JPH0868765A - Foreign object detection method by image processing - Google Patents

Abstract

(57) [Summary] (Corrected) [Purpose] To detect water droplets, dust, scratches, and stains separately from non-metallic inclusions. [Structure] An image of an object is picked up to generate a grayscale image, a density profile on a line passing through a central portion of a rounded foreign object in the grayscale image is obtained, and the first derivative of this density profile passes through 0 points. Is calculated and the number of points passing through this 0 point is 3
In the above case, the foreign matter is treated as water droplets. Also, dust, scratches, and stains are determined from the density and shape of the foreign matter.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a foreign matter detection method by image processing for determining dust, scratches, water drops, etc. on a sample by image analysis when inspecting a metallographic structure with a microscope.

[0002]

2. Description of the Related Art A television camera is connected to a metallographic microscope to measure non-metallic inclusions in a steel material to be measured from a captured image. These are JIS-G-05
55, ASTM-E45 / 1245, etc., a binary image is created from the captured image, and the size and shape of particles and the number of consecutive particles within a predetermined interval are measured to detect non-metallic inclusions. To do.

The surface of the sample to be measured is one that has been polished into a mirror-finished state by a polishing machine or that has been etched. These polishing machines may cause scratches on the sample surface. Further, after polishing, the surface is washed with water, dried with hot air or wiped with a solvent to set the surface on a microscope, but water droplets may remain. In addition, after the water droplets are dried, dust floating in the water droplets may be dried and adhere to the sample to cause stains. In addition, dust may adhere to the sample during wiping.

[0004]

Regarding the measurement of non-metallic inclusions, automatic measurement is being carried out by using image processing technology. However, when water drops, dust, scratches, stains, etc. are mixed, these judgments are made. In many cases, it is difficult for the inspector to judge them by directly looking at the microscopic image, which is a major obstacle to automatic measurement.

The present invention has been made in view of the above problems, and an object of the present invention is to provide a method for automatically determining water drops, dust, scratches, stains and the like.

[0006]

In order to achieve the above object, in the invention of claim 1, an object is imaged to generate a grayscale image, and the grayscale image appears in the grayscale image and passes through the central portion of the foreign matter having a shape close to a circle. The concentration profile on the line is determined, and the number of points where the first derivative curve of the concentration profile passes through 0 is determined. When the number is 3 or more, the foreign matter is determined to be a water drop.

According to the second aspect of the present invention, an object is imaged to generate a grayscale image, the area of the foreign matter appearing in the grayscale image is larger than a predetermined value, and the density on a plurality of lines passing through the foreign matter is high. The profile is determined, and the number of points at which the first-order differential curve of each concentration profile passes 0 is determined. When the maximum value of this number is 3 or more, the foreign matter is determined to be a water drop or dust.

According to the third aspect of the present invention, an object is imaged to generate a grayscale image, the area of the foreign matter appearing in the grayscale image is larger than a predetermined value, and the density on a plurality of lines passing through the foreign matter is high. The profile is obtained, and when the gradient at the end of the foreign substance of the first derivative curve of the concentration profile is less than or equal to a predetermined value, the foreign substance is determined to be dust.

Further, in the invention of claim 4, the object is imaged to generate a grayscale image, and the ratio of the length to the width of the foreign matter appearing in the grayscale image is not less than the first set value, and the length is the first. There are linear particles in a direction different from the rolling direction with a setting value of 2 or more, and a line is extended in the length direction of the particle, and a band-shaped region having the thickness of this line as Sk is defined as L, and within this band-shaped region L All particles are judged as scratches.

Further, in the invention of claim 5, an object is imaged to generate a grayscale image, and the ratio of the length to the width of the foreign matter appearing in the grayscale image is not less than the first set value, and the length is the first. There is a linear particle of less than 2 set values, a line is extended in the length direction of this particle, and the band-shaped region where the thickness of this line is Sk is L, and the number of particles in this band-shaped region L is a predetermined value or more. At this time, the particles in the band-shaped region L are determined as scratches.

In the sixth aspect of the present invention, an object is imaged to generate a grayscale image, and for the particles appearing in this grayscale image, a rectangular h × v inspection frame is set around each particle, and m When an inspection frame that includes the above particles and that overlaps with other inspection frames is taken out, and the minimum rectangle that encloses all the particles included in this overlapping inspection frame group is set, and this rectangle becomes almost square , The particles in this rectangle are determined as stains.

[0012]

In the invention of claim 1, in the case of a water drop, it is close to a circle, and the degree of the circle is determined by, for example, a circle coefficient πML ² / (4A
0) (ML is the maximum length of the foreign matter, A0 is the area of the foreign matter), and if the circular coefficient is close to 1, it is close to a circle (for example, 0.7 or more). In the case of water drops, as shown in FIG. 2, the center is bright and the surroundings are dark like a donut. Therefore, as shown by the broken line, the concentration profile on the line passing through the substantially central portion of the grayscale image of the water drop has a dark peak at the donut portion and a bright peak at the central portion. Therefore, if the first derivative is taken, it becomes 0 at each peak position, and three such 0 points occur. This makes it possible to detect water drops. It may become a non-concentric double ring depending on the swell of the water drop and the condition of the illumination that hits it.

According to a second aspect of the present invention, there is shown a method for detecting a water drop or dust by scanning a foreign material which is a binarized black particle in a plurality of directions, and when a water drop has a bright portion due to reflection or a depth of focus. If there are one or more white spots in the dust of the inside thickness, and if there are three or more points that pass through 0 of the first derivative curve of the concentration profile on any of the scanning lines, then it is judged as a water drop or dust. it can.

In the third aspect of the present invention, in the case of dust, the area thereof is larger than a predetermined value, for example, 1 μm ² , and is present higher than the sample surface. Since the focus of the microscope is on the sample surface, the dust looks blurry. Therefore, the density profile on the line passing through the foreign matter appearing in the grayscale image and this primary differential curve are obtained, and when the gradient at the end of the foreign matter of this primary differential curve is smaller than a predetermined value, it can be determined as dust. In other words, a non-metallic inclusion that is in focus, such as a non-metallic inclusion, has a sharp gradient image of the first-order derivative curve at its end and is a clear gray image, and dust can be distinguished.

In a fourth aspect of the present invention, when the ratio of the length to the width of the foreign matter is not less than the first set value and is elongated, and the length is not less than the second setting, it is a linear particle extending in a direction different from the rolling direction. In this case, extend this linear particle in its length direction and set its width to S
A band-shaped region L having a length of k is provided, and all particles in the band-shaped region L are determined to be scratched. Sk is preferably about 1 μm, for example.

In the invention of claim 5, when the ratio of the length to the width of the foreign matter is longer than the first set value and elongated but the length is shorter than the second set value and is short, the particles are extended in the length direction and the width thereof is increased. Is set as Sk, and when the number of particles in the belt-shaped region L is equal to or larger than a predetermined value, all the particles in the belt-shaped region L are determined to be scratched.

In the invention of claim 6, a rectangular h × v inspection frame is set around each particle, and m is set in this inspection frame.
Take out the inspection frame containing more than one particle. Furthermore, take out the inspection frame groups that overlap each other from this inspection frame, enclose all the particles contained in this inspection frame group with the smallest rectangle, and when this rectangle is almost square, stain the particles in this rectangle. To determine. In many cases, stains that are floated on water are dried, and at this time, the moisture becomes a circle, and the dust that floats here also remains in a circle, so it gathers in a square and becomes a stain.

[0018]

Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 is a block diagram showing the arrangement of an apparatus for carrying out this embodiment. An image pickup lens 16 is attached to the eyepiece portion of the microscope 1 and an image pickup device 16 for picking up an image through the image pickup lens is attached. The stage 17 on which the measurement sample is placed is subjected to plane position adjustment by a plane moving mechanism 18 for moving the stage 17 back and forth and left and right by a pulse motor provided on the stand in response to a signal from the auto stage driver 10.
Vertical movement mechanism 19 by autofocus driver 11
Is operated to move the stage 17 in the vertical direction to focus.

The A / D converter 2 converts the input data from the image pickup device 16 from analog to digital, and the input buffer 3 temporarily stores the digital data. The bus 4 transmits signals, the program memory 5 stores a program that defines the operation of the apparatus, and the CPU 6 controls the entire apparatus according to this program.

The image processor 7 performs grayscale processing, binarization processing, image analysis, etc. of the input image data, the grayscale image memory 8 stores the grayscale image data, and the binarization memory 9 stores 2
Stores value or multi-valued image data. The auto stage driver 10 controls the plane moving mechanism 18 to move the stage 17 on which the measurement sample is placed according to an instruction from the CPU 6 to move X, Y.
Direction, and set the measurement position and area of the measurement sample. The autofocus driver 11 receives a control command from the CPU 6 to the vertical movement mechanism 19 and controls the vertical movement mechanism 19 to automatically focus. The output buffer 12 temporarily stores the data to be output, and the D / A converter 13 converts this output data from digital to analog, and the CRT
14 displays this output data on the screen. Keyboard 1
From 5 the operator inputs instructions and data.

Next, the first embodiment will be described. This example shows a method for detecting water drops. When measuring non-metallic inclusions in metal with a metallurgical microscope, the surface of the sample to be measured is mirror-finished with a polishing machine, washed with water to evaporate or wipe off water, and then set on the metallographic microscope. It may remain. FIG. 2 shows a grayscale image of water particles, a concentration profile y of a line A passing through the central portion of the water particles, and its first derivative curve y '. The gray-scale image of water particles has a bright central part and a dark donut-shaped part around it. The broken line shows the concentration profile y of the line A, the vertical axis y shows the concentration, and the horizontal axis x shows the position of the particles. A dark peak occurs in the donut part and a bright peak value occurs in the center part. The solid line shows the first derivative curve y ′ of the concentration profile y, the vertical axis y ′ represents the first derivative value of the concentration, and the horizontal axis x represents the position of the particle. The first derivative curve is 0 at the position where the concentration profile has a peak or bottom. The point that becomes 0 is called a zero cross point. Since three broken line peaks or bottoms occur in one particle region, three zero cross points also occur. It can be identified from the fact that the particles representing the water droplets are close to a circle, and the concentration profile of the line passing through the substantially central portion of the water particles is obtained, and this first-order differential curve has three or more zero-cross points. The region where the zero-cross points are counted is between the edges of the particles.

FIG. 3 is an operation flow chart for identifying water drops. First, the area to be measured is set (S1). in this case,
Since each particle is measured, the measurement area is one particle.
Next, the circular shape factor is used to investigate the closeness of particles to a circle. The circular shape factor is expressed by the following equation. Circular shape factor = πML ² / (4A0) (1) Here, ML represents the maximum length of the particle, and A0 represents the area of the particle. The circular shape factor becomes 1 when the particle is a perfect circle, and becomes a smaller value as it goes away from the perfect circle. The set value K is preferably set to 0.6 to 0.7 as a reference of whether or not it is close to a circle, and is set to 0.6 in the embodiment. In addition, the judgment of whether it is close to a circle is
It is not necessary to limit to using the above circular shape factor,
It is only necessary to be able to make a rough classification such as whether it is close to a circle or elongated. In this way, check if it is close to a circle (S
2) If it is not a circle, it is not a water drop (S9).
If it is determined that the concentration is close to a circle, a concentration profile passing through the center of the particle is obtained as shown in FIG. 2 (S3), and the concentration profile is smoothed as necessary (S3).
4) Obtain the first derivative curve of the concentration profile (S
5).

The point where the first derivative curve passes through zero is examined and n
It is assumed that individual pieces have been obtained (S6). It is checked whether or not the number n is the reference value n1 or more (S7). As described with reference to FIG. 2, the reference value n1 is set to three because there are three in the case of water droplets. When n is 3 or more, the particles are regarded as water droplets (S8), and when less than 3 particles are not water droplets (S9). This makes it possible to determine the particles as water droplets.

Next, a second embodiment will be described. In the present embodiment, unlike the first embodiment, water particles, mica, alumina foil, plastic, etc., which have no thickness and have a reflection part or white part inside, and dust having a general thickness are detected. . Mica and other thin and reflective parts have the same bright range as water droplets and are sometimes indistinguishable from water droplets, so they are detected together with water droplets. FIG. 4 shows the concentration profile y of the line A passing through the central portion and the first derivative curve y of the particle such as a water drop in which the shape of the particle is deviated from the circular shape and the bright range is eccentric, or a particle such as mica having a partially bright range. ′ Is shown. The concentration profile y shown by the broken line is the vertical axis y
Represents the concentration, and the horizontal axis x represents the position of the particle. The first-order differential curve y ′ shown by the solid line represents the first-order differential value of the concentration on the vertical axis y ′ and the position of the particle on the horizontal axis x. FIG. 5 shows the direction of the particle lines for which the concentration profile is determined. B is orthogonal to A with A as a reference, C is inclined 45 ° counterclockwise from A, and D is C
Is orthogonal to. Further, C and D may be diagonal lines of the rectangular area. In consideration of the fact that the water droplets and dust particles of this embodiment have a reflection part and a white spot apart from the center, the density profile is obtained for a plurality of lines, usually four lines A, B, C, and D. The first-order differential curve is obtained and determination is performed.

In the case of water drops or mica, since there is a bright range within the particles, when there are three or more points where the first derivative curve of the concentration profile of the particles passes through 0, as in the first embodiment, water drops and dust particles are present. To determine. However, in this case, there are a plurality of lines for obtaining the primary differential curve, and the determination is made based on the primary differential curve of any one of these lines. In addition, these particles generally have an area of a certain value or more, and when the area of a predetermined value or more and the number of points where 0 of any one of the first-order differential curves passes is three or more points, they are regarded as water drops or dust. judge. FIG.
When scanning in the B direction, there is only one point that passes 0 of the first-order differential curve, but as shown in FIG. 5, one or more of the four lines of ABCD have three or more points that pass 0. Sometimes it is determined to be a water drop or dust. FIG. 4 shows an example in which there is only one water drop or bright range of mica on the left side,
There may be more than one. At this time, the point passing through 0 of the first-order differential curve is 5 on the line passing through the two bright regions.
There are more than one. Similarly, the number of passing points increases according to the number of bright areas.

FIG. 6 is a diagram for explaining the characteristics of non-metallic inclusions (a) and dust (b). The concentration profile y of the line A passing through the central portion and the first derivative curve y'of each particle of non-metallic inclusions or dust are shown. The concentration profile y shown by the broken line represents the concentration on the vertical axis y, and the horizontal axis x represents the position of the particles. The first-order differential curve y ′ shown by the solid line represents the first-order differential value of the concentration on the vertical axis y ′ and the position of the particle on the horizontal axis x.
Since the non-metallic inclusions are at the level of the surface of the measurement sample and the microscope is focused on this surface, a clear gray-scale image can be obtained. On the other hand, since dust is raised on the surface of the sample, the dust is out of focus and looks blurred. This sharpness, that is, the degree of blurring is clearly shown in the gradient of the first-order derivative curve near the edge of the particle. An example of how to obtain the slope of the first-order differential curve will be described for non-metallic inclusions.
As shown in (a), the differential value is, for example, 25% with respect to the peak value ho of the first-order differential curve near the edge.
The distance between the point at 0.25 ho and the peak point is x, and the gradient is calculated by the following equation. Gradient = (ho-h) / x (2) The gradient of dust shown in (b) is smaller than the gradient of non-metallic inclusions shown in (a). In addition, dust generally has an area of a certain value or more, and it is possible to detect the dust from the fact that this area is a specified value or more and the gradient.

Regarding the particles to be measured, there are a plurality of lines,
Usually, four lines A, B, C, and D are provided, and the gradient at the edge of the particle of the first-order differential curve of each line is obtained from equation (2). Four
Eight edges are obtained in the line of the book. If the maximum of these gradients is less than or equal to the predetermined value P, the whole is regarded as blurred and regarded as dust. Further, even if there is a value exceeding the predetermined value P (even if there is a part that is not partially blurred), if the minimum gradient is equal to or less than the predetermined value Q, the degree of partial blur is considered to be large and dust is generated. The predetermined values P and Q are set in advance according to the composition of the particles to be measured.

FIG. 7 is an operation flow chart of the second embodiment.
First, one particle is set as a measurement region (S10). For example, the range surrounded by both ends of the X coordinate and the Y coordinate of the particle is taken. The outer circumference of the binarized particles may be used as the area. Next, it is checked whether the area of the particles is the set value A1 or more (S11). As the set value A1, 1 μm ² is used in this embodiment,
Since water drops and dust vary depending on the measurement environment, set the value according to the environment. Next, the concentration profiles for k lines are acquired (S12). As k, the four lines A, B, C, and D described above are suitable, but the number may be increased or decreased.
However, at least one is provided. Next, each density profile is smoothed (S13) and first-order differentiated (S1).
4). Next, the numbers n1, n2, ... Of the respective zero-cross points of the primary differential curves Po to Pk are detected (S15), and n
It is checked whether the maximum value of 1, n2, ... Is smaller than the set value N (S16). If it is not smaller, it is determined as a water drop or dust (S17). Note that N is usually 3, but it can be changed. This makes it possible to detect water drops as shown in FIG. 4 or dust such as mica.

Next, the primary differential curve Po near the edge of the particle
The absolute values L1, L2, ... Of the gradient of .about.Pk are obtained using the equation (2) (S18). Since the gradients of the edges at both ends of the particle are obtained for one primary differential curve, two gradients are obtained, and in the k concentration profile, absolute values of 2k gradients are obtained. Next, it is checked whether or not the maximum value of the absolute values L1, L2, ... Of the gradient is larger than the set value P (S19), and if it is not larger, it is regarded as dust (S21). If it is larger, it is checked whether the minimum value of the absolute values L1, L2, ... Is larger than the set value Q (S20). If it is not larger, it is considered as dust (S2).
1). If it is larger than the set value Q, it is determined that it is not dust or water droplets (S22). Appropriate values are set in advance for the set values P and Q according to the composition of the particles to be measured. The reason is that even if dust has a part of the edge thin and even if it is in close contact with the sample, if there are multiple scanning lines, it will float from the surface somewhere on the edge and the focus will be blurred, so the focus will be This is because it can be distinguished from inclusions that are fitted over the entire circumference.

Next, a third embodiment will be described. The present embodiment is a method for detecting scratches that occur when polishing a sample. The scratches appear as elongated particles, but they may also appear as pieces. If elongated particles are generated in the rolling direction, there is a high possibility of non-metallic inclusions, so in this embodiment it is determined that they are not scratches, and the subsequent inspection is entrusted.

FIG. 8 is an operation flow chart of this embodiment. First, it is checked whether a value obtained by dividing the maximum particle length ML by the width Bd is larger than the set value B1 (S31). The set value B1 is 30 in this embodiment. This value may also be an appropriate value depending on the shape of the scratches that occur. If it is smaller than the set value B1, it is determined that there is no scratch (S32). If the value is equal to or larger than the set value B1, it is checked whether the maximum length ML of the particles is equal to or larger than the set value B2 (S33). If the value is equal to or larger than the set value B2, it is further checked whether the arrangement direction (length direction) of the particles is the rolling direction (S34). In the case of a non-rolled material, the grain orientation direction is not the rolling direction, and the process proceeds to S35 and thereafter. The set value B2 is 20 μm in this embodiment.
However, this value may also be an appropriate value depending on the shape of the scratches that occur. In the case of S34 in the rolling direction, there is a possibility of non-metallic inclusions, so there is no flaw (S3
2) Leave it to the subsequent inspection.

If it is not in the rolling direction in S34, the line segment S extended in the length direction of the particles is set (S35). FIG. 9 is a diagram for explaining drawing when detecting a flaw, and shows a state in which a line segment S is drawn in the maximum length ML direction of the particle F. A band-shaped region L having the width of this line segment as Sk is set (S36). Line segment S
Has a range in which other particles are arranged in the length direction, and has a width Sk of about 1 to 3 μm and 1 μm in this embodiment. All particles in this band-shaped region L are scratched (S3
7). Since scratches are usually present on the same line integrally or in pieces, all the scratches in the strip-shaped region L are scratches. Particles that intersect the strip region L are not scratched.

Even when the maximum particle length ML is shorter than the set value B2 in S33, the line segment S is extended in the particle direction (S3
8), a strip-shaped region L having a line segment width of Sk is obtained (S3
9). S38 and S39 have the same contents as S35 and S36. Next, the number i of particles in the strip region L is obtained (S40).
It is checked whether or not this i is equal to or larger than the set value I (S41), and if it is smaller than the set value I (S42), it is determined that all the particles in the band-shaped region L are scratched (S43). Set value I
Is three in this embodiment. In the case of scratches, this occurs in a somewhat long range, but short particles are selected in S33, and if at least three particles are lined up in the strip region L, it is determined as a scratch. Particles that intersect the strip region L are not scratched.

Next, a fourth embodiment will be described. The present embodiment relates to a stain detection method. The stains are left behind after the water droplets are dried after the water droplets are dried, and are present together in a certain amount, and are often concentrated in a rectangle close to a square. This results from the round water droplets.

FIG. 10 is an operation flow chart of this embodiment.
FIG. 11 is a diagram showing a specific example of the operation flow of FIG. First, a v × h inspection frame is set around each particle in the visual field (S51). In this case, the particles are particles 1-10 and 10
Since there are pieces, 10 inspection frames are also set. Where v and h
Are both 20 μm. This size can also be changed according to the size of the stain. Next, an inspection frame containing m or more particles and overlapping with other inspection frames is taken out (S5).
2). Although m depends on the size of v × h, v and h are 2
In the case of 0 μm, about 10 pieces are suitable. In this case, m = 3 for ease of explanation. Then NO. 1, N
O. 8 to NO. No. 10 inspection frame no longer applicable, NO. Two
~ NO. 7 inspection frames have 3 or more. And NO. Two
NO. The 7 inspection frames overlap each other.

NO. 2 to NO. The 7 inspection frames include the following particles. NO. 2 inspection frames are particles 2, 3, 7
Have. NO. The three inspection frames have particles 1, 2, 3, 4, 6. NO. The 4 inspection frame has particles 3, 4, 5. N
O. The 5 inspection frame has particles 4, 5 and 6. NO. The 6 inspection frame has particles 3, 5, 6, 7. NO. The 7 inspection frame has particles 2, 6, 7. This NO. 2 to NO. The smallest rectangle that encloses all particles 1 to 7 included in the 7 inspection frame group is shown in FIG.
1 is set as indicated by R (S53), and since this rectangle is almost square, particles 1 to 7 within this rectangle R are used as spots (S54).

As an important function of this embodiment, the visual field image in which the foreign matter is mechanically judged as described above is stored in a storage medium, and an operator call (operation) is performed after the inspection is completed.
You can re-measure the image by calling and correct the image, or you can re-measure the field of view by calling (operating) the operator.
We are working to reduce inspection labor and improve accuracy.

[0038]

As is clear from the above description, according to the present invention, when inspecting a metallographic structure with a microscope, water drops, dust, scratches, and stains on a sample can be detected almost automatically by image processing. The burden on the inspector can be greatly reduced.

[Brief description of drawings]

FIG. 1 is a block diagram of an apparatus that realizes the present exemplary embodiment.

FIG. 2 is a diagram showing a concentration profile of a water drop and its first derivative curve.

FIG. 3 is an operation flow chart for detecting water drops.

FIG. 4 is a diagram showing a concentration profile of a water drop or dust having a bright portion at an eccentric position and its first derivative curve.

FIG. 5 is a diagram showing an example in a line direction for obtaining a concentration profile for detecting water drops or dust.

FIG. 6 is a diagram showing a concentration profile of non-metallic inclusions and dust raised on a sample and a first derivative curve thereof.

FIG. 7 is an operation flow chart for detecting dust and water droplets.

FIG. 8 is an operation flow chart for detecting a flaw.

FIG. 9 is a band-shaped region setting diagram when a flaw is detected.

FIG. 10 is an operation flow chart for detecting a stain.

FIG. 11 is a diagram showing an example of an inspection frame setting when detecting a stain.

[Explanation of symbols]

 1 Microscope 6 CPU 7 Image Processor 8 Gray Image Memory 9 Binary Memory 10 XY Auto Stage Driver 11 Auto Focus Driver 14 CRT 16 Imaging Device 17 Stage 18 Plane Moving Mechanism 19 Vertical Moving Mechanism

Claims (6)

[Claims] 1. An image of an object is picked up to generate a grayscale image, a density profile on a line passing through a central portion of a foreign substance, which appears in the grayscale image and is close to a circle, is obtained, and the first derivative curve of the density profile is 0. A foreign matter detection method by image processing, characterized in that the number of passing points is determined, and when the number is 3 or more, the foreign matter is determined to be a water drop. 2. An object is imaged to generate a grayscale image, and the area of the foreign matter appearing in the grayscale image is larger than a predetermined value,
The concentration profiles on a plurality of lines passing through the foreign matter are determined, and the number of points at which the first derivative curve of each concentration profile passes 0 is determined. When the maximum value of this number is 3 or more, the foreign matter is regarded as a water drop or dust. A foreign matter detection method by image processing characterized by making a determination. 3. An object is imaged to generate a grayscale image, and the area of the foreign matter appearing in the grayscale image is larger than a predetermined value,
By the density processing on a plurality of lines passing through the foreign matter, when the gradient at the end of the foreign matter of the first derivative curve of the density profile is less than a predetermined value, the foreign matter is determined to be dust. Foreign matter detection method. 4. An object is imaged to generate a grayscale image, and the ratio of the length and width of the foreign matter appearing in the grayscale image is equal to or greater than a first set value, and the length is equal to or greater than a second set value. There is a linear particle in a direction different from that, and a line is extended in the length direction of this particle, and a band-shaped region having the thickness of this line as k is defined as L, and all particles in this band-shaped region L are determined as scratches. A foreign matter detection method characterized by the above. 5. An object is imaged to generate a grayscale image,
There is a linear particle whose length-width ratio of the foreign matter appearing in this grayscale image is equal to or greater than the first set value and whose length is less than the second set value.
A line is extended in the length direction of the particle, and a band-shaped region having the thickness of the line as k is defined as L. When the number of particles in the band-shaped region L is a predetermined value or more, the particles in the band-shaped region L are A foreign matter detection method by image processing, which is characterized by a flaw determination. 6. An object is imaged to generate a grayscale image, and particles appearing in the grayscale image are surrounded by a rectangle h with each particle at the center.
The inspection frame of xv is set, and the inspection frame that includes m or more particles and overlaps with other inspection frames is taken out, and the smallest rectangle that encloses all particles included in this overlapped inspection frame group is set. A foreign matter detection method by image processing, which is set, and when the rectangle becomes substantially square, the particles in the rectangle are determined to be spots.