JP2000105166A - Optical member inspection device, image-processing device and method, and computer-readable medium - Google Patents

Abstract

PROBLEM TO BE SOLVED: To digitize the degree of the entire quality of a member to be inspected by normalizing the amount of measured feature according to the position for each extraction element, and by performing an operation based on a normalized value for the entire extraction elements. SOLUTION: A device for inspecting an optical member such as a lens is composed of an image pickup device consisting of an image pickup lens and an image pickup element, and a control device consisting of a CPU, memories, and the like. Then, for example, the image of a lens to be inspected is picked up by a specific method using the circumferential light of a shade, thus detecting an optical defect as a bright part for storing into the memory. A fault candidate element that is extracted by binarizing the luminance value of each pixel is judged to be dust, scratch, or the like according to width, length, or the like, and the area corresponds to, for example, regions A-D of a concentric circle with the optical axis of an optical member to be inspected as a center and at the same time is normalized by reference values S/2-8S being set for each element type. According to each normalization value for each region or type, an evaluation function for judging quality is calculated.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an optical member inspection apparatus, an image processing apparatus, an image processing method, and a computer-readable medium for detecting an optical defect such as an abnormal shape of an optical member such as a lens.

[0002]

2. Description of the Related Art An optical member such as a lens or a prism is used to refract an incident light beam regularly or to travel in parallel.
It is designed to converge or diverge at one point or linearly. However, when forming the optical member, lint or the like is mixed into the optical member (so-called “burr”), scratches or the like are generated on the surface of the optical member due to human handling after molding, and dust is generated. If it is attached, the incident light beam will be disturbed, so that desired performance cannot be obtained.

For this reason, various optical member inspection apparatuses have been proposed which detect a cause of a defect in an optical member and automatically judge the quality of the optical member. This type of optical member inspection apparatus generally captures an image of an optical member by a method in which the cause of the defect is imprinted on the image data, and performs binarization processing on the obtained image data to thereby obtain the defect factor. Was extracted. Then, when the area of any of the regions indicating the cause of failure exceeds a predetermined determination threshold, the inspection target optical member is determined to be defective.

[0004]

However, according to the above-described conventional determination algorithm, the information obtained as a result of the inspection is only whether the optical member to be inspected is a non-defective or defective product. Therefore, for example, in an optical member in which only one failure factor having an area exceeding the determination threshold is formed and in an optical member in which many such failure factors are formed, it can be said that the degree of failure is greater in the latter than in the former. However, according to the above-described conventional determination algorithm, it is not possible to evaluate the degree of such a defect. Further, the optical member formed with only one defect factor having an area slightly exceeding the determination threshold and the optical member formed with a large number of defect factors having an area slightly smaller than the determination threshold, as the degree of failure, The latter may exceed the former in some cases, but according to the above-described conventional determination algorithm, the former is always determined to be defective and the latter is determined to be good. As described above, according to the above-described determination algorithm, it is not possible to evaluate the degree of failure of the optical member to be inspected. If the degree of failure of the optical member to be inspected cannot be evaluated in this way, even if the statistical processing of the inspection result for a predetermined number of optical members is performed,
Only the breakdown of the types of failure factors that have exceeded the determination threshold of the optical member can be clarified.

In view of the above-mentioned problems in the prior art, the present invention considers the location of each defect factor and evaluates the influence of the defect factor on the performance of the entire optical member to be inspected. An object of the present invention is to provide an optical member inspection device, an image processing device, an image processing method, and a computer-readable medium capable of quantifying the degree of overall defect or non-defective optical member.

[0006]

The invention described in each claim is
It has been made to solve the above problems.

That is, in the optical member inspection apparatus according to the first aspect, an image pickup apparatus that images an optical member to be inspected and outputs image data including an image of the optical member to be inspected, and a surrounding in the image data. Extraction means for extracting portions having different luminances as extraction elements; graphic feature quantity measurement means for measuring the graphic feature quantity of each of the extracted elements; and position for measuring the position of each of the extracted elements in the image data A normalizing unit that normalizes a graphic feature amount measured by the graphic feature amount measuring unit for each of the extracted elements according to the position measured by the position measuring unit to obtain a normalized value. Means, and arithmetic means for executing a predetermined evaluation function based on all normalized values obtained by the normalizing means for all extracted elements. .

With such a configuration, the graphic feature of each extracted element extracted from the image data by the extracting means is normalized by the normalizing means in accordance with the position of the extracted element in the image data. You. At this time, the normalizing means has a relatively large normalized value of the graphic feature amount of the extracted element caused by an optical defect formed at a position that greatly affects the performance of the inspection target optical member, and Normalization can be performed such that the graphic feature amount of the extracted element caused by an optical defect formed at a position where the effect on performance is small has a relatively small normalized value. Therefore, when the calculating means executes the evaluation function with respect to the normalized values of all the extracted elements extracted by the extracting means, a numerical value indicating the degree of overall failure of the inspection target optical member is obtained.

The graphic feature quantity measured by the graphic feature quantity measuring means may be, specifically, the area of the extracted element, the maximum fillet diameter, or the average brightness. Or the maximum luminance. Further, the position measuring means may measure the distance itself from the position corresponding to the optical axis of the optical member to be inspected to the extracted element, or any one of a plurality of concentric regions around the optical axis. It may be identified whether the extracted element is located within the. Further, when the position measurement means measures the distance itself, the normalization means may perform normalization by executing some function on the distance and the graphic feature amount. Is identified, the graphic feature quantity may be normalized based on a reference value previously associated with the identified area.

According to a second aspect of the present invention, there is provided an optical member inspection apparatus, wherein the imaging apparatus of the first aspect illuminates the inspection target optical member from outside the imaging optical axis to thereby remove the optical defect from the inspection target optical member. This is specified by generating scattered light and outputting image data in which this optical defect is reflected as a bright portion.

According to a third aspect of the present invention, in the optical member inspection apparatus, the extracting means of the first or second aspect binarizes the luminance value of each pixel constituting the image data by comparing the luminance value with a predetermined threshold value. It is specified by extracting the extraction element.

According to a fourth aspect of the present invention, in the optical member inspection apparatus, the graphic feature measuring means of the first aspect specifies the area by measuring the area of each of the extracted elements.

According to a fifth aspect of the present invention, the position measuring means of the first aspect measures the position of each of the extracted elements with reference to a position corresponding to the optical axis of the optical member to be inspected. It is specified.

According to a sixth aspect of the present invention, in the optical member inspection apparatus, the normalizing means according to the fifth aspect is configured such that the position of the extracted element measured by the position measuring means is determined from a position corresponding to the optical axis of the inspection target optical member. A reference value is set for each of the extracted elements so as to increase as the distance increases, and the normalized value is obtained by dividing the graphical feature of the extracted element by the set reference value. It is.

According to a seventh aspect of the present invention, in the optical member inspecting apparatus, the position measuring means according to the sixth aspect is arranged such that each of the plurality of concentric regions centered on a position corresponding to the optical axis of the optical member to be inspected. Whether or not the extracted element is located is measured, and the normalization means specifies the reference value by setting the reference value corresponding to each of the plurality of concentric regions.

According to an eighth aspect of the present invention, in the optical member inspection apparatus, the arithmetic means of the seventh aspect adds the normalized value of each of the extracted elements obtained by the normalizing means for each of the regions, and It is specified by calculating the square root of the sum of squares of the addition result for each area.

According to a ninth aspect of the present invention, in the optical member inspection apparatus, the arithmetic means of the eighth aspect classifies the extracted elements according to the type of optical defect that has caused the extracted elements, and for each type, calculates the sum of squares. This is specified by calculating the square root and further calculating the square root of the sum of squares of the calculation result for each type.

According to a tenth aspect of the present invention, in the optical member inspection apparatus according to the ninth aspect, the calculation means of the ninth aspect sets the normalized value less than a predetermined value at an interval within a predetermined distance from an extraction element corresponding to the normalized value. Unless a predetermined number or more of the extracted elements are connected, this is specified by not performing the addition.

According to an eleventh aspect of the present invention, in the optical member inspection apparatus according to the ninth aspect, the arithmetic means according to the ninth aspect is configured such that at least a predetermined number of extracted elements are separated from each other by a predetermined distance from the extracted elements corresponding to the normalized values less than the predetermined value. If they are connected, a value obtained by multiplying the number of the extracted elements connected at intervals within the predetermined distance by a certain value is set as a normal value for all the extracted elements, and the normal value for the entire extracted elements is set as a normal value. It is specified by classifying independently from other normal values.

According to a twelfth aspect of the present invention, in the optical member inspection apparatus according to the first aspect, when the value obtained by executing the evaluation function by the calculation means exceeds a predetermined pass / fail judgment reference value, The optical member is specified by further including a pass / fail determination unit that determines that the optical member is defective.

According to a thirteenth aspect of the present invention, the optical member inspection apparatus is specified by measuring the area and luminance of each of the extracted elements by the graphic feature quantity measuring means of the first aspect.

According to a fourteenth aspect of the present invention, the optical member inspection apparatus is specified by measuring the position of the center of gravity of each of the extracted elements by the position measuring means of the first aspect.

According to a fifteenth aspect of the present invention, a portion having a luminance different from the surroundings in image data including an image of the inspection target optical member obtained by imaging the inspection target optical member is extracted as an extraction element. Extracting means, a graphical feature quantity measuring means for measuring a graphical feature quantity of each of the extracted elements, a position measuring means for measuring a position of each of the extracted elements in the image data, and A normalizing unit that normalizes the graphic feature amount measured by the graphic feature amount measuring unit according to the position measured by the position measuring unit to obtain a normalized value; and Calculating means for executing a predetermined evaluation function based on all normalized values obtained by the converting means.

In the image processing method according to the present invention, a portion having a brightness different from the surroundings in image data including an image of the optical member to be inspected obtained by imaging the optical member to be inspected is extracted as an extraction element. Measuring the graphical feature of each of the extracted elements, measuring the position of each of the extracted elements in the image data, and for each of the extracted elements, calculating the graphical feature of the extracted element in the image data. Is characterized in that a normalized value is obtained by normalizing according to the position in and a predetermined evaluation function is executed based on all the normalized values obtained for all the extracted elements.

According to a seventeenth aspect of the present invention, there is provided a computer readable medium having a portion having a brightness different from the surroundings in image data including an image of the optical member to be inspected obtained by imaging the optical member to be inspected. Is extracted as an extraction element, the graphical feature amount of each of the extracted elements is measured, the position of each of the extracted elements in the image data is measured, and the graphical feature amount of each of the extracted elements is calculated as the image data. This stores a program that causes a normalized value to be obtained by normalizing according to the position in the inside and executes a predetermined evaluation function based on all the normalized values obtained for all the extracted elements.

[0026]

Embodiments of the present invention will be described below with reference to the drawings.

[0027]

First Embodiment <Configuration of Optical Member Inspection Apparatus> A schematic configuration of an optical member inspection apparatus according to a first embodiment of the present invention is shown in a side sectional view of FIG. As shown in FIG. 1, the illumination lamp 1, the diffusion plate 2, and the imaging device 3 that constitute the optical member inspection device are arranged on the same optical axis l.

The image pickup device 3 includes an image pickup lens 4 which is a positive lens system, and an image pickup device 5 which picks up an image by light converged by the image pickup lens 4.
In FIG. 1, the image sensor 5 is installed so that its pixel row is directed to the left and right. The pixel array of the image sensor 5 intersects with the optical axis 1 of the image pickup lens 4 perpendicularly at the center. Note that the imaging lens 4 is movable forward and backward with respect to the imaging device 5 in the imaging device 3 (focus adjustment is possible).
The imaging device 3 itself is also mounted on a frame (not shown) of the optical member inspection device so that the imaging device 3 itself can be advanced and retracted in the optical axis l direction.

The image pickup element 5 repeatedly picks up an image in a line at predetermined time intervals (a time period in which electric charges are appropriately accumulated in each pixel), and scans each pixel in the order in which the pixels are arranged. Output the accumulated charge. The charges output from the image sensor 5 in this way are subjected to predetermined amplification processing and A / D conversion processing, and then input to the image processing device 6 as image data composed of luminance signals for one line.

The optical member 14 to be inspected is a circular lens as shown in FIG. 1 and FIG. 2 which is a plan view seen from the side of the image pickup device 3, and is attached to a frame (not shown) of the optical member inspection device. The imaging lens 4 is held by the holder 15 so that the surface thereof (the surface facing the imaging lens 4) is conjugate with the imaging surface of the imaging element 5. The holder 15 has an annular shape centered on a central axis O that is offset in parallel with the optical axis l of the imaging lens 4, and holds the periphery of the optical member 14 to be inspected over the entire periphery thereof. I do.

The holder 15 is rotatable about a central axis O in a plane perpendicular to the optical axis l. An annular gear 16 is formed on the periphery of the holder 15. This annular gear 16 meshes with the pinion gear 7 attached to the drive shaft of the drive motor 8. Accordingly, when the drive motor 8 rotates its drive shaft, the holder 15 is rotationally driven via the two gears 7 and 16, and the inspection target optical member 14 held by the holder 15 is orthogonal to the optical axis l. It is rotationally driven in the plane.

The magnification of the image pickup lens 4 (ie, the position of the image pickup apparatus 3 itself and the position of the image pickup lens 4 with respect to the image pickup element 5) is determined by the total width of the inspection target optical member 14 in the radial direction. Is adjusted so that an image can be formed. Therefore, the image pickup device 5 includes the inspection target optical member 1
It is possible to capture an image of the surface of No. 4 (an image for one line over the entire area in the radial direction). In FIG. 2, an imaging target area for one line that can be imaged by the imaging element 5 is indicated by a two-dot chain line.

The illumination lamp 1 is an incandescent lamp that emits illumination light (white light), and is fixed to a frame (not shown) of the optical member inspection device.

The illumination lamp 1 and the optical member 14 to be inspected
As shown in FIG. 2, the diffusion plate 2 disposed between them has a disk shape having a diameter larger than the radius of the inspection target optical member 14, and the surface thereof is processed as a rough surface. . Accordingly, the diffusion plate 2 can receive the illumination light emitted from the illumination lamp 1 on the entire back surface and diffuse the illumination light toward the optical member 14 to be inspected. The diffusion plate 2 is fixed to a frame (not shown) of the optical member inspection device so that the center of the diffusion plate 2 is orthogonal to the optical axis 1 of the imaging lens 4.

A light-shielding plate 9 having a band shape is attached on the surface of the diffusion plate 2 with its longitudinal direction oriented parallel to the direction of the pixel rows of the image sensor 5. The center of the light shielding plate 9 coincides with the optical axis l of the imaging lens 4. The total length of the light shielding plate 9 in the longitudinal direction is longer than the radius of the optical member 14 to be inspected. Further, as shown in FIG. 4 which is a cross-sectional view of the optical member inspection apparatus in a direction orthogonal to the direction of the pixel rows of the image sensor 5, the width of the light shielding plate 9 is the margin of light incident on each pixel of the image sensor 5. It is wider than the interval between the light beams m and m.

The control device 6 determines whether the optical member 14 to be inspected is a non-defective product or a defective product based on the image data input from the image pickup device 3, and in accordance with this determination, the drive motor 8 This is a processing device that supplies a drive current.

FIG. 3 is a block diagram showing the internal circuit configuration of the control device 6. As shown in FIG.
Is composed of a CPU 60, a frame memory 61, a host memory 62, and a motor drive circuit 63, which are interconnected via a bus B.

The frame memory 61 is a buffer in which image data input from the image pickup device 3 is written.

The host memory 62 has an image memory area 62
a, work memory area 62b, classification table storage area 62
c and an image processing program storage area 62d. The image memory area 62a is an area in which the image data written in the frame memory 61 is written every predetermined time from the top row in row units. The image data written in the image memory area 62a is image data (polar coordinate data) based on a polar coordinate system because of the imaging method of the imaging device 3. Further, the working memory area 62b is transcribed in a state where the image data (polar coordinate data) in the image memory area 62a is subjected to coordinate conversion (polar coordinate-rectangular coordinate conversion), and is binarized according to a predetermined threshold. ,
This is an area in which the evaluation of “bad candidate elements (extracted elements extracted as a group of pixels having a high luminance value) in the binarized image data obtained by the binarization. Classification table storage area 62c Is an area for storing a classification table shown in Fig. 8. Details of this classification table will be described later, and an image processing program storage area 62d stores an image processing program executed by the CPU 60. It is an area as a computer-readable medium.

The motor drive circuit 63 supplies the drive motor 8 with a drive current for driving the drive motor 8 so that the holder 15 and the optical member 14 to be inspected rotate at a constant speed in the clockwise direction when viewed from the imaging device 3 side. I do.

The CPU 60 is a computer that controls the entire control device 6, and corresponds to an extracting unit, a graphic feature measuring unit, a position measuring unit, a normalizing unit, a calculating unit, and a pass / fail determining unit. That is, the CPU 60 executes the image processing program stored in the image processing program storage area 62d of the host memory 62, and periodically writes the image data written in the frame memory 61 to the image memory area 62a of the host memory 62. At the same time, when image data (polar coordinate data) corresponding to the entire inspection target optical member 14 is synthesized in the image memory area 62a,
This image data is subjected to a coordinate conversion process, and the image data (rectangular coordinate data) after the coordinate conversion is stored in the work memory
Copy to 2b. Then, by binarizing this image data (orthogonal coordinate data), defective candidate elements are extracted, the area of each extracted defective candidate element is normalized according to its position, and classified according to various criteria described later. , Are totaled in the classification table in the classification table storage area 62c.
After the classification and totalization of all the defective candidate elements are completed, an evaluation function F described later is calculated based on the totalization result, and based on whether the evaluation function F exceeds a predetermined determination reference value. The quality of the inspection target optical member 14 is determined. Note that the CPU 60 instructs the motor drive circuit 63 to supply a drive current to the drive motor 8 in synchronization with the capture of image data from the frame memory 61.

The motor drive circuit 63 supplies a drive current for driving the drive motor 8 so that the holder 15 and the optical member 14 to be inspected rotate clockwise at a constant speed when viewed from the imaging device 3 side. I do. <Principle of Detecting Optical Defect> In the optical member inspection apparatus configured as described above, in the plane of FIG.
Is incident on each pixel of the imaging element 5 and is a light beam having a light ray along the optical axis l of the imaging lens 4 as a principal ray, and passes between the marginal rays m, m shown in FIG. Only light. When the marginal rays m, m are traced in the opposite direction, they intersect at the surface of the inspection target optical member 14 and then spread toward the diffusion plate 2. Then, on the diffusion plate 2,
The space between the marginal rays m, m is blocked by the light shielding plate 9. Therefore, as shown in FIG.
Area (image pickup lens 4)
Assuming that there is no optical defect at (a part conjugate to the light receiving surface of the pixel array of the image sensor 5 and its vicinity in the optical axis direction), no light is incident on each pixel of the image sensor 5. That is, the light n diffused from the side of the light-shielding plate 9 on the surface of the diffusion plate 2 passes through the imaging target area in the inspection target optical member 14, but passes outside the marginal rays m, m. Does not enter. Further, the light is diffused from the side of the light shielding plate 9 on the surface of the diffusion plate 2 to be inspected.
The light transmitted through a portion other than the imaging target area in, can enter the imaging lens 4, but is not converged on each pixel of the imaging element 5. Therefore, the image data output from the imaging device 3 is dark over the entire area except for a bright portion (due to diffused light on the side surface) corresponding to the outer edge of the inspection target optical member 14.

On the other hand, as shown in FIG. 2, when there is a flaw β and dust γ in the imaging target area on the surface of the optical member 14 to be inspected, as shown in FIG. When the light n diffused from the side portion of the plate 9 hits these flaws β and dust γ, this light
And dust γ. The diffused light n ′ diverges around the intersection of the marginal rays m, m, and a part of the light is incident on the pixels of the image sensor 5 via the image pickup lens 4. Accordingly, images of the scratches β and dust γ (images brighter than the surroundings) are formed on the imaging surface of the imaging element 5.

The image pickup (charge accumulation and scanning) by the image pickup device 5 is performed every time the inspection target optical member 14 rotates by a predetermined angle in synchronization with the rotation of the inspection target optical member 14 by the drive motor 8. Then, every time imaging (charge accumulation and scanning) is performed by the imaging element 5, the linear image data is written into the frame memory 61 of the control device 6 and taken into the image memory area 62 a of the host memory 62. As a result, as the inspection target optical member 14 rotates, each line of image data captured by the imaging device 3 is sequentially written in each row of the image memory area 62a from the top row.

Since the image data stored in the image memory area 62a of the host memory 62 at the time when the optical member 14 to be inspected makes one rotation is polar coordinate data, it is subjected to coordinate conversion (polar coordinate-rectangular coordinate conversion) processing. , Are converted into orthogonal coordinate data as shown in FIG. In the rectangular coordinate data, the shape of the bright part (failure candidate element) indicating the cause of the defect matches the shape of the optical defect that caused the cause. <Principle of Quality Evaluation of Inspection Optical Member> The CPU 60 compares the luminance value of each pixel constituting the image data in the image memory area 62a with a predetermined threshold value, and determines a pixel having a luminance value exceeding the threshold value. Binarization is performed by giving a new luminance value “255” and giving a new luminance value “0” to a pixel having a luminance value equal to or less than the threshold value. Note that the threshold value is set to a value higher than the brightness value of a bright portion caused by noise but lower than the brightness value of a bright portion caused by an optical defect. As a result, from the image data,
Only defective candidate elements are extracted.

The defective candidate elements are classified according to the positions where they are formed. That is, as shown in FIG. 6, the inspection target optical member 14 is divided into four regions A to D in a concentric manner centered on a position corresponding to the optical axis. In the area A, the optical defect formed on the area A has the greatest effect on the performance of the inspection target optical member 14, and the optical defects of the area A will be described below in the order of B, C, and D. The effect on performance is getting smaller.

Therefore, in the present embodiment, for each region, the area of a defect candidate element to be determined as a defect occurring in the region is divided by a reference value R prepared for the region. It is normalized and converted into a normalized value (a score indicating the degree of influence on the performance of the optical member to be inspected) P. Specifically, when the reference value R to be applied to the region A is “S (= determination reference value used in the conventional determination algorithm)”, as shown in FIG. “2S” is applied as the reference value R, “4S” is applied as the reference value R for the region C, and the region D is applied.
Is applied as “8S” as the reference value R. That is,
The reference value R is applied such that the position of the defective candidate element increases as the position moves away from the position corresponding to the optical axis of the inspection target optical member 14. The calculated normalized value P is “1” if the area of the defective candidate element is the same as the reference value R.
Therefore, the defective candidate element having the area S generated in the area A, the area B
In the case of the defective candidate element having the area 2S generated in the above, the defective candidate element having the area 4S generated in the area C, and the defective candidate element having the area 8S generated in the area D, the calculated normalized value P is " 1 ". As described above, the area of each defective candidate element is converted into a normalized value P indicating the degree of influence on the performance of the inspection target optical member 14 by normalizing as described above.

As the reference value R, the reference value Rk is applied when the defective candidate element is caused by a flaw, and the reference value Rd is applied when the defective candidate element is caused by dust. These reference values Rk and Rd are different from each other. That is, the reference value Rk applied to the defective candidate element caused by the flaw formed in the area A is “Sk”, and the reference value Rd applied to the defective candidate element caused by the dust formed in the area A is “Sk”. Sd ", these Sk and Sd
Have different values from each other. Therefore, the CPU 60
Determines whether the defective candidate element to be processed is due to scratches or dust before calculating the normalized value P.

This determination is made using a threshold function shown in FIG. Specifically, the CPU 60 measures the maximum width (X fillet) in the X-axis direction (horizontal direction in FIG. 6) and the maximum width (Y fillet) in the Y-axis direction (vertical direction in FIG. 6) of each defective candidate element. Then, the ratio (filler ratio) of the smaller of the X fillet and the Y fillet to the other is calculated according to the following equation (1).

Filler ratio = small fillet / larger fillet × 100 (1) On the other hand, the CPU 60 executes the following equation (2) for the X fillet and the Y fillet and the area of the defective element candidate. ,
Calculate the occupancy.

Occupancy = Area of defective element candidate / (X fillet × Y fillet) × 100 (2) Then, the CPU 60 compares the calculated fillet ratio and occupancy with the threshold function shown in FIG. It is determined whether the defective candidate element is caused by dust or a flaw. That is,
If the intersection of the fillet ratio and the occupancy is located above the threshold function on the graph of FIG. 9, the CPU 60 determines that the defective candidate element is caused by dust, and the intersection of the fillet ratio and the occupancy is determined. If it is located below the threshold function, it is determined that the defective candidate element is caused by the flaw.

In this manner, when it is determined whether the defective candidate element is caused by a scratch or dust, the CPU 60
Calculates a reference value Rk or Rd according to the determined type of the scratch or dust and the position of the area where the defective candidate element has occurred. Then, the above-described normalized value P is calculated based on the obtained reference value Rk or Rd.

Next, if the normalized value P calculated as described above is 0.5 or more, the CPU 60 determines the determined dust in the classification table stored in the classification table storage area 62c. The normalized value P calculated for the defective candidate element is added to a column corresponding to the distinction between (ds) or flaw (k) and the area A to D where the defective candidate element has occurred.

If the calculated normalized value P is less than 0.5, it is assumed that the optical defect that caused the defect is not large enough to affect the performance of the optical member 14 to be inspected. In principle, it is not added to the classification table.
However, when a plurality of small defective candidate elements in which the calculated normalized value P is less than 0.5 are consecutively formed at an interval within a predetermined distance, a bright portion caused by the same optical defect is binary. It is highly probable that it has been divided into a plurality of defective candidate elements in the process of conversion. Therefore, when such a small defective candidate element is detected, another defective candidate element within a certain distance from the defective candidate element is searched for. When a defective candidate element is found in this way, another defective candidate element within a certain distance from the defective candidate element is further found. The search for such defective candidate elements is continuously performed until another defective candidate element cannot be found within a certain distance from the last found defective candidate element. If the number of defective candidate elements found within a certain distance from each other is three or more, the number of defective candidate elements is 0.25.
Is added to the column of concentration (m) corresponding to the areas A to D in which the defective candidate element has occurred in the classification table. In other words, when a predetermined number (two or more) of defective candidate elements continue at intervals within a predetermined distance from a defective candidate element for which a normalized value P less than the predetermined value is calculated, the defective candidate elements are within the predetermined distance. Number of defective candidate elements (3 or more)
Is multiplied by a constant value (0.25) as a normalized value P corresponding to the entirety of these defective candidate elements, and is classified independently of the defective candidate elements for which a score of 0.5 or more is calculated. .

When the addition to the classification table as described above is performed for all the defective candidate elements, the CPU 60 first calculates the normalized value P written in all the columns corresponding to dust (ds) in the classification table. The square root of the sum of squares of the sum is calculated according to the following equation (3), and the square root of the sum of the squares of the sum of the normalized values P similarly written in all the columns corresponding to the flaw (k) is calculated according to the following equation (4). And the normalized values P written in all the columns corresponding to the concentration (m)
Is calculated according to the following equation (5). Then, the square root of the further sum of squares of the values calculated for each is calculated as the evaluation function F according to the following equation (6).

Ds = ((dsA) ² + (dsB) ² + (dsC) ² + (dsD) ² ) ^0.5 (3) k = ((kA) ² + (kB) ² + (kC) ² + ^{(kD) 2) 0.5 ...... (} 4) m = ((mA) 2 + (mB) 2 + (mC) 2 + (mD) 2) 0.5 ...... (5) F = (ds 2 + k 2 + m 2) ^0.5 (6) The value of the evaluation function F is a value indicating the overall performance of the entire optical member 14 to be inspected, and the larger the value, the worse the performance. Therefore, the value of the evaluation function F is compared with a predetermined pass / fail judgment reference point. If the value of the evaluation function F is less than the pass / fail judgment reference point, the inspection target optical member 14 is checked.
Are judged to be non-defective, and if the value of the evaluation function F is equal to or more than the pass / fail judgment reference point, the inspection target optical member 14 is judged to be defective. As described above, the evaluation function F further calculates the square root of the sum of squares of the sum of the normalized values P calculated for each type of the failure factor.

The sum of the normalized values P and the value of the evaluation function F written in the classification table are stored in a disk device (not shown) as the inspection result of the optical member 14 to be inspected. . Then, the quality is statistically managed by obtaining an average value, a standard deviation, a histogram, and the like for the inspection results stored for each inspected optical member for each lot. Thereby, in addition to the overall quality of the optical member 14 to be inspected, statistical processing can be performed for each area and each type of the cause of failure, which is very effective in quality control of the manufacturing process. <Control Processing> Next, in order to make a pass / fail judgment based on the above-described principle of defect factor detection and quality evaluation, the control device 6 (CPU 60) executes the image processing program read from the image processing program storage area 62d. The contents of the control processing to be performed will be described with reference to the flowcharts of FIGS.

The control process shown in FIG. 10 is started when an unillustrated test start button connected to the control device 6 is pressed. In the first S01 after the start, the CPU 60
Instructs the motor drive circuit 63 to supply a drive current to the drive motor 8, and rotates the inspection target optical member 14 at a constant speed.

In the next S02, the CPU 60 stores the image data written from the imaging device 3 into the frame memory 61 in the image memory area 62a of the host memory 62.

In the next S03, the CPU 60 checks whether or not the image data (polar coordinate data) corresponding to the entire inspection optical member 14 has been combined in the image memory area 62a by storing the image data in S02. If the image data (polar coordinate data) corresponding to the entire inspection target optical member 14 has not been synthesized, the process returns to S02, and the imaging device 3 writes the image data obtained by the next imaging into the frame memory 61. Wait for. On the other hand, if the polar coordinate data corresponding to the entire inspection target optical member 14 has been synthesized, the process proceeds to S04.

In S04, the CPU 60 performs a coordinate conversion (polar coordinate-rectangular coordinate conversion) process on the image data (polar coordinate data) stored in the image memory area 62a to convert the image data into rectangular coordinate data. The image data (orthogonal coordinate data) is stored in the work memory area 62b.

In the next S05, the CPU 60 performs the above-described binarization processing on the image data in the working memory area 62b to obtain binarized data including the emphasized defective candidate element (extraction means). Equivalent). That is, CPU
Reference numeral 60 rewrites the luminance value of a pixel having a luminance value equal to or less than a predetermined threshold value to “0” and rewrites the luminance value of a pixel having a luminance value exceeding the threshold value to “255”.

In the next S06, the CPU 60 executes a masking process for removing the outer edge of the inspection target optical member 14 (the outer edge of the area D in FIG. 6) in the image data (corresponding to the extracting means).

In the next S07, the CPU 60 assigns a unique number (label) n (n = 1, 2, 3,...) To each of the defective candidate elements extracted in the binary data. (Corresponding to the extraction means). Specifically, the CPU 60 searches the luminance value of each line sequentially from the top line in the masked binary data, and receives the luminance value (= 255) corresponding to the defective candidate element. When a pixel is detected, the luminance value of that pixel is identified by a unique label n
To the value of However, if the detected pixel is adjacent to a pixel whose luminance value has already been rewritten to the value of the label n (≠ 0, 255), the CPU 60 determines the luminance value of the detected pixel as an adjacent pixel. Brightness value (n
{0, 255).

Next, the CPU 60 executes a loop process from S08 to S26 in order to calculate the normalized value P for each defective candidate area in the order of the labels n added in S07. In the first S08 after entering this loop, the CPU 6
0 specifies the smallest label among unspecified labels n as the label to be processed.

In the next S09, the CPU 60 determines whether the defective candidate element with the label n specified in S08 is due to dust or a flaw as described above. If it is determined that the defective candidate element is caused by dust, the CPU 60 proceeds from S10 to S1.
Proceed to 1. On the other hand, when it is determined that the defect factor is caused by the scratch, the CPU 60 advances the processing from S10 to S16.

In S11, the CPU 60 obtains the center of gravity of the defective candidate element with the label n specified in S08,
Among the respective areas in the binarized data, the area where the obtained center of gravity exists is determined (corresponding to a position measuring means).

In the next S12, the CPU 60 sets a reference value Rd according to the area determined in S11. Specifically, when the determined area is the area A, the reference value R
d is set to Sd, and when the determined area is the area B, the reference value Rd is set to 2Sd. When the determined area is the area C, the reference value Rd is set to 4Sd. In the case where the obtained region is the region D, the reference value Rd is set to 8Sd
Set as

In the next S13, the CPU 60 measures the area of the defective candidate element with the label n specified in S08. That is, the total number of pixels constituting the defective candidate element is counted. Here, a graphic feature amount other than the area, for example, a maximum fillet diameter may be measured (corresponding to a graphic feature amount measuring unit).

In the next S14, the CPU 60 divides the area of the defective candidate element measured in S13 by the reference value Rd set in S12 to obtain a numerical value (normalized value) P obtained by normalizing this area. (Corresponding to normalization means).

In the next S15, the CPU 60 checks whether or not the normalized value P obtained in S14 is 0.5 or more. If the normalized value P is 0.5 or more, CP
In S21, U60 stores the classification table storage area 62
After adding the normalized value P obtained in S14 to the dust (ds) column corresponding to the area determined in S11 in the classification table stored in c, the process proceeds to S26. On the other hand, when determining in S15 that the normalized value P is less than 0.5, the CPU 60 advances the processing to S23.

On the other hand, in S16, the CPU 60 executes S08
The center of gravity of the defective candidate element with the label n specified in is determined, and among the regions in the binarized data, the region where the determined center of gravity is located is determined (corresponding to a position measuring unit).

In the next S17, the CPU 60 sets a reference value Rk according to the area determined in S16. Specifically, when the determined area is the area A, the reference value R
k is set to Sk, and when the determined area is the area B, the reference value Rk is set to 2Sk. When the determined area is the area C, the reference value Rk is set to 4Sk. In the case where the set area is the area D, the reference value Rk is set to 8Sk
Set as

In the next S18, the CPU 60 measures the area of the defective candidate element with the label n specified in S08. That is, the total number of pixels constituting the defective candidate element is counted. Here, a graphic feature amount other than the area, for example, a maximum fillet diameter may be measured (corresponding to a graphic feature amount measuring unit).

In the next S19, the CPU 60 divides the area of the defective candidate element measured in S18 by the reference value Rk set in S17 to obtain a numerical value (normalized value) P obtained by normalizing this area. (Corresponding to normalization means).

In the next S20, the CPU 60 checks whether or not the normalized value P obtained in S19 is 0.5 or more. If the normalized value P is 0.5 or more, CP
In S22, U60 determines whether the classification table storage area 62
After adding the normalized value P obtained in S19 to the flaw (k) column corresponding to the area determined in S16 in the classification table stored in c, the process proceeds to S26. On the other hand, when determining in S20 that the normalized value P is less than 0.5, the CPU 60 advances the processing to S23.

In S23, the CPU 60 checks whether or not two or more other defective candidate elements are connected at an interval within a predetermined distance from the defective candidate element with the label n specified in S08. If two or more other defective candidate elements are not continuous at intervals within a predetermined distance, the CP
U60 advances the process directly to S26. On the contrary,
If two or more other defective candidate elements are continuous at intervals within the predetermined distance, the CPU 60 determines in S24 that S0
A numerical value obtained by multiplying the total number of defective candidate elements with the label n specified in 8 and the total number of other defective candidate elements (the number of concentrated elements) connected at intervals within a predetermined distance by 0.25 by 0.25 In the classification table stored in the storage area 62c, the value is added to the concentration (m) column corresponding to the area determined in S09. In the next S25, the CPU 60 treats the label attached to the defective candidate element determined to be continuous at an interval within the predetermined distance in S23 as the specified treatment. After the completion of S25, the CPU 60 advances the processing to S26.

In S26, it is checked whether or not all the labels n added in S07 have already been specified. If all the labels n have not yet been specified, the CPU 6
If it is 0, the process returns to S08. On the other hand, as a result of repeating the loop processing from S08 to S26, all the labels n
If has been specified, the CPU 60 advances the processing to S27.

In S27, the CPU 60 calculates the value of the evaluation function F with respect to the sum of the normalized values P described in each column of the classification table using the above-described equations (3) to (6) (operation Means).

In the next step S28, the CPU 60 checks whether or not the value of the evaluation function F obtained in step S27 is less than a predetermined pass / fail judgment reference value. If the value of the evaluation function F is smaller than the pass / fail judgment reference value, the CPU 6
In the case of 0, in S29, it is determined that the inspection target optical member 14 is a non-defective product, and the fact is externally output (image display, audio output). On the other hand, when the value of the evaluation function F is equal to or larger than the pass / fail determination reference value, the CPU 60 determines in S30 that the inspection target optical member 14 is defective, and outputs the fact to the outside ( Image display, sound output). In any case, the CPU 60 thereafter ends this control processing. <Operation of the Embodiment> For the sake of easy understanding, it is assumed that only one type of defective candidate element is formed in the areas A and B of the optical member 14 to be inspected. In this case, the equation for calculating the evaluation function F is represented by the following equation (7).

F = (A ² + B ² ) ^0.5 (7) where A is the normalized value P of the defective candidate element caused by any type of optical defect formed in the area A. B is the sum of the normalized values P of the defective candidate elements caused by the same type of optical defect formed in the area B. At this time, assuming that the pass / fail judgment reference point is 1.2, the range of the combination of A and B when the inspection target optical member 14 is determined to be non-defective, and the range of the combination when A and B are determined to be defective. The range of the combination of A and B is divided as shown in FIG. On the other hand, for the sake of comparison, the division of each range in the case where a pass / fail judgment is made based on whether one of the sums A and B calculated for each region exceeds 1 without using the evaluation function F is shown in FIG. 21.

As is apparent from FIG. 13, when the sum of the normalized values P in the area A and the sum of the normalized values P in the area B are both slightly smaller than 1 (that is, when the sum of the normalized values P corresponds to the hatched portion α). In this case, according to the evaluation method shown in FIG. 21, the inspection target optical member 14 is determined to be non-defective. However, when the inspection target optical member 14 is viewed as a whole, it is considered that its performance has deteriorated. In this regard, according to the present embodiment, since the value of the evaluation function F exceeds the pass / fail determination reference point 1.2, the inspection target optical member 14 is determined to be defective. Conversely, when the sum of the normalized values P in the area A slightly exceeds 1, but the sum of the normalized values P in the area B is close to 0, or when the sum of the normalized values P in the area B is 1 If it slightly exceeds, but the sum of the normalized values P of the area A is close to 0 (that is, if it corresponds to the hatched portion β), the optical member 14 to be inspected is determined by the evaluation method shown in FIG. The product is determined to be defective. However, if the inspection target optical member 14 is viewed as a whole, it is considered that it can be regarded as a non-defective product. In this regard, according to the present embodiment, since the value of the evaluation function F is lower than the pass / fail determination reference value 1.2, the inspection target optical member 14 is determined to be non-defective.

In addition, the sum of the above-described normalized values P is
This is a value normalized so as to indicate the degree to which the defect candidate element generated in each area affects the performance of the inspection target optical member 14. Therefore, according to the present embodiment, the inspection target optical member 14 is evaluated as a whole, and a score indicating the degree of the defect or non-defective product is obtained.

[0084]

[Embodiment 2] An optical member inspection apparatus according to a second embodiment of the present invention is different from the above-described first embodiment in the configuration in the host memory 62 and the image processing program storage area 62d of the host memory 62. Only the contents of the control processing executed by the CPU 60 in accordance with the image processing program stored in are different, and other configurations are common.

The second embodiment has been devised for the following reason. According to the first embodiment described above, only the area of each defective candidate element extracted by binarization is determined by the reference value Rk,
Normalization is performed based on Rd, and the quality of the inspection target optical member is determined based on the normalized area. Therefore, when the defect candidate elements having the same area are compared, it is determined that the degree of the defect is the same regardless of the magnitude of the luminance value. However, as shown in FIG. 14, when the defect candidate elements having the same area are compared with each other, the defect candidate element having a higher luminance value is more conspicuous than the defect candidate element having a lower luminance value. A high judgment can be made. Based on this kind of judgment, Fig. 1
As shown in FIG. 5, among the various defective candidate elements in which the luminance value and the area are distributed, those having a relatively high luminance value (FIG. 15)
The defective candidate elements within the range surrounded by the broken line above)
Even if the area is relatively small, it should have a large effect on the normalized value (point indicating the degree of influence on the performance of the optical member to be inspected) P, which is the basis for calculating the evaluation function F. Therefore, the control device 6 of the optical member inspection apparatus according to the second embodiment is characterized in that the normalized value P is calculated in consideration of not only the area of each defective candidate element but also the luminance. Of course, when calculating the normalized value P based on the luminance of each defective candidate element, the control device 6 uses a reference value appropriately set according to the position of the defective candidate element in the inspection target optical member. , The luminance of the defective candidate element is normalized.

Hereinafter, the configuration of the optical member inspection apparatus according to the second embodiment will be described. However, in the following, description of the configuration common to that of the above-described first embodiment will be omitted.

The optical member inspection apparatus according to the second embodiment having the configuration shown in FIG. 1 includes a control device 6 having a circuit configuration as shown in FIG. The circuit configuration shown in FIG. 16 has two working memory areas (a first working memory area 62e and a second working memory area 62f) in the host memory 62, as compared with the circuit configuration of the first embodiment shown in FIG. Only the points that are formed differ. The first working memory area 62e is an area where image data (polar coordinate data) in the image memory area 62a is copied and written in a state where coordinate conversion (polar coordinate-rectangular coordinate conversion) is performed. Further, the second working memory area 62f copies the image data (orthogonal coordinate data) in the first working memory area 62e and binarizes the copied image data in accordance with a predetermined threshold value, thereby indicating a "bad". This is an area where the position of a candidate element (an extracted element extracted as a group of pixels having a high luminance value) is specified.

The CPU 60 executes the image processing program stored in the image processing program storage area 62d of the host memory 62, and periodically stores the image data written in the frame memory 61 in the image memory area 62a of the host memory 62. Copy and copy the image memory area 62a
When the image data (polar coordinate data) corresponding to the entire inspection target optical member 14 is synthesized, the image data is subjected to a coordinate conversion process, and the image data (rectangular coordinate data) after the coordinate conversion is converted into the first data. Copy to the working memory area 62e. Next, the CPU 60 sets the first working memory area 62e
The defective candidate elements are extracted by copying the image data (rectangular coordinate data) in the second working memory area 62f and binarizing the image data (rectangular coordinate data) copied in the second working memory area 62e. I do. Next, the CPU 60
Is used to normalize the area of each extracted defective candidate element according to its position, and to set the pixel at the same position as the defective candidate element in the image data (orthogonal coordinate data) in the first working memory area 62e. Is normalized according to the position, and a predetermined operation is performed on the normalized area and brightness value to calculate a normalized number P of the defective candidate element. Next, the CPU 60 classifies the normalized number P calculated for each defective candidate element in accordance with various criteria described later, and
Aggregate in the classification table in c. Then, after the classification and totalization of the normalized numbers P for all the defective candidate elements are completed, an evaluation function F described later is calculated based on the totalization result, and the evaluation function F exceeds a predetermined determination reference value. Whether the inspection target optical member 14 is good or not is determined based on whether the optical member 14 is present. Note that the CPU 60 instructs the motor drive circuit 63 to supply a drive current to the drive motor 8 in synchronization with the capture of image data from the frame memory 61. <Principle of Quality Evaluation of Inspection Optical Member> A defect candidate element extracted in the binarized data in the second working memory area 62f is first determined at the position where it was formed, as in the first embodiment. Classified according to. As shown in FIG. 6, the inspection target optical member 14 is divided into four regions A to D concentrically around a position corresponding to the optical axis.

For each defective candidate element generated in each area, the CPU 60 normalizes the area by dividing the area by an area reference value R1 prepared for the area in accordance with the following equation (8). At the same time, in the first working memory 62e, the average value of the luminance values of all the pixels included in the range overlapping with the defective candidate element is normalized by dividing by the luminance reference value R2 prepared corresponding to the area, By calculating the square root of the sum of squares of the normalized area and the average luminance value, a normalized value (point indicating the degree of influence on the performance of the inspection target optical member) P is calculated. P = ((area / R1) ² + (average luminance / R2) ² ) ^0.5 (8) Specifically, the area reference value R1 to be applied to the region A
Is “S”, the CPU 60 applies “2S” as the area reference value R1 for the area B, applies “4S” as the area reference value R1 for the area C, and applies the area “8S” is applied as the reference value R1 for use. Similarly, when the average luminance reference value R2 to be applied to the area A is “L”, “2L” is applied as the average luminance reference value R2 to the area B, and the average luminance reference value R2 is applied to the area C. “4L” is applied as the value R2, and the area D
Is applied as “8L” as the reference value R2 for average luminance. That is, the position of the defect candidate element is determined by the optical member 1 to be inspected.
The two reference values R1 and R2 are applied so that the reference values R1 and R2 increase as the distance from the position corresponding to the optical axis No. 4 increases. Further, the square root of the sum of squares of the normalized area and the average luminance value is the normalized value P
Therefore, if the average luminance is very high even if the area is relatively small, the normalized value P is calculated to be large.
Conversely, if the area is very large even if the average luminance is relatively low, the normalized value P is calculated to be large.

When the defective candidate element is caused by a flaw, the reference values R1 and R2 are used as the reference values R1 and R2.
When k is applied and the defective candidate element is caused by dust, reference values R1d and R2d are applied. These reference values R1k and R2k, R1d and R2d are different values. That is, the reference value R1k applied to the area of the defective candidate element caused by the flaw formed in the region A is set to “Sk”.
If the reference value R1d applied to the area of the defective candidate element caused by dust formed in the region A is “Sd”, Sk and Sd are different from each other.
Similarly, the reference value R2k applied to the average luminance of the defective candidate element caused by the flaw formed in the area A is set to “Lk”, and the reference value R2k is applied to the average luminance of the defective candidate element caused by the dust formed in the area A. When the reference value R2d to be obtained is “Ld”, Lk and Ld are different from each other.
Therefore, prior to calculating the normalized value P, the CPU 60 determines whether the defective candidate element to be processed is due to a scratch or dust. This determination is made using the threshold function shown in FIG. Then, the CPU 60 determines the reference value Rk or Rd and Lk according to the type of the determined scratch or dust and the position of the area where the defective candidate element is formed.
Or Ld is determined, and the obtained reference values Rk or Rd and Lk are determined.
Alternatively, the above-described calculation of the normalized value P is performed based on Ld.

Next, if the normalized value P calculated as described above is 0.5 or more, the CPU 60 determines the determined dust in the classification table stored in the classification table storage area 62c. The normalized value P calculated for the defective candidate element is added to a column corresponding to the distinction between (ds) or flaw (k) and the area A to D where the defective candidate element has occurred.

On the other hand, if the calculated normalized value P is less than 0.5, another defective candidate element within a certain distance from the defective candidate element is searched for. When a defective candidate element is found in this way, another defective candidate element within a certain distance from the defective candidate element is further found. The search for such defective candidate elements is continuously performed until another defective candidate element cannot be found within a certain distance from the last found defective candidate element. If the number of defective candidate elements located within a certain distance from each other is three or more, the score obtained by multiplying the number of defective candidate elements by 0.25 is classified. In the table, the value is added to the column of concentration (m) corresponding to the areas A to D where the defective candidate element has occurred.

When the addition to the classification table as described above is performed for all the defective candidate elements, the CPU 60 calculates the above-mentioned (3) to (6) for each normalized value P written in each column of the classification table. Is performed, the evaluation function F is calculated, and the quality of the inspection target optical member 14 is determined according to the calculated evaluation function F, as in the case of the above-described first embodiment. <Control Processing> Next, the contents of the control processing executed by the control device 6 (CPU 60) according to the image processing program read from the image processing program storage area 62d in the second embodiment will be described. In the control processing of the second embodiment, S31 to S42 shown in FIG. 17 are executed instead of S09 to S20 (FIG. 11) in the control processing of the first embodiment shown in FIGS. That is, the CPU 60 performs the processing from S01 to S08 shown in FIG.
When the process of S08 is completed, the process proceeds to S31 of FIG. In this S31, the CPU 60
Determines whether the defective candidate element with the label n specified in S08 is caused by dust or a flaw as described above. When determining that the defective candidate element is caused by dust, the CPU 60 advances the processing from S32 to S33. On the other hand, when it is determined that the defect factor is caused by the scratch, the CPU 60
The process proceeds from S32 to S38.

In S33, the CPU 60 obtains the center of gravity of the defective candidate element with the label n specified in S08,
Among the respective areas in the binarized data, the area where the obtained center of gravity exists is determined (corresponding to a position measuring means).

In the next S34, the CPU 60 sets the area reference value R1d and the average luminance reference value R2d according to the region determined in S33. Specifically, when the determined area is the area A, the area reference value R1d is set to S
d, the average luminance reference value R2d is set to Ld, and if the determined region is the region B, the area reference value R1d
Is set to 2Sd, and the reference value for average luminance R2d is set to 2Ld. When the determined area is the area C, the reference value for area R1d is set to 4Sd, and the reference value for average luminance R2d is set to 4Ld. If the obtained region is the region D, the area reference value R1d is set to 8Sd and the average luminance reference value R2d is set to 8Ld.

In the next S35, the CPU 60 measures the area of the defective candidate element to which the label n specified in S08 is attached. That is, the total number of pixels constituting the defective candidate element is counted. Here, a graphic feature amount other than the area, for example, a maximum fillet diameter may be measured (corresponding to a graphic feature amount measuring unit). At the same time, the CPU 60 specifies the position of each pixel constituting the defective candidate element to which the label n specified in S08 is attached, and stores the image data (before binarization) stored in the first working memory area 62e. The luminance values of all the pixels existing at the corresponding positions in the rectangular coordinate data) are read, and the average value of the read luminance values is calculated.

In the next step S36, the CPU 60 divides the area of the defective candidate element measured in step S35 by the area reference value R1d set in step S34 and calculates the result in step S35 according to the following equation (8-1). The average brightness value obtained is S
Divide by the average luminance reference value R2d set at 34, and calculate the square root of the sum of squares of each division result as a normalization value P for this defective candidate element (corresponding to normalization means).

P = ((Area / R1d) ² + (Average luminance / R2d) ² ) ^0.5 (8-1) In the next S37, the CPU 60 sets the normalized value P calculated in S36 to 0. Check if it is 5 or more.
Then, if the normalized value P is 0.5 or more, the process proceeds to S21 in FIG. 12, and if the normalized value P is less than 0.5, the process proceeds to S23 in FIG.

On the other hand, in S38, the CPU 60 determines in S08
The center of gravity of the defective candidate element with the label n specified in is determined, and among the regions in the binarized data, the region where the determined center of gravity is located is determined (corresponding to a position measuring unit).

In the next S39, the CPU 60 sets an area reference value R1k and an average luminance reference value R2k corresponding to the region determined in S38. Specifically, when the determined area is the area A, the area reference value R1k is set to S
k and the average brightness reference value R2k are set as Lk, respectively. If the determined area is the area B, the area reference value R1k
Is set to 2Sk and the average luminance reference value R2k is set to 2Lk. If the determined area is the area C, the area reference value R1k is set to 4Sk and the average luminance reference value R2k is set to 4Lk. If the obtained region is the region D, the area reference value R1k is set to 8Sk and the average luminance reference value R2k is set to 8Lk.

In the next S40, the CPU 60 measures the area of the defective candidate element with the label n specified in S08. That is, the total number of pixels constituting the defective candidate element is counted. Here, a graphic feature amount other than the area, for example, a maximum fillet diameter may be measured (corresponding to a graphic feature amount measuring unit). At the same time, the CPU 60 specifies the position of each pixel constituting the defective candidate element to which the label n specified in S08 is attached, and stores the image data (before binarization) stored in the first working memory area 62e. The luminance values of all the pixels existing at the corresponding positions in the rectangular coordinate data) are read, and the average value of the read luminance values is calculated.

In the next step S41, the CPU 60 divides the area of the defective candidate element measured in step S40 by the area reference value R1k set in step S39 and calculates the result in step S40 according to the following equation (8-2). The average brightness value obtained is S
The result is divided by the average luminance reference value R2k set at 39, and the square root of the sum of squares of each division result is calculated as a normalized value P for this defective candidate element (corresponding to a normalizing means).

P = ((area / R1k) ² + (average luminance / R2k) ² ) ^0.5 (8-2) In the next S42, the CPU 60 sets the normalized value P calculated in S41 to 0.5. Check if this is the case. Then, the CPU 60 advances the process to S22 in FIG. 12 if the normalized value P is 0.5 or more, and advances the process to S23 in FIG. 12 if the normalized value P is less than 0.5. These S2
The processing after S1, S22, and S23 is exactly the same as that in the first embodiment described above, and thus the description thereof is omitted. <Operation of the Embodiment> FIG. 18 shows the area and average luminance (each of the reference values R1d, R2) of each defective candidate element caused by dust extracted from the area A when an inspection target optical member 14 is inspected.
9 is a graph showing the distribution of the area and the luminance value after division by d. Each of the arcs depicted in FIG. 18 indicates a range in which the normalized value P calculated for each defective candidate element is 1, 2.7, and 4, respectively. As is apparent from FIG. 18, even if the defective candidate elements have the same area,
The calculated normalized value P varies depending on the average luminance. Therefore, the optical member inspection apparatus according to the second embodiment can obtain the evaluation function F that further reflects the actual “conspicuousness” of the defective candidate element in the inspection target optical member 14. <Modification of Embodiment> According to the second embodiment, S3
In 5 and S40, the average value of the luminance values of the defective candidate elements is calculated. However, the maximum luminance value of the defective candidate elements may be obtained by the MAX calculation. However, in this case, S36 and S41
In calculating the normalized value P, the reference value for average luminance R2d,
A maximum luminance reference value is used instead of R2k. The maximum luminance reference value is set as a value larger than the average luminance reference values R2d and R2k in S34 or S39.

[0104]

Embodiment 3 An optical member inspection apparatus according to a third embodiment of the present invention is different from the optical member inspection apparatus according to the above-described second embodiment in that the CPU 60 according to the image processing program stored in the image processing program storage area 62d of the host memory 62. Are different only in the content of the control processing executed by the other, and have the other configurations in common. <Principle of Quality Evaluation of Inspection Optical Member> The CPU 60 controls each part of the optical member inspection apparatus in the same manner as in the above-described first embodiment, so that the inspection target optical member 14 is controlled.
Is converted from a polar coordinate system to a rectangular coordinate system, and the image data formed on the first working memory area 62e by this conversion is
Copy to the second working memory area 62f. And CPU
60 compares the luminance value of each pixel constituting the image data copied to the second working memory area 62f with a predetermined threshold value, and adds a new luminance value “255” to the pixel having a luminance value exceeding the threshold value. And a new luminance value “0” is given to a pixel having a luminance value equal to or less than the threshold value, thereby performing binarization and extracting a defective candidate element from the image data.

The defective candidate elements are classified according to the positions where they are formed, as in the first embodiment.
That is, as shown in FIG. 6, the inspection target optical member 14 is divided into four regions A to D in a concentric manner centered on a position corresponding to the optical axis.

The CPU 60 determines, for each defective candidate element generated in each area, the area of the defective candidate element and the total area included in the area overlapping with the defective candidate element in the first working memory 62e according to the following equation (9). The normalized value (point indicating the degree of influence on the performance of the inspection target optical member) P is calculated by integrating the average value of the luminance values of the pixels and dividing the integration result by the reference value R3. P = (area × average luminance) / R3 (9) Specifically, the reference value R3 to be applied to the region A is “S
In the case of “L”, the CPU 60 applies “2SL” as the reference value R3 for the area B, applies “4SL” as the reference value R3 for the area C, and “8SL” as the reference value R3 for the area D. That is, the reference value R3 is applied so that the position of the defective candidate element increases as the position moves away from the position corresponding to the optical axis of the inspection target optical member 14. The product of the area and the average luminance value is used as the reference. Value R3
, The normalized value P is calculated, so that even if the area is relatively small, if the average luminance is very high, the normalized value P is calculated to be large. Conversely, if the area is very large even if the average luminance is relatively low, the normalized value P is calculated to be large.

If the defective candidate element is caused by a flaw, the reference value R3k is applied as the reference value R3.
If the defective candidate element is caused by dust, the reference value R3d
Is applied. These reference values R3k and R3d are different from each other. That is, the reference value R3k applied to the area of the defective candidate element caused by the flaw formed in the area A is set to “SLk”, and the standard applied to the area of the defective candidate element caused by the dust formed in the area A is set to “SLk”. Change the value R3d to “SL
d ", SLk and SLd have different values from each other. Therefore, the CPU 60 sets the normalized value P
Prior to the calculation of, it is determined whether the defective candidate element to be processed is due to a scratch or dust. This determination is made using the threshold function shown in FIG. And C
The PU 60 sets a reference value R3 according to the determined type of the flaw or dust and the position of the area where the defective candidate element is formed.
k or R3d is obtained, and the above-described normalized value P is calculated based on the obtained reference value R3k or R3d.

Next, if the normalized value P calculated as described above is 0.5 or more, the CPU 60 determines that the determined dust in the classification table stored in the classification table storage area 62c. The normalized value P calculated for the defective candidate element is added to a column corresponding to the distinction between (ds) or flaw (k) and the area A to D where the defective candidate element has occurred.

On the other hand, if the calculated normalized value P is less than 0.5, another defective candidate element within a certain distance from the defective candidate element is searched for. When a defective candidate element is found in this way, another defective candidate element within a certain distance from the defective candidate element is further found. The search for such defective candidate elements is continuously performed until another defective candidate element cannot be found within a certain distance from the last found defective candidate element. If the number of defective candidate elements located within a certain distance from each other is three or more, the score obtained by multiplying the number of defective candidate elements by 0.25 is classified. In the table, the value is added to the column of concentration (m) corresponding to the areas A to D where the defective candidate element has occurred.

When the addition to the classification table as described above is performed for all the defective candidate elements, the CPU 60 applies the above-mentioned (3) to (6) to each normalized value P written in each column of the classification table. Is performed, the evaluation function F is calculated, and the quality of the inspection target optical member 14 is determined according to the calculated evaluation function F, as in the case of the above-described first embodiment. <Control Processing> Next, in the third embodiment, the contents of control processing executed by the control device 6 (CPU 60) according to the image processing program read from the image processing program storage area 62d will be described. In the control process of the third embodiment, S51 to S62 shown in FIG. 19 are executed instead of S09 to S20 (FIG. 11) in the control process of the first embodiment shown in FIGS. That is, the CPU 60 performs the processing from S01 to S08 shown in FIG.
When the process of S08 is completed, the process proceeds to S51 of FIG. In this S51, the CPU 60
Determines, as described above, whether the defective candidate element labeled with the label specified in S08 is caused by dust or a flaw. When determining that the defective candidate element is caused by dust, the CPU 60 advances the processing from S52 to S53. On the other hand, when it is determined that the defect factor is caused by the scratch, the CPU 60 advances the processing from S52 to S58.

In S53, the CPU 60 obtains the center of gravity of the defective candidate element with the label n specified in S08,
Among the respective areas in the binarized data, the area where the obtained center of gravity exists is determined (corresponding to a position measuring means).

In the next S54, the CPU 60 sets a reference value R3d according to the area determined in S53. Specifically, when the determined area is the area A, the reference value R3d is set to SLd, and when the determined area is the area B, the reference value R3d is set to 2SLd, and the determined area is set to SLd. Is the area C, the reference value R3d is set to 4SLd.
When the determined area is the area D, the reference value R3d is set to 8SLd.

In the next S55, the CPU 60 measures the area of the defective candidate element with the label n specified in S08. That is, the total number of pixels constituting the defective candidate element is counted. Here, a graphic feature amount other than the area, for example, a maximum fillet diameter may be measured (corresponding to a graphic feature amount measuring unit). At the same time, the CPU 60 specifies the position of each pixel constituting the defective candidate element to which the label n specified in S08 is attached, and stores the image data (before binarization) stored in the first working memory area 62e. The luminance values of all the pixels existing at the corresponding positions in the rectangular coordinate data) are read, and the average value of the read luminance values is calculated.

In the next step S56, the CPU 60 multiplies the area of the defective candidate element measured in step S55 by the average luminance value calculated in step S55 according to the following equation (9-1), and adds the multiplication result to step S54. By dividing by the reference value R3d set as above, a normalization value P for this defective candidate element is calculated (corresponding to normalization means).

P = (area × average luminance) / R3d (9-1) In the next S57, the CPU 60 checks whether or not the normalized value P calculated in S56 is 0.5 or more. I do.
Then, if the normalized value P is 0.5 or more, the process proceeds to S21 in FIG. 12, and if the normalized value P is less than 0.5, the process proceeds to S23 in FIG.

On the other hand, in S58, the CPU 60 determines in S08
The center of gravity of the defective candidate element with the label n specified in is determined, and among the regions in the binarized data, the region where the determined center of gravity is located is determined (corresponding to a position measuring unit).

In the next step S59, the CPU 60 sets a reference value R3k corresponding to the area determined in step S58. Specifically, when the determined area is the area A, the reference value R3k is set to SLk. When the determined area is the area B, the reference value R3k is set to 2SLk. Is the area C, the reference value R3k is set to 4SLk.
When the determined area is the area D, the reference value R3k is set to 8SLk.

In the next S60, the CPU 60 measures the area of the defective candidate element with the label n specified in S08. That is, the total number of pixels constituting the defective candidate element is counted. Here, a graphic feature amount other than the area, for example, a maximum fillet diameter may be measured (corresponding to a graphic feature amount measuring unit). At the same time, the CPU 60 specifies the position of each pixel constituting the defective candidate element to which the label n specified in S08 is attached, and stores the image data (before binarization) stored in the first working memory area 62e. The luminance values of all the pixels existing at the corresponding positions in the rectangular coordinate data) are read, and the average value of the read luminance values is calculated.

In the next step S61, the CPU 60 multiplies the area of the defective candidate element measured in step S60 by the average luminance value calculated in step S60 according to the following equation (9-2), and adds the result of multiplication to step S59. By dividing by the set reference value R3k, a normalized value P for this defective candidate element is calculated (corresponding to a normalizing means).

P = (area × average luminance) / R3k (9-2) In the next S62, the CPU 60 checks whether or not the normalized value P calculated in S61 is 0.5 or more. . Then, the CPU 60 advances the process to S22 in FIG. 12 if the normalized value P is 0.5 or more, and advances the process to S23 in FIG. 12 if the normalized value P is less than 0.5. These S2
The processing after S1, S22, and S23 is exactly the same as that in the first embodiment described above, and thus the description thereof is omitted. <Operation of the Embodiment> FIG. 20 shows an optical member 1 to be inspected.
4 is a graph showing the area and average luminance distribution of each defective candidate element caused by dust extracted from the area A when inspecting No. 4; In addition, each hyperbola drawn in FIG.
When Ld = 150, the range in which the normalized value P calculated for each defective candidate element is 1, 2.3, 3.6 is
Each is shown. As is apparent from FIG. 20, even if the defective candidate elements have the same area, the calculated normalization value P differs depending on the average luminance. Therefore, the optical member inspection apparatus according to the second embodiment can obtain the evaluation function F that further reflects the actual “conspicuousness” of the defective candidate element in the inspection target optical member 14. <Modification of Embodiment> According to the third embodiment, S5
In 5 and S60, the average value of the luminance values of the defective candidate elements is calculated. However, the maximum luminance value of the defective candidate elements may be obtained by the MAX calculation. However, in this case, S56 and S61
Values R3d, R3 used for calculating the normalized value P in
k is set as a larger value.

[0121]

According to the optical member inspection apparatus, the image processing apparatus, the image processing method, and the computer-readable medium of the present invention configured as described above, the failure factors are determined in consideration of the formation position of each failure factor. By evaluating the influence on the performance of the entire optical member to be inspected, the degree of overall failure of the optical member to be inspected can be quantified.

[0122]

[Brief description of the drawings]

FIG. 1 is a side sectional view showing a schematic configuration of an optical member inspection apparatus according to a first embodiment of the present invention.

FIG. 2 is a plan view of the inspection target optical member and the like in FIG. 1 viewed from a position of an imaging device.

FIG. 3 is a block diagram showing an internal circuit configuration of the control device of FIG. 1;

FIG. 4 is a diagram showing a light traveling state when an optical member to be inspected has no optical defect;

FIG. 5 is a diagram showing a traveling state of light when the inspection target optical member has an optical defect.

FIG. 6 is a diagram showing an area of image data.

FIG. 7 is a graph showing a relationship between each area in FIG. 6 and a normalized reference value.

FIG. 8 is a table showing a classification table.

FIG. 9 is a graph showing a threshold function used for shape determination;

FIG. 10 is a flowchart showing control processing executed by the CPU of FIG. 3;

FIG. 11 is a flowchart showing control processing executed by the CPU of FIG. 3;

FIG. 12 is a flowchart showing control processing executed by the CPU of FIG. 3;

FIG. 13 is an explanatory diagram of an evaluation function.

FIG. 14 is a graph showing a difference in luminance value between defective candidate elements having the same area.

FIG. 15 is a graph showing the distribution of the area and the luminance value of each defective candidate element.

FIG. 16 is a block diagram showing an internal circuit configuration of a control device of the optical member inspection device according to the second embodiment of the present invention.

FIG. 17 shows a second embodiment of the present invention;
Flowchart showing control processing executed by PU

FIG. 18 is a graph showing a distribution of normalized values.

FIG. 19 is a flowchart showing control processing executed in a third embodiment of the present invention.

FIG. 20 is a graph showing a distribution of normalized values.

FIG. 21 is a graph showing a comparative example for an evaluation function.

[Explanation of symbols]

 REFERENCE SIGNS LIST 3 imaging device 4 imaging lens 5 imaging device 6 control device 8 drive motor 14 inspection target optical member 15 holder 60 CPU 62 host memory 62a image memory area 62b classification table storage area

Claims (17)

[Claims] An image pickup apparatus for picking up an image of an optical member to be inspected and outputting image data including an image of the optical member to be inspected, and extracting a portion having a luminance different from the surroundings in the image data as an extraction element. Extracting means, a graphical feature quantity measuring means for measuring a graphical feature quantity of each of the extracted elements, a position measuring means for measuring a position in the image data of each of the extracted elements, for each of the extracted elements, Normalizing means for obtaining a normalized value by normalizing the graphic feature quantity measured by the graphic feature quantity measuring means according to the position measured by the position measuring means; and An optical member inspection apparatus comprising: an operation unit that executes a predetermined evaluation function based on all normalized values obtained by the normalization unit. 2. The imaging device according to claim 1, wherein the imaging device illuminates the inspection target optical member from outside the imaging optical axis to generate scattered light from an optical defect portion of the inspection target optical member. 2. The optical member inspection apparatus according to claim 1, wherein image data in which a portion is reflected as a bright portion is output. 3. The method according to claim 1, wherein said extracting means extracts said extracted element by binarizing a luminance value of each pixel constituting said image data by comparing it with a predetermined threshold value. Or the optical member inspection device according to 2. 4. An optical member inspection apparatus according to claim 1, wherein said graphic feature amount measuring means measures an area of each of said extracted elements. 5. The optical member inspection apparatus according to claim 1, wherein said position measuring means measures the position of each of said extracted elements with reference to a position corresponding to an optical axis of said optical member to be inspected. 6. The extraction unit according to claim 1, wherein the position of the extraction element measured by the position measurement unit increases with distance from a position corresponding to an optical axis of the inspection target optical member. 6. A standard value is obtained by dividing a graphical feature amount of the extracted element by the set standard value.
An optical member inspection device according to claim 1. 7. The position measuring means measures which of the plurality of concentric regions around the position corresponding to the optical axis of the inspection target optical member is located in each of the extraction elements, 7. The method according to claim 6, wherein the normalizing unit sets a reference value corresponding to each of the plurality of concentric regions.
An optical member inspection device according to claim 1. 8. The arithmetic unit adds a normalized value of each of the extracted elements obtained by the normalizing unit for each of the regions, and calculates a square root of a sum of squares of the addition result for each of the regions. The optical member inspection apparatus according to claim 7, wherein the inspection is performed. 9. The computing means classifies the extracted elements according to the type of optical defect that caused them, calculates the square root of the sum of squares for each type, and calculates the calculation result for each type. 9. The optical member inspection apparatus according to claim 8, further comprising calculating a square root of a sum of squares. 10. The arithmetic means, for the normalized value less than a predetermined value, as long as a predetermined number or more of extracted elements are not continuous at an interval within a predetermined distance from an extracted element corresponding to the normalized value. The optical member inspection apparatus according to claim 9, wherein the addition is not performed. 11. The method according to claim 1, wherein when a predetermined number or more of extracted elements are connected at intervals within a predetermined distance from the extracted elements corresponding to the normalized values less than the predetermined value, A value obtained by multiplying the number of extracted elements connected at intervals by a fixed value is set as a normal value for all of the extracted elements, and the normal value of such an entire extracted element is classified independently of other normal values. 2. The method according to claim 1, wherein
0 optical member inspection apparatus. 12. A pass / fail judgment means for judging that said optical member to be inspected is defective when a value obtained by executing said evaluation function by said arithmetic means exceeds a predetermined pass / fail judgment reference value. The optical member inspection apparatus according to claim 1, further comprising: 13. An optical member inspection apparatus according to claim 1, wherein said graphic feature amount measuring means measures an area and a brightness of each of said extracted elements. 14. The optical member inspection apparatus according to claim 1, wherein said position measuring means measures the position of the center of gravity of each of said extracted elements. 15. An extraction means for extracting, as an extraction element, a portion having a luminance different from the surroundings in image data including an image of the inspection target optical member obtained by imaging the inspection target optical member, A graphical feature quantity measuring means for measuring a graphical feature quantity of the extracted element; a position measuring means for measuring a position of each of the extracted elements in the image data; and a graphical feature quantity measurement for each of the extracted elements. Normalizing means for obtaining a normalized value by normalizing the graphical feature quantity measured by the means in accordance with the position measured by the position measuring means; and And a calculating means for executing a predetermined evaluation function based on all the normalized values. 16. A portion having a brightness different from that of surroundings in image data including an image of the inspection target optical member obtained by imaging the inspection target optical member is extracted as an extraction element. Measuring a graphical feature, measuring a position of each of the extracted elements in the image data, and normalizing a graphical feature of the extracted element for each of the extracted elements in accordance with the position in the image data Performing a predetermined evaluation function based on all the normalized values acquired for all the extracted elements. 17. A computer having a part having a luminance different from that of surroundings in image data including an image of the inspection target optical member obtained by imaging the inspection target optical member, as an extraction element, The graphical feature amount of each of the extracted elements is measured, and the position of each of the extracted elements in the image data is measured. For each of the extracted elements, the graphical feature amount is determined according to the position in the image data. A computer-readable medium storing a program for causing a normalized value to be obtained by normalization and executing a predetermined evaluation function based on all normalized values obtained for all extracted elements.