JPH08145907A - Inspection equipment of defect - Google Patents

Abstract

PURPOSE: To make it possible to detect various defects from the same image at a high speed by one inspecting equipment and to determine them synthetically by determining synthetically the defects detected by minute defect inspection, defect inspection by binarization of image data and nonuniformity inspection. CONSTITUTION: Gray level information on each pixel obtained by picking up 1 an image of a material to be inspected is taken in an image data input part 3 and subjected to A/D conversion. A defect (minute defect) inspecting part 5 determines each amount of variation of the gray level information between pixels in horizontal and vertical directions, adds up and averages the amount of variation for individual pixels and determines the amount as a defect when it is larger than a reference amount. In a binarizing inspection part 7, the defect is detected by binarizing image data by a prescribed threshold, while in a nonuniformity inspecting part 9, the gray level information on individual pixels are divided into lattices composed of matrixes being in prescribed numbers longitudinally and laterally, the gray level of each pixel in each lattice is added up, the amount of variation of an addition value between the lattices is determined in horizontal and vertical directions and it is determined as the defect when it is larger than a reference amount. In a part 13 for synthetic determination of the defect, the defect of the material to be inspected is determined synthetically on the basis of the results of these inspections.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a defect inspection apparatus for automatically inspecting the surface appearance of a film, a nonwoven fabric, a coil of steel, aluminum, copper or the like picked up by a line sensor camera.

[0002]

2. Description of the Related Art It is necessary to inspect whether or not these products have defects in a production line for sheet-like objects such as films, non-woven fabrics, steel, aluminum and copper coils. Conventionally, these inspections take an image of the inspection target with a camera, perform image processing on the image data,
A method of determining a defect based on the processing defect is adopted. In the image processing in this case, binarization processing is generally performed in which the image data (density data) from the camera is binarized with a predetermined threshold value to determine pass / fail.

[0003]

However, in the conventional defect inspection apparatus, since the inspection is generally executed by a single image processing represented by the binarization processing, various inspections occur. It was impossible to address all of the defects. For example, in a color filter for a liquid crystal display, an uneven defect such as a part of the color filter that becomes vaguely bright or dark due to uneven dyeing may occur, but this unevenness is a simple binarization. It cannot be detected by processing. Further, when processing image data with a lot of noise, accurate inspection cannot be performed by binarization. Further, it is impossible to perform a binarization process to detect a defect of one pixel or less such as a pinhole defect.

As described above, the conventional defect inspection apparatus mainly based on image processing cannot detect various kinds of defects, and it is necessary to use a dedicated image processing apparatus for each defect type to deal with each defect individually. There was no way.

The present invention has been made in view of the above circumstances, and an object thereof is to detect a plurality of types of different defects from the same image data at a high speed with a single inspection apparatus, and to detect detected defects. An object of the present invention is to provide a defect inspection apparatus capable of informing the type and comprehensively determining defects.

[0006]

In order to achieve the above-mentioned object, the invention described in claim 1 converts the density information for each pixel obtained by imaging the inspection object into digital image data. The image data input unit for fetching, the change amount calculating means for obtaining the change amount of the density information between each pixel in the input image data in the horizontal direction and the vertical direction, and the obtained change amounts in the horizontal direction and the vertical direction are described above. Change amount adding means for adding each pixel, averaging means for averaging the added change amount for each pixel, and detection for determining a fine defect when the averaged change amount is larger than a reference amount A micro-defect inspection unit having a means, a binarization inspection unit that binarizes the input image data with a predetermined threshold value to determine whether the image data is good or bad, and the input image data has predetermined vertical and horizontal directions. Number of pixel rows And a density adding means for obtaining density added value for each grid by adding density information of each pixel in each grid, and a change amount of the density added value between each grid in the horizontal and vertical directions. A variation amount calculating unit for each of the above, and a detection unit for determining unevenness when the obtained horizontal variation amount and vertical variation amount are larger than a predetermined amount, The present invention is characterized by comprising a defect comprehensive judgment unit for comprehensively judging defects of the inspection object respectively inspected by the defect inspection unit, the binarization inspection unit and the unevenness inspection unit.

According to a second aspect of the present invention, in the defect inspection apparatus according to the first aspect, the image data supplied from the image data input unit is emphasized by a vertical line along the conveyance direction of the inspection object. The enhancement means and the enhanced image data are added for each predetermined number of lines to obtain the density addition value,
Based on the density addition means that stores the current addition value, the difference calculation means that calculates the difference between the current addition value and the previous addition value, and the density addition data from the density addition means. Streak-like defect detection means for determining the presence or absence of streak-like defects in the conveyance direction of the inspection object,
And a defect determination unit, which is a fine defect inspection unit, a binarization inspection unit, an unevenness inspection unit, and a streak defect inspection unit. It is characterized by doing.

According to a third aspect of the present invention, in the defect inspection apparatus according to the first or second aspect, the image data input section converts the input analog image data into 10-bit digital image data A / D converter,
An A / D conversion circuit for converting the 10-bit converter output into 8-bit data is provided.

[0009]

According to the structure of the first aspect, the fine defect inspection section obtains the respective variation amounts between the density information for each of a plurality of pixels in the horizontal direction and the vertical direction, and determines the variation amounts in the horizontal direction and the vertical direction. Addition is performed for each pixel, the added change amount is averaged for each pixel, and when the averaged change amount is larger than a predetermined value, it is detected as a defect.

Further, the binarization inspection section binarizes the input image data by a predetermined threshold value and detects a defect by the binarized data.

Further, in the unevenness inspection section, the density information for each pixel is divided into grids each consisting of a predetermined number of vertical and horizontal pixel matrices, the density information of each pixel in each grid is added, and the density addition between each grid is added. The amount of change in the value is obtained for each of the horizontal direction and the vertical direction, and if the obtained amount of change in the horizontal direction and the amount of change in the vertical direction are larger than a predetermined amount, it is detected as unevenness.

The defect determination section is a fine defect inspection section,
The defects of the inspected object respectively inspected by the binarization inspection unit and the unevenness inspection unit are comprehensively determined.

According to the second aspect of the present invention, vertical lines are emphasized in the input image data along the conveyance direction of the object to be inspected, and the emphasized image data is added every predetermined number of lines to obtain the density addition value. Together with the current addition value, the difference between the present addition value and the previous addition value is calculated, and based on the calculated difference data or the density addition data from the density addition means Determine the presence or absence. Then, the defect determination section comprehensively determines the defects of the inspection object respectively inspected by the fine defect inspection section, the binarization inspection section, the unevenness inspection section, and the streak defect inspection section.

According to the third aspect of the invention, the analog image data is converted into 10-bit digital image data. Therefore, the image resolution is improved as compared with the case of 8-bit conversion. However, in order to image-process 10-bit data as it is, a device structure corresponding to 10-bit must be used. Therefore, this 10-bit data is converted into 8-bit data for image processing. This makes it possible to improve the image resolution without making the circuit configuration complicated and expensive as compared with the case where 8-bit data is handled from the beginning.

[0015]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the defect inspection apparatus according to the present invention will be systematically described below with reference to the drawings.

<Overall Configuration of Apparatus> FIG. 1 is a block diagram showing the overall configuration of an embodiment of the defect inspection apparatus according to the present invention.
FIG. 2 is a block diagram showing the configuration of the entire system of the defect inspection apparatus.

As shown in FIG. 1, the defect inspection apparatus of this embodiment has a CCD line sensor camera 1, an image data input section 3, a flaw (fine defect) inspection section 5, and a binarization inspection section 7.
And a nonuniformity inspection unit 9, an image display unit 11, and a comprehensive defect determination unit 13. 3 to 6 show detailed configurations of the respective component parts of the defect inspection apparatus shown in FIG.

As shown in FIG. 2, the system configuration includes a main CPU 15 and a plurality of local CPUs 17 centered on the main CPU 15. Each local CPU 17 is connected to a RO via a local bus 19.
The main CPU 15 is connected to the M21, the RAM 23, the image memory 25, and the VME interface 27.
7 is connected. 3 to 6,
The connection relationship between each component of the defect inspection apparatus and the local CPU 17 is omitted.

<Image Data Input Unit 3> The image data input unit 3 has a function of inputting image pickup data picked up by a CCD line sensor camera (hereinafter referred to as a line sensor) 1 and converting the image pickup data into digital image data. As shown in FIG.
An A / D conversion unit 31, a shading correction unit 33, a projection calculation unit 35, and a brightness correction unit 37 are provided.

<< A / D converter 31 >> In this embodiment,
The CCD line sensor 1 is used, the A / D conversion is set to 10 bits, and the digital signal processing unit is set to 8 bits.

Generally, digital data is handled in units of 8 bits, but in order to improve the accuracy of A / D conversion, for example, the accuracy of the A / D converter is changed from 8 bits to 10 bits.
In the case of using bits, it is necessary to change all digital circuits to those capable of 10-bit processing. However, it is not practical to change all the digital circuits to those of 10-bit processing because of high cost.

Therefore, in the A / D conversion section 31 of this embodiment, a bit conversion circuit 41 for converting 10-bit data into 8-bit data is provided at the next stage of the A / D converter 39. The bit conversion circuit 41 is specifically configured by an LUT (look-up table) as shown in FIG.

FIG. 7A shows an example of conversion at 4: 1 and FIG. 4B shows an example of conversion at 1: 1. (B)
When only a part of the input analog signal is used, as in the case of, by converting only that range,
It can be processed by an existing 8-bit digital processing circuit,
And the accuracy is also improved.

Therefore, even when the existing digital processing unit is used, by adding the A / D converter 39 and the bit conversion circuit 41, it is possible to perform more accurate data processing. Further, the bit conversion circuit 41 has an LU
By using the T method, the conversion accuracy can be freely changed at low cost. So, for example,
Processing such as inversion output and conversion of only some data is possible.

<Scratch (Fine Defect) Inspection Unit 5> The defect inspection unit 5 includes a change amount calculation unit 51 including a spatial filter (for example, a 3 × 3 Sobel filter) and a division table, and a change amount addition unit (pixels). Interval calculation unit) 53, average calculation unit (spatial filter and division table) 55, and sector addition unit (detection unit) 57. The sector addition unit 57 includes a density conversion circuit, a sector addition circuit, and an LUT. It consists of and.

<Binarization Inspection Unit 7> The binarization inspection unit 7 has a well-known structure, and is provided with an average value filter 71 for removing noise, a bright defect detection side, and a dark defect detection side. Each sector addition section 73 is provided, and each of the sector addition sections 73A and 73B includes a binarization circuit, a sector addition circuit, an L
It is composed of UT and.

<Mura Inspecting Section 9> The mura inspecting section 9 is composed of two-channel mura inspecting sections 9A and 9B. The unevenness detected by the unevenness inspection unit 9A of the first channel is the unevenness 1,
The unevenness detected by the unevenness inspection unit 9B of the second channel is referred to as unevenness 2. The size of the unevenness to be detected (uneven frequency) is different between the unevenness 1 and the unevenness 2. That is, the unevenness 1 refers to small-sized (high frequency) unevenness, and the unevenness 2 refers to large-sized (low frequency) unevenness. Therefore, the set mesh width is small in the first channel, and the set mesh width is large in the second channel. Each channel has the following configuration.

The uneven sector adding unit 91 is provided for the line sensor 1.
The density information for each pixel obtained by the above is divided into a grid consisting of a plurality of vertical and horizontal pixel matrices as shown in FIG. 8, and the density information of each pixel in each grid is added to obtain a density addition value for each grid.

The subtractor 93 executes a process of reducing the 20-bit density addition data obtained by the addition to 16-bit density addition data.

The change amount calculation unit (spatial filter A, spatial filter B and each absolute value circuit) 95 indicates the change amount in the horizontal direction and the change amount in the vertical direction of the density addition value between the respective grids in 3 rows and 3 columns. In the matrix of.

The determination circuits 97A and 97B detect unevenness of the object to be inspected based on the amount of change in the vertical direction and the amount of change in the horizontal direction obtained by the spatial change amount calculator 7.

<Image Display Unit 11> The image display unit 11 displays the output to the output shown in the figure (the output is a combined output of the light side binarization output and the dark side binarization output). Input through the display bus, select a desired image by the image display selection function 111, and select the VRAM circuit 113 and the DAC circuit 1.
It is supplied to a monitor device (not shown) as an NTSC signal via 15.

<Process in Scratch Inspecting Unit 5 (Micro Defect Detection Process)> First, the image data of the inspection object picked up by the line sensor 1 has density information for each pixel represented by 256 gradations of 0 to 255. ing.

A change amount calculator 5 for the density information of each pixel
1 obtains the amount of change H _ij in the horizontal direction and the amount of change V _ij in the vertical direction between density information in a matrix of 3 rows and 3 columns.

[0035] For example, in the horizontal direction of the variation H _ij and vertical arbitrary pixel L _ij variation V _ij is the three rows and three columns of nine lattice around the pixel L _ij as shown in FIG. 9 Based on the density information, the horizontal Sobel filter shown in FIG.
(2) and the vertical Sobel filter (second coefficient table) shown in FIG. 11 are used to obtain the values according to the following equations (1) and (2).

[0036]

H _ij = | -L _{i-1, j-1} -2L _{i, j-1} -L _{i + 1, j-1} + L _{i-1, j + 1} + 2L _{i, j + 1} + L _{i + 1, j + 1} | ... Expression (1) V _ij = | -L _{i-1, j-1} + L _{i + 1, j-1} -2L _{i-1, j} + 2L _{i + 1, j} -L _{i-1 , j + 1} + L _{i + 1, j + 1} | Equation (2) When the horizontal change amount H _ij and the vertical change amount V _ij are obtained, the change amount adding section 53 changes the horizontal change. The amount H _ij and the vertical direction change amount V _ij are added to obtain the change amount T _{ij of the} pixel.

When the amount of change for each pixel is obtained, the average calculator 55 uses the obtained amount of change as the average of the amounts of change in the central pixel with the average value of the nine amounts of change in 3 rows and 3 columns. 12
Averaging is performed using the filter shown in, and the average variation is obtained for each pixel.

For example, the change amount of the pixel L _{i-1, j-1} is T
_{i-1, j-1} , pixel L _{i, j-1} change amount is T _{i, j-1} , pixel L
_{i + 1, j-1} of the change amount T _{i + 1, j-1,} the pixels L _i-1, T the variation of _{j i-1, j,} a T _ij variation amount of the pixel L _ij, pixel L _{i + 1, j} of the change amount T _{i + 1, j,} pixel L _{i-1, j + 1} of the amount of change T
_{i-1, j + 1} , pixel L _{i, j + 1} change amount is T _{i, j + 1} , pixel L
_When the change amount of _{i + 1, j + 1} is T _{i + 1, j + 1} , the average change amount A _ij of the pixel L _ij is obtained by the following equation (3).

[0039]

## EQU00002 ## A _ij = (T _{i-1, j-1} + T _{i, j-1} + T _{i + 1, j-1} + T _{i-1, j} + T _ij + T _{i + 1, j} + T _{i-1, j +1} + T _{i, j + 1} + _{i + 1, j + 1} ) × 1/9 Equation (3) Then, in the sector addition unit 57, the calculated average change amounts and the preset defect determination thresholds are set. When the calculated average change amount is larger than a preset defect determination threshold value, it is determined to be a defect.

As described above, the horizontal variation amount and the vertical variation amount of the density information obtained by the line sensor 1 are emphasized in a plurality of vertical and horizontal matrices, for example, a horizontal Sobel filter and a vertical contour. Obtained by a vertical Sobel filter, the obtained amounts of change in the horizontal and vertical directions are added for each pixel, and the added amount of change is centered on, for example, the average value of nine changes in 3 rows and 3 columns. Since a defect is detected by the average amount of change using a filter that averages the change amount of pixels, even a defect that is minute (before and after the pixel area) with respect to the pixel of the line sensor 1 can be accurately detected. It becomes possible. Even when the density information obtained by imaging the inspection object includes a lot of shading and high frequency noise, the flaw inspection unit 5 of the present embodiment.
Is used, both the shading and the high frequency noise are reduced by obtaining the average change amount.

<Processing of Binarization Inspection Unit 7> In the binarization inspection unit 7, 0 to 25 supplied from the image data input unit 3 are supplied.
After removing the noise component from the average value filter 71 with respect to the density information for each pixel represented by 256 gradations of 5, a bright defect such as a hollow flaw and a dark defect such as a scratch are caused by predetermined threshold values. Detect each defect.

<Processing of unevenness detecting section 9> As described above,
The unevenness detecting unit 9 is set so that unevenness having a high unevenness frequency can be detected by the channel 1 and unevenness having a low unevenness frequency can be detected by the channel 2. The individual processing is exactly the same for both channels.

When the density information for each pixel is supplied from the image data input unit 3, the sector addition unit 91, which adds the density information, divides the density information into a grid consisting of a plurality of vertical and horizontal pixel matrices, as shown in FIG. The density information of each pixel in each grid is added to obtain a density added value for each grid. Note that in FIG. 8, the grid is divided into 10 rows and 10 columns by 10 pixels in each of the vertical and horizontal directions, but the number is not limited to this and this number can be arbitrarily changed.

The change amount calculation unit 95 obtains the change amount H _ij of the horizontal density addition value and the change amount V _ij of the vertical density addition value between the respective grids in a matrix of 3 rows and 3 columns.

For example, the amount of change H _{ij in} the horizontal direction and the amount of change V _{ij in the} vertical direction of the density addition value Σ _ij of an arbitrary grid are shown in FIG.
As shown in FIG. 11, the horizontal Sobel filter (spatial filter A) shown in FIG. 10 and the vertical direction shown in FIG. 11 are based on the density addition values of nine lattices of 3 rows and 3 columns centered on Σ _ij . Using the Sobel filter (spatial filter B) of Equation (4) and Equation (5) shown below, the values are obtained.

[0046]

H _ij = | −Σ _{i-1, j-1} −2Σ _{i, j-1} −Σ _{i + 1, j-1} + Σ _{i-1, j + 1} +2 Σ _{i, j + 1} + Σ _{i + 1, j + 1} | ... Expression (4) V _ij = | −Σ _{i-1, j-1} + Σ _{i + 1, j-1} −2Σ _{i-1, j} + 2Σ _{i + 1, j} −Σ _{i-1 , j + 1} + Σ _{i + 1, j + 1} | ... (5) Then, in the determination circuits 97A and 97B, the obtained horizontal variation amount H _ij and vertical variation amount V _ij are set in advance. When the variation amounts H _ij and V _ij thus obtained are compared with a non-uniformity determination threshold value and are larger than a preset non-uniformity determination threshold value, non-uniformity is determined.

As described above, the density information of the object to be inspected for each pixel obtained by the line sensor 1 is divided into a grid consisting of a plurality of vertical and horizontal pixel matrices, and the density information of each pixel in each grid is added to form a grid. The unevenness is emphasized by obtaining the density addition value for each grid, and the change amount H _ij of the density addition value in the horizontal direction between the lattices and the change amount V _ij of the density addition value in the vertical direction are _stored in a matrix of 3 rows and 3 columns. The unevenness is further emphasized by seeking at. Then, the obtained variation amounts H _ij and V _{ij of} each lattice are compared with a preset mura determination threshold value, and the obtained variation amounts H _ij and V _{ij of} each lattice are determined. If it is larger than the threshold value, it is determined as unevenness.

Therefore, since the variations H _ij and V _ij are obtained in both the horizontal and vertical directions to further emphasize the unevenness, the unevenness can be detected regardless of the detection direction of the inspection object, and the inspection object can be detected. Even if the surface of the object is finely divided into a grid like a color filter for LCD and there is a boundary line, the detection accuracy of unevenness does not depend on the inclination of the object to be inspected. further,
Even when the line sensor is used, it is possible to detect unevenness that occurs in a predetermined cycle that is difficult to detect regardless of the cycle.

It should be _noted that the sum of the obtained horizontal variation H _ij and vertical variation V _ij is used as the lattice variation C _ij , and this variation C _ij is determined for each lattice, and the decision circuit 97A is determined. , 97B compares the obtained variation amount C _ij with a preset unevenness determination threshold value, and if the obtained variation amount C _ij is larger than the preset unevenness determination threshold value, it determines that there is unevenness. It may be determined.

<Defect Comprehensive Judgment Processing> Each Inspection Unit 5, 7, 9
The inspection data from is supplied to the comprehensive defect determination unit 13. The defect comprehensive determination unit 13 has, for example, a defect type determination table 131 formed therein, and determines the defect type of the inspected object based on the inspection data supplied based on this table data. It should be noted that, as shown in FIG. 14, the outputs of the inspection units 5, 7, and 9 are classified into strength and weakness depending on the strength in the contrast direction, that is, whether or not the amount of change exceeds a threshold value. In the case of fine defects, there are two types of micro-large and micro-small, in the case of binarization, there are two types of binarization large and binarization small, and in the case of unevenness, uneven 1 strong, uneven 1 weak, uneven Two
A total of 8 types of inspections are possible, 4 types of strong and 2 less uneven. The threshold value at this time is preset in the defect comprehensive judgment unit 13. Therefore, at this time, the defect type determination table may be accessed using the 8-bit data as an index.

15 and 16 show examples of various defect names, their original images, density sections, judgment results, and processing screens. In this example, (1) very small hole (pinhole),
FIG. 4 is a schematic diagram showing the determination results by the apparatus of this embodiment for (2) holes, (3) large holes, (4) light dirt, (5) dirt, and (6) dark dirt.

(1) In case of extremely small hole (pinhole defect) As shown in FIG. 15 (1), extremely small hole (pinhole defect)
Is a minute defect, the defect is detected by the flaw inspection unit and is determined to be "micro small".

(2) In the case of hole As shown in FIG. 15 (2), the hole is judged to be "micro large" or "binarization small".

(3) Large hole As shown in FIG. 15 (2), a large hole has a defect detected in "large binarization" or "large binarization and all other judgments".

(4) Case of Light Stain As shown in FIG. 16A, "light stain" is judged to be "unevenness 1 weak".

(5) In the case of stains, as shown in FIG. 16 (2), the stains correspond to "strong unevenness 1" or "strong unevenness 1" and "micro large".

(6) In the case of dark stains As shown in FIG. 16 (2), dark stains are "strong unevenness 1 strong".
And "Mura 2".

As described above, according to the apparatus of this embodiment, a plurality of defects can be comprehensively determined from one image data, and by setting the defect information in the defect type determination table in advance, each defect type can be determined. Can be determined.

<Streak Defect Detection Process> FIG. 17 shows an example of the structure of a streak defect inspection section. This streak defect inspection section 151 is provided with a flaw inspection section after the image data input section 3. 5 and the like are provided in parallel.

First, the principle will be described. Steel plate, strip steel,
A continuous sheet product such as a resin sheet or paper or a sheet product cut out one by one is shown in FIG. 18 (1) along the conveyance direction (flow direction) of the product in which defect detection is difficult. A low-contrast streak-like defect or a fine-width streak as shown in FIG. 18B may occur.

A density cross section of one line of a low-contrast streak defect as shown in FIG. 18A is shown in FIG.
As shown in (3), the width of the streak portion is wide and the noise is relatively small. On the other hand, in the concentration cross section of the fine scratch having a fine width as shown in FIG.
As shown in (4), the streak width is narrow and a large amount of noise occurs randomly. When these respective image data are integrated in the flow direction (conveyance direction), in the low-contrast streak in FIG. 18 (1), the density added value becomes clear in the streak part as shown in FIG. 19 (1). It becomes possible to detect the streak portion by cutting at a certain threshold value. However, as shown in FIG. 19B, since the density added value of the image data with fine-width streak in FIG. 18B has a lot of noise, the streak portion and the noise portion cannot be distinguished from each other. Judgment becomes difficult. Therefore, a streak of a fine width cannot be detected by the method of simply integrating the image data (density value).

Therefore, as shown in FIG. 17, as a pre-process for the image data output from the image data input unit 3, a spatial filter unit 153 is provided and the filter output is added by the integration processing unit 155. Also,
As a post-process of the output of the integration processing unit 155, a difference processing unit 57 that takes a difference for each block is provided. Further, as the determination unit, a first determination unit 159 that determines based on the output data of the integration processing unit 155 and a second determination unit 161 that determines based on the output data of the difference processing unit 157 are provided.

The spatial filter section 153 can employ, for example, the above-mentioned vertical Sobel filter, and the streak portion in the image data is emphasized by this Sobel filter. This concentration cross section is shown in FIG.

When integration processing is performed on the data in which the streak portion is emphasized by the spatial filter section 153, the difference between the noise portion and the streak portion becomes clearer as shown in FIG. When the noise level is low, the first determining unit 159 determines the streak defect by thresholding the integrated data.

On the other hand, when the difference between the noise and the streak portion is not clear only from the integrated data, the difference processing unit 157 calculates the difference from the previous integrated value as shown in FIG. The second determination unit 161 makes a determination based on the difference data.

These judgment results are supplied to the defect comprehensive judgment unit 13, and the defect inspection of the object to be inspected is comprehensively executed based on the judgment results. If there is a defect, what kind of defect is displayed. Of.

As described above, according to the streak defect inspection section 151, it is possible to stably detect a streak defect having a low contrast or a fine width, and the detection accuracy is improved. It is also possible to detect spot-like low-contrast defects.

Further, since the difference between the previous integrated value and the present integrated value is obtained and the streak defect is inspected based on this difference value, the starting point of the streak scratch and the end point of the streak scratch can be detected. This makes it possible to detect defects more clearly.

[0069]

As described above, according to the invention described in claim 1, a plurality of different kinds of defects can be detected at high speed from the same image data by one inspection apparatus, and the kind of the detected defects can be detected. It becomes possible to comprehensively judge the defect by informing the user.

According to the second aspect of the present invention, it becomes possible to stably detect a low-contrast or fine-width streak-like defect.

According to the third aspect of the present invention, even when the existing digital processing unit is used, by providing the A / D converter and the conversion circuit, it is possible to perform more accurate data processing.

[Brief description of drawings]

FIG. 1 is a block diagram showing the configuration of an embodiment of a defect inspection apparatus according to the present invention.

FIG. 2 is a block diagram showing a system configuration of a defect inspection apparatus according to the present invention.

FIG. 3 is a block diagram showing a configuration of an image data input unit in the defect inspection apparatus according to the present invention.

FIG. 4 is a block diagram showing a configuration of each inspection unit in the defect inspection apparatus according to the present invention.

FIG. 5 is a block diagram showing a configuration of a nonuniformity inspection unit in the defect inspection apparatus according to the present invention.

FIG. 6 is a block diagram showing a configuration of an image display unit in the defect inspection apparatus according to the present invention.

FIG. 7 is an explanatory diagram showing a table configuration in an A / D conversion circuit.

FIG. 8 is an explanatory diagram showing an example in which the density information for each pixel is divided into a grid composed of a plurality of vertical and horizontal pixel matrices.

FIG. 9 is an explanatory diagram for obtaining a horizontal variation amount and a vertical variation amount of luminance information of an arbitrary pixel.

FIG. 10 is an explanatory diagram showing a horizontal Sobel filter.

FIG. 11 is an explanatory diagram showing a vertical Sobel filter.

FIG. 12 is an explanatory diagram showing a filter of 3 rows and 3 columns for obtaining an average change amount.

FIG. 13 is an explanatory diagram for obtaining a horizontal variation amount and a vertical variation amount of a luminance added value of an arbitrary grid.

FIG. 14 is an explanatory diagram showing a comprehensive defect determination process.

FIG. 15 is an explanatory diagram showing a result of defect comprehensive determination processing.

FIG. 16 is an explanatory diagram showing a result of defect comprehensive determination processing.

FIG. 17 is a block diagram showing a configuration example of a stripe defect inspection unit.

FIG. 18 is an explanatory diagram of a stripe defect.

FIG. 19 is an explanatory diagram showing a density addition value of a stripe defect.

FIG. 20 is an explanatory diagram showing a density cross section after the streak-like defect is filtered.

FIG. 21 is an explanatory diagram showing a density cross-section after the streak-like defect filter processing + integration processing.

FIG. 22 is an explanatory diagram showing a density cross section after difference processing of streak defects.

[Explanation of symbols]

 1 CCD line sensor camera 3 image data input unit 5 flaw (fine defect) inspection unit 51 change amount calculation unit 53 change amount addition unit 55 average calculation unit 57 sector addition unit 7 binarization inspection unit 9 unevenness inspection unit 91 uneven sector addition unit 95 change amount calculation unit 11 image display unit 13 defect comprehensive determination unit 39 A / D converter 41 bit conversion circuit 151 streak defect inspection unit

 ─────────────────────────────────────────────────── ─── Continuation of front page (72) Inventor Akihiro Tsubakihara 66-2 Horikawa-cho, Saiwai-ku, Kawasaki-shi, Kanagawa Toshiba Engineering Co., Ltd. (72) Noriyuki Arai 66-2 Horikawa-cho, Kawasaki-shi, Kanagawa Toshiba Engineering Co., Ltd. (72) Inventor Masaharu Ishida 66-2 Horikawa-cho, Sachi-ku, Kawasaki-shi, Kanagawa Prefecture Toshiba Engineering Co., Ltd. (72) Masayuki Takahashi 66-2 Horikawa-cho, Kawasaki-shi, Kanagawa Prefecture In Toshiba Engineering Co., Ltd. (72) Inventor Hiroshi Iwasaki 66-2 Horikawa-cho, Sachi-ku, Kawasaki-shi, Kanagawa Inside Toshiba Engineering Co., Ltd.

Claims (3)

[Claims] 1. An image data input section for converting density information of each pixel obtained by picking up an image of an inspection object into digital image data and loading the image data, and an amount of change in density information between pixels in the input image data. Is calculated for the horizontal and vertical directions, a change amount adding means for adding the calculated respective change amounts in the horizontal direction and the vertical direction to each pixel, and the added change amount for each pixel is averaged. A fine defect inspection section having an averaging means for converting the input image data into a predetermined threshold value, and a detecting means for determining that there is a fine defect when the averaged change amount is larger than a reference amount. Binarization inspection section that determines the quality of image data by binarizing with, and divides the input image data into a grid consisting of a predetermined number of pixel matrix in each of the vertical and horizontal directions, and adds the density information of each pixel in each grid. Each grid A density addition means for obtaining the density addition value for each, a change amount calculation means for obtaining the change amount of the density addition value between the respective grids in the horizontal direction and the vertical direction, and the obtained horizontal change amount and the vertical direction. When the variation amount of is larger than a predetermined amount, unevenness detection means for determining that there is unevenness,
And a comprehensive defect determination unit for comprehensively determining defects of the inspected object respectively inspected by the fine defect inspection unit, the binarization inspection unit, and the irregularity inspection unit. Defect inspection equipment. 2. The defect inspection apparatus according to claim 1, further comprising: an emphasis unit for emphasizing vertical lines in the image data supplied from the image data input unit along a conveyance direction of the inspection object, and the emphasized image data. The density addition value is calculated by adding every predetermined number of lines, the density addition means is stored as the current addition value, the difference calculation means for calculating the difference between the current addition value and the previous addition value, and the calculated difference data. Based on, or based on the density addition data from the density addition means, streak defect detection means for determining the presence or absence of streak defects in the conveyance direction of the object to be inspected, and a streak defect inspection unit having, The defect determination unit is a defect inspection apparatus characterized by comprehensively determining defects of the inspected object respectively inspected by the fine defect inspection unit, the binarization inspection unit, the unevenness inspection unit, and the streak defect inspection unit. 3. The defect inspection apparatus according to claim 1 or 2, wherein the image data input unit converts the input analog image data into 10-bit digital image data.
A defect inspection apparatus comprising a converter and an A / D conversion circuit for converting the 10-bit converter output into 8-bit data. A defect inspection apparatus comprising: