed from above, and an RGB color image is synthesized from the three image data obtained by assigning R, G, and B to the obtained image data at each elevation angle. A defect inspection method for a metal sample surface for inspecting defects from the shades of R, G, and B,
Extract an image taken by illuminating from the maximum of the three different elevation angles ,
The image is binarized with a threshold higher than the brightness of the normal metal sample surface,
A defect inspection method for a surface of a metal sample, wherein a defect is determined based on the binarized image. In claim 1,
Further, for each pixel of the composite image, a hue value is obtained from the light intensity of each of the R, G, and B color components, and from the obtained hue value of each pixel, a surface state to be inspected and a range of hue values A defect inspection method for a surface of a metal sample, characterized by assigning a hue value classification to each pixel of a composite image in accordance with a database in which the correspondence is made and discriminating between a defect and a pseudo defect based on the classification result . In claim 2,
A defect inspection method for a surface of a metal sample, wherein the hue value is a numerical value obtained from a hue conversion formula defined by Haydn or a predetermined multiple of the numerical value. In claim 1,
Defect inspection on the surface of a metal sample characterized in that noise is removed by performing a smoothing process on each image data or composite image to which R, G, and B respectively obtained by photographing are assigned Method. In claim 2,
A defect inspection method for a surface of a metal sample, characterized in that an isolated minute region removal process is performed to remove a region smaller than necessary classified as a defect in pixel data that has been classified. In claim 2,
A method for inspecting a defect on a surface of a metal sample, wherein a portion classified as a defect is labeled on the classified image data, and the quality of the defective portion is determined based on the label. In claim 1,
A defect inspection method for a metal sample surface, characterized by performing a masking process for limiting an inspection region on the surface of the metal sample. An illumination unit that illuminates the surface of the metal sample from three different elevation angles, an imaging unit that captures an image of reflected illumination from each position of the illumination unit, and an image captured by reflected illumination from each position are R , G, B, an input means for inputting each image, and an image processing means for detecting defects based on a composite image obtained by synthesizing each input image, ,
Means for illuminating with illumination means at the maximum elevation angle position among the three different elevation angle positions, and extracting an image photographed by the photographing means;
Means for binarizing the extracted image with a threshold higher than the brightness of the surface of a normal metal sample;
A defect inspection apparatus for a metal sample surface, comprising: means for determining a defect based on the created binarized image. In claim 8,
For each pixel of the composite image created from the three image data or image data obtained by performing image processing on the three image data, the image processing means calculates the hue from the light intensity of each of the R, G, and B color components. For each pixel of the composite image, referring to the database in which the hue conversion means for obtaining the value and the surface state of the target to be inspected based on the processing result of the hue conversion unit and the range of the hue value are made, A defect inspecting apparatus for a surface of a metal sample, comprising: a classifying unit that provides a classification of the hue value. In claim 8,
The surface of the metal sample is illuminated from three different elevation angles and the same region is photographed from above, and three image data are created by assigning R, G, and B to the obtained image data by illumination at each elevation angle. An apparatus for inspecting a defect on a surface of a metal sample, comprising an image capturing unit. In claim 10,
The inspection image capturing unit includes at least an illumination unit that illuminates the surface of the metal sample from three different elevation angle positions, and an imaging unit that captures an image of reflected illumination from each position of the illumination unit from above. An apparatus for inspecting a defect on a surface of a metal sample, comprising an image color assigning unit for assigning a photographed image by reflected illumination from each position to R, G, and B, if necessary. In claim 10 or 11,
When the inspection image capturing unit further captures a masking image for masking a region other than the inspection region of the inspection image, the inspection region of the metal sample and the surrounding background portion are transmitted through the illumination of the metal sample. A defect inspection apparatus for a surface of a metal sample, characterized by comprising a transmission illumination means for taking a picture at a point. In claim 11 or 12,
The photographing means is a monochrome camera, and the illumination means is provided with fixed light sources fixed at three different elevation positions, and each light source is turned on for a predetermined time with a different time. A defect inspection apparatus for a surface of a metal sample, comprising a lighting part. In claim 11 or 12,
An apparatus for inspecting a defect on a surface of a metal sample, wherein the photographing means is a monochrome camera, and the illumination means includes a movable part that sequentially moves one light source to three different elevation positions. In claim 11 or 12,
An apparatus for inspecting a defect on a surface of a metal sample, wherein the photographing means is a color camera, and the illumination means is a fixed light source having illumination light of each of RGB colors at three different elevation positions.

Images