JP2003242511A - Similar periodic pattern inspection device and inspection method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To dispense with the preparation of a reference pattern or a judgment threshold pattern. <P>SOLUTION: This device has an imaging apparatus 2; an image processor 3 for inputting the data of the imaging apparatus 2; and a position detector 4 for outputting a carrying position detection signal PG. The image processor 3 comprises an image memory 31; an image memory write control circuit 32 for storing the data in the image memory 31; a small block memory 34 for dividing the data in the image memory 31 into small blocks followed by storing; a middle block memory 35 for storing middle blocks; a middle block N-value memory 36A for performing N-value processing of the middle block memory 35 with an N-value processing threshold determined by dividing the difference between standard deviation maximum value and minimum value of the middle blocks by N; a middle block carrying directional projection memory 36B for storing the total value of middle blocks of the same position on the scanning direction of N-value processing; and a scanning direction projection memory 37 for dividing the center part of the middle blocks of the same position into three parts in the scanning direction and storing the scanning directional total value. <P>COPYRIGHT: (C)2003,JPO

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an inspected object having a similar periodic pattern, for example, a semiconductor wafer, a liquid crystal display panel, a similar periodic pattern inspection apparatus and an inspection method for inspecting a print pattern.

[0002]

2. Description of the Related Art As a conventional device of this kind, Japanese Patent Laid-Open No. 05-
There is a print pattern inspection device shown in 172542.
However, the first problem in this apparatus is that it is necessary to capture a grayscale image of a perfect product in advance and create a differential processed image and a binarized processed image of the perfect product, and inspect various print patterns. In this case, there was a problem that pretreatment was required. Further, as a second problem, there is a problem that an erroneous determination occurs due to a change in the conveyance speed of the pattern to be inspected or a change in the image size caused by the image capture.

As a method that does not use a complete product pattern in advance, which has occurred in the above-mentioned apparatus, Japanese Patent Laid-Open No. 2001-209798.
There is a visual inspection device shown in. This device sets the region of interest in the repeated pattern part, creates a reference image that has been image-processed (smoothed) from the captured image, compares the image with the next captured image, and binarizes the sharp gradation change part. The presence or absence of a defective portion is determined, and the standard deviation value of the repeated portion is obtained to obtain a binarization threshold value. However, even in this apparatus, the region of interest in the repetitive pattern portion needs to be preset, and an erroneous determination occurs due to a change in the transfer speed of the pattern to be inspected or a change in the image size caused by image capture, and the standard deviation of the repetitive portion In the method of obtaining the value, the binarization threshold value may not be properly obtained in the image part including the defect, and the standard deviation value becomes large in the case of a complicated image. In this case also, the binarization threshold value is appropriate. False determination may occur due to the fact that it is not obtained.

As a device capable of inspecting a pattern defect even if various patterns exist, there is a pattern defect inspection device disclosed in Japanese Patent Laid-Open No. 04-332809. In this device, the same position pattern between chips is scanned by the same electron beam, the obtained image signal is measured and stored by the binarization circuit unit and the edge number counting unit, and then the same region of another chip is scanned. When different parts are detected, by performing image processing such as pattern position, pattern length, rotation, etc. of one chip, pseudo defects due to drawing error are removed and highly accurate alignment is performed in the entire area of the chip. . In this apparatus, the processing time delay due to image processing such as pattern position, pattern length, and rotation becomes a problem, and in order to scan the same position pattern between chips, the transport speed and the scanning speed must be kept the same. The problem arises that it must be done.

As a method for detecting and determining a pattern, there is a chip size and chip pitch detecting method disclosed in Japanese Patent Laid-Open No. 03-074855. In this method, image data of a certain area is sampled, a histogram of gray values of the area is created, and then the density of the sampling area is determined from the characteristics of the histogram. This process is repeated for adjacent areas in the secondary direction, the chip size and chip pitch are determined from the density of all areas, and the obtained chip size and chip pitch are obtained from the same location in the adjacent chip areas. This is a method of comparing and determining image data and reference data. This method has a problem that the chip size cannot be determined unless the pattern shape is a quadrangle such as a semiconductor chip and there is a plain area outside the pattern area. In addition, there is a problem in that reference data for making a determination must be created.

As an apparatus in which the arrangement of patterns varies and it does not take time and effort for each product type, Japanese Patent Laid-Open No. 2001-0596.
There is a defect inspection method for a pseudo periodic pattern shown in FIG.
This method determines the array element pixel area as the pixel area occupied by each array element corresponding to the predetermined array pitch and array direction of the image data to be inspected, and adds the value of each pixel corresponding to the array element pixel area. , Image processing such as integration processing for obtaining an integrated value for each array element and differential processing for emphasizing the difference in the integrated value for each array element pixel area from the integration processing result, and the result obtained from the image processing result On the other hand, it is a device for extracting a defective portion from the extraction result obtained by repeating the process of extracting a defect candidate by comparing it with a predetermined slice level a plurality of times as necessary. The problem with this apparatus is that the defect candidates are extracted, and therefore, it may not be possible to extract the defects with small grayscale differences by performing the differential processing. In addition, there is a problem that it is necessary to set the condition setting process and the pseudo periodic pattern area from the defect sample in advance in order to determine the array element pixel area.

[0007]

DISCLOSURE OF THE INVENTION Problems to be Solved by the Invention As described above,
The first problem in the existing pattern inspection apparatus is that it is necessary to prepare a reference pattern or a determination threshold pattern for determining the quality of an inspection target, and it is necessary to perform a particularly detailed inspection. Indeed, it takes time to prepare.

A second problem is that it is necessary to preset the area to be inspected or to extract features by setting a certain limit on the object to be inspected. This has been a further problem because the conveyance speed of the inspection object to be inspected is varied or the inspection object is slightly deformed.

It is an object of the present invention to provide a similar periodic pattern inspection apparatus and inspection method which are capable of automatically inspecting various similar periodic patterns with only a very simple setting.

[0010]

In order to solve the above-mentioned problems, the invention of a similar periodic pattern inspection apparatus according to the first aspect of the present invention is to inspect an object to be inspected having a similar pattern conveyed in the conveying direction in the scanning direction. The data from the image pickup device for imaging is divided into small blocks, and a small block memory that stores the average value and standard deviation value of the brightness values of the small blocks, and the standard deviation value of the brightness values of the middle blocks configured by the small blocks. And a threshold value of N-value obtained by dividing the difference between the standard deviation maximum value and the standard deviation minimum value of the middle block by N (N is a positive integer),
A medium block N-valued memory for N-valued storing the contents of the medium block memory, and a medium-block carrying direction projection memory for storing a total value of the middle blocks at the same position in the N-valued scanning direction, A central portion of the middle block at the same position in the scanning direction is divided into three, and a scanning direction projection memory that stores a total value in each scanning direction is provided. The start and end of the pattern are detected, the size of the pattern is normalized, the difference between the patterns is calculated, the threshold value for determination is calculated from the middle block, and the pattern is determined using the threshold value for determination. It is characterized in that the mutual difference is binarized, the continuity of defect candidate locations is judged, and the defects are output.

The invention of a similar periodic pattern inspection apparatus according to claim 2 is an image pickup apparatus for picking up an image of an inspection object having a similar pattern in the scanning direction, and an image processing apparatus to which data from the image pickup apparatus is input. And a position detection device that outputs a transfer position detection signal of a transfer device that transfers the inspection object in the transfer direction, and the image processing device has an image memory that stores data from the imaging device, An image memory writing control circuit that stores data from the image pickup device in the image memory based on the transport position detection signal, and the data stored in the image memory is divided into small blocks, and the brightness value of the small blocks is divided. A small block memory that stores the average value and standard deviation value of the, and a medium block composed of the small blocks. The medium block memory, and the medium block memory according to an N-valued threshold value obtained by dividing the difference between the maximum standard deviation and the minimum standard deviation of the medium blocks by N (N is a positive integer). A N-valued medium block storing N-valued contents, a medium block carrying direction projection memory storing a total value of the middle blocks at the same position in the N-valued scanning direction, and the scanning direction In the image processing apparatus, the central portion of the middle block at the same position is divided into three and a scanning direction projection memory for storing the total value in each scanning direction and an initial setting value of the pattern are input. A CPU for performing the processing of 1., a threshold value memory for storing a judgment threshold value calculated based on the contents of the medium block N-valued memory, and a difference memory for storing the difference between the patterns. Characterized in that it obtain.

Further, in the invention of a similar periodic pattern inspection apparatus according to a third aspect, the small block according to the first or second aspect has one pixel in the carrying direction and the scanning direction of the inspection object. It is characterized in that it is constructed so as to overlap.

Further, in the invention of the similar periodic pattern inspection apparatus according to claim 4, the small block memory according to claim 1 or 2 stores the average value of the brightness values of the small blocks. An average value memory, a small block deviation value memory that stores the standard deviation value of the brightness values in each small block, and a small block standard deviation that stores the maximum standard deviation value of the brightness values of the small blocks. A maximum value memory and a small block standard deviation maximum value memory that stores a minimum value of standard deviation values of the lightness values of the small blocks;
The medium block memory stores a medium block standard deviation value memory that stores a standard deviation value of the lightness values in the medium block for each medium block, and a maximum standard deviation value of the lightness values of the medium block. A medium block standard deviation maximum value memory, and a medium block standard deviation maximum value memory storing a minimum value of the standard deviation value of the lightness value of the medium block,
The scanning direction projection memory divides the central portion of the middle block at the same position in the scanning direction into three, and stores a total value in the scanning direction of each of the scanning direction start end projection memories and scanning direction central projection. A memory and a projection memory in the scanning direction are provided.

Further, in the invention of the similar periodic pattern inspection apparatus according to the fifth aspect, the number of pulses of the transport position detection signal according to the second aspect is measured, and the inspection target is inspected only by the resolution in the transport direction. The next data is written in the image memory only when the target surface moves.

Further, in the invention of a similar periodic pattern inspecting apparatus according to a sixth aspect, in the case where the conveying apparatus according to the second aspect conveys the inspection object surface of the inspection object at a constant speed, The image memory writing control circuit does not control writing to the image memory, and the image memory stores all data from the imaging device.

Further, in the invention of a similar periodic pattern inspection apparatus according to claim 7, if it is necessary for the transfer apparatus according to claim 2 to inspect only when the speed exceeds a certain speed, The image memory writing control circuit is configured so that the image speed is stored in the image memory only when the transfer speed of the transfer device is calculated from the pulse cycle of the transfer position detection signal and becomes higher than a specified transfer speed. Is configured.

Further, in the invention of a similar periodic pattern inspection device according to an eighth aspect, the transport device according to the second aspect outputs an inspection start signal for starting the inspection of the pattern, and outputs the inspection start signal as the inspection start signal. The image memory writing control circuit is operated in synchronization with each other.

Further, in the invention of the similar periodic pattern inspection method according to the ninth aspect, an image memory storing step of storing data of an image of an inspection object having a similar pattern in an image memory, and the data of the image memory are small. A small block dividing step of dividing the block into blocks and calculating an average value and a standard deviation value of the brightness values; and a medium block dividing step of dividing into small blocks composed of the small blocks and calculating a standard deviation value of the brightness values , An N-value conversion step of obtaining a threshold value from the standard deviation value of the medium block and converting the data of the medium block into N-value (N is a positive integer), and projecting and detecting the N-valued data A pattern detection step of detecting the start and end of the pattern in the small block in the middle block, normalizing the size of the pattern, and calculating the difference between the patterns. A difference memory storing step of storing the difference memory in a difference memory, a judgment threshold calculating step of calculating a judgment threshold from the medium block, and a binary data of the difference memory using the judgment threshold. And a defect output step of determining the continuity of defect candidate locations and outputting the defect,
It is characterized by including.

[0019]

DESCRIPTION OF THE PREFERRED EMBODIMENTS << Summary of the Invention >> The present invention relates to a similar cycle in which the quality of a product is judged by comparing similar patterns of an inspection object having a plurality of similar patterns in the conveyance direction within the same inspection plane. The pattern inspection apparatus is characterized in that the inspection can be automatically performed without setting the reference determination level and the inspection area in advance and even when the transport speed changes.

FIG. 1 is a block diagram of the present invention. Figure 1
At, the digital image data signal VD obtained from the solid-state imaging device 2 that scans the inspection target surface 1 is input to the image processing device 3. On the other hand, the inspection target surface 1 is transported by a transport device not shown in FIG. 1, but the transport position detection signal PG is obtained by the position detection device 4 and input to the image processing device 3.

FIG. 2 is a block diagram of the image processing apparatus 3. The solid-state imaging device 2 images the inspection target surface 1 along the scanning direction B. The image memory writing control circuit 32 uses the digital image data signal V obtained by the solid-state imaging device 2.
D is stored in the image memory 31 based on the transport position detection signal PG obtained by the position detection device 4. (Image memory storage step) The data stored in the image memory 31 is divided into small block areas obtained from the relationship between the defect size to be inspected and the image resolution, and the average value of the lightness values is calculated for each small block. The calculated value is stored in the small block average value memory 34A of the small block memory 34, the standard deviation value of the lightness value in the small block is calculated for each small block, and the small block standard deviation value memory 34B of the small block memory 34 is calculated. To store. (Small Block Dividing Step) It should be noted that each small block area is configured to overlap by one pixel or more in the carrying direction and the scanning direction. The small blocks stored in the small block average value memory 34A are divided into the medium block scanning direction small block number MBX and the medium block carrying direction small block number MBY in the medium block area, and the standard deviation value of the lightness value for each medium block is calculated. To do. The obtained standard deviation value is stored in the middle block standard deviation value memory 35A of the middle block memory 35, and the maximum and minimum standard deviation values of the middle block are respectively stored in the middle block standard deviation maximum value memory 35. 35B and medium block standard deviation minimum value memory 3
Store in 5C. (Medium Block Dividing Step) When the stored data amount is equal to or larger than a certain value or when an inspection start is instructed by an external signal, the medium block standard deviation maximum value memory 35B and the medium block standard deviation minimum value memory 35
The data in the medium block standard deviation value memory 35A is converted into an N-value based on the threshold value obtained by C (N is a positive integer). (N-value conversion step) The obtained N-valued data is used to transfer the transfer-direction pattern size PYS and the transfer-direction pattern number PY.
N, the pattern size PXS in the scanning direction and the number of patterns in the scanning direction PXN are divided and projected, and the middle block of the change point from the plain area to the picture area is detected, and the maximum deviation in the middle block is detected. The beginning and end of each pattern is detected by extracting the pattern change portion in the small block having a value. (Pattern detection process) After the pattern size in the transport direction and the scanning direction is determined,
Next, after normalizing the pattern size, the difference between the patterns is calculated and stored in the difference memory 39. (Differential Memory Storing Step) A judgment threshold value S for obtaining a defect candidate pixel from the data stored in the differential memory 39 is required. The determination threshold value S is calculated based on the contents of the medium block N-valued memory 36A that includes the pixel to be compared. (Determination Threshold Calculation Step) Medium block N-valued memory 36 including two pixels to be compared
Depending on the combination of the maximum value and the minimum value of the contents of A, the fixed threshold coefficient K8 or the adjacency of the pixel to be compared 5 ×
Either 5 pixels + fixed threshold value coefficient K8 or the maximum difference of 3 × 3 pixels + fixed threshold value coefficient K8 is selected. Here, the selected one is the judgment threshold value S.
Becomes When the binarization of the image data in the difference memory 39 is completed using the judgment threshold value S, the continuity of the obtained defect candidate portions is judged, and when the continuity in the horizontal / vertical direction is equal to or more than a certain value, the defect signal is given. Output E. (Defect Output Process) In this way, according to the present invention, the pattern area is automatically determined and the determination level is automatically created even if the transport speed is changed or a similar pattern, which is extremely simple. By default, pattern-independent inspection can be performed automatically.

<< Structure of First Embodiment >> The structure of the first embodiment will be described with reference to FIGS.

In FIG. 1, a digital image data signal V obtained from a solid-state image pickup device 2 for scanning the surface 1 to be inspected.
Input D to the image processing apparatus 3. On the other hand, the inspection target surface 1 is transported by a transport device not shown in the figure, but the transport position detection signal PG is obtained by the position detection device 4 and input to the image processing device 3. The surface 1 to be inspected is assumed to be transported in the transport direction A in the figure,
The solid-state imaging device 2 performs scanning in the scanning direction B direction in the figure.
Here, when the conveyance in the conveyance direction A is stopped, the conveyance direction A and the scanning direction B are in a relationship orthogonal to each other.

In FIG. 2, the image processing apparatus 3 includes an image memory 31 for storing digital image data, an image memory writing control circuit 32 for storing the image data in the image memory 31 based on the carrying position, and various processes. And the CPU 33 for inputting the initial setting value, the data stored in the image memory 31 is divided into small block areas, and the small block memory 34 for storing the data of this small block (the small block memory for storing the average value of the lightness values of the small blocks). Block average value memory 3
4A, a small block standard deviation value memory 34B that stores the standard deviation value of the small block brightness values, a small block standard deviation maximum value memory 34C that stores the maximum standard deviation value of the small block brightness values, and a small block brightness Small block standard deviation minimum value memory 34D storing the minimum value of standard deviation value
A medium block memory 35 (medium block standard deviation value memory 35A, medium block standard deviation maximum value memory 35B, medium block standard) which stores data of a medium block composed of a plurality of small blocks. The deviation minimum memory 35C) is connected.

Further, the medium block N-valued memory 36A, the medium block carrying direction projection memory 36B, the scanning direction projection memory 37 (the scanning direction start end projection memory 37A, the scanning direction center projection memory 37B, the scanning for performing the pattern matching process) Direction end projection memory 37C), determination threshold value S calculated based on the contents of the medium block N-valued memory 36A
And a difference memory 39 for storing the difference between patterns to be compared. It should be noted that there is no problem if all of these configurations are realized by the CPU 33 unless there is a limitation on the processing time.

The configuration of the embodiment has been described in detail above.
Since the solid-state image pickup device 2 is well known to those skilled in the art and is not directly related to the present invention, its detailed configuration will be omitted.

Although the solid-state image pickup device 2 is composed of a one-dimensional camera in the present application, it is natural that a two-dimensional camera may be used as long as there is no problem in driving frequency and field of view.

Similarly, the position detecting device 4 of FIG. 1 may use a rotary encoder (which outputs a pulse) or the like, and since it is not directly related to the present invention, its detailed configuration will be omitted.

Further, in the above embodiment, although the illumination section is not shown, the reflected light may be used when inspecting the picture or the like of the printed matter, and the transmitted light may be used when inspecting the perforation or the like. As a matter of course, both are appropriately selected and used in combination as necessary.

Further, how to sort the inspection objects according to the conveying device and the inspection result is not directly related to the present application, and therefore the detailed configuration will be omitted. << Operation of the First Embodiment >> Next, the present invention will be described. The operation will be described in detail with reference to the drawings. As items to be set in advance before the inspection, it is assumed that the conveyance direction pattern number PYN, the conveyance direction pattern size PYS, the scanning direction pattern number PXN, and the scanning direction pattern size PXS are input to the CPU 33, respectively. In addition, when there are a plurality of inspection targets, a plurality of conveyance direction pattern sizes PYS may be set or a range value may be set. Further, when the number of patterns in the scanning direction PXN is 1, the position SSP in the scanning direction at the starting end side and the position SE in the scanning direction at the end side SE are used to determine the inspection range.
It is necessary to set either P or the scanning direction start plain area SSW and the scanning direction end plain area SEW.

Before describing the inspection operation, an example of specific numerical values of each device will be described below. When the pattern size PXS in the scanning direction of the inspection surface 1 is 80 mm, the field of view WS of the solid-state imaging device 2 needs to be WS> K1.PXS. This coefficient K1 may be set appropriately in consideration of the positional deviation of the pattern to be inspected in the scanning direction, the lens system shading of the solid-state imaging device 2, etc. However, if K1 = 1.2 here, WS> 1.2PXS WS is 96mm
As described above, the field of view WS is set to 100 here for simplicity.
mm.

Assuming that the defect detection size ES required next is 1 mm, the resolution WB of the solid-state image pickup device 2 in the scanning direction is WB.
Needs to be K2 · WB = ES, and the scanning direction decomposition coefficient K2 is set for more stable inspection. The scanning direction resolution coefficient K2 may be set to 1 or more and may be increased so that the inspection can be performed with higher accuracy. However, if it is too large, it is disadvantageous in terms of inspection processing time. . Here, the scanning direction decomposition coefficient K2 is set to 10
, The scanning direction resolution WB = ES / K2 =
It becomes 0.1 mm. As a result, the number of elements LB of the solid-state imaging device 2 becomes 1,000 from LB = WS / WB, but in consideration of the number of commercially available elements, here, the number of elements LB of the solid-state imaging device 2 is set to 1,024. To do.

On the other hand, assuming that the maximum transport speed SMAX at which the surface 1 to be inspected is transported is 1,000 mm / S and the resolution HB in the transport direction of the solid-state imaging device 2 is the same as the resolution WB in the scanning direction, HB = WB = 0. It will be 1 mm. In order to eliminate inspection omissions in the transport direction A, the next scan must be started at least while the inspection target surface 1 moves in the transport direction A by the transport direction resolution HB.
Drive frequency F is F = (LB + BB) / HB × SMAX
Ask more. In this case, the number of blank bits in the scanning direction B
B is a value determined by the specifications of the solid-state imaging device 2 and is assumed to be 100 here. From the above, the driving frequency F is at least 11.2 MHz.

The position detecting device 4 of FIG. 1 is provided with a device for outputting a position synchronizing pulse, such as a rotary encoder, at a portion where a position error such as a slip does not occur in a drive portion of the conveying device. The position resolution PW of the position detecting device 4 is determined by the defect size ES to be detected, the carrying direction resolution HB, and the carrying direction decomposition coefficient K3. Similarly to the scanning direction decomposition coefficient K2, the carrying direction decomposition coefficient K3 is preferably about 2 to 20, but if K3 = 10 is set here, PW = HB / K2.
= 0.01 mm.

The operation of each part will be described in detail with reference to the specific numerical values described above. Naturally, the above-mentioned numerical values are shown in order to more specifically describe the operation, and do not limit the scope of the present application.

In FIG. 1, a digital image data signal VD obtained when the solid-state image pickup device 2 scans the surface 1 to be inspected is sent to the image processing device 3. Further, the inspection object 1 is transported by a transport device (not shown in FIG. 1), and the position detection device 4 is attached to the transport device drive unit or the like.
The transport position detection signal PG of is also sent to the image processing apparatus 3.

FIG. 2 is a block diagram of the image processing apparatus 3. The operation of each part of the image processing apparatus 3 will be described below. Digital image data signal VD input to the image processing device 3
Are sequentially stored (stored) in the image memory 31. Here, while the solid-state imaging device 2 always operates at a constant drive frequency, the carrier device is driven / stopped and the speed fluctuates. Therefore, it is necessary to simply store all the obtained digital image data signals VD. Since it becomes inefficient in terms of storage capacity and processing time, the image memory writing control circuit 32 controls the image memory writing. This control method measures the number of pulses of the carrying position detection signal PG obtained from the position detecting device 4, and writes the digital image data signal VD of the next scanning line only when the inspection target surface 1 moves by the carrying direction resolution HB. This can be executed by (storing) or by overwriting the same storage portion until the inspection target surface 1 moves for a certain distance or more. Actually, since a pulse interval error of the position detecting device 4 occurs, it is necessary to make it smaller than the carrying direction resolution HB.

The image data stored in the image memory 31 is divided into small block areas. This division into small blocks is used to prevent the influence of minute defects due to the smoothing processing of image data, and to extract the change points of the pattern and to calculate the threshold value S for judgment. It should be noted that these small blocks are configured to overlap at least one pixel in the scanning direction B and the transport direction A. The small block size may be set in consideration of the minute defect removal size and the processing time. However, since the small block scanning direction pixel number SBX is preferably 2 or more and the scanning direction decomposition coefficient K2 or less, the small block scanning direction pixel is used here. The number SBX = 5. Similarly, the small block carrying direction pixel number SBY is preferably 2 or more and the carrying direction decomposition coefficient K3 or less, so that the small block carrying direction pixel number SBY is set to 5 here. The overlap amount is one pixel in both the scanning direction and the transport direction. The area configuration of these small blocks is shown in FIG. FIG. 3 shows the relationship between pixels, small blocks, and patterns.

The average value of the brightness values of the image data in the small blocks is calculated for each small block and stored in the small block average value memory 34A, and the brightness value of the image data in the small blocks is calculated for each small block. Calculate the standard deviation value. The obtained standard deviation value is stored in the small block standard deviation value memory 34B, and the maximum and minimum standard deviation values of the small blocks are respectively stored in the small block standard deviation maximum value memory 34.
C, small block standard deviation minimum value memory 34D.

The small blocks stored in the small block average value memory 34A are set to the number MB of small blocks in the medium block scanning direction.
X, the number of small blocks in the medium block carrying direction MBY is divided into medium block areas. The size of the medium block may be determined in consideration of the complexity of the image data and the processing time, but here, the number of small blocks in the medium block scanning direction MBX = 5 and the number of small blocks in the medium block carrying direction MBY = 5. Therefore, the number of small blocks forming one medium block is 25. The standard deviation value of the lightness value of the image data for each medium block is calculated, the obtained standard deviation value is stored in the medium block standard deviation value memory 35A, and the maximum and minimum standard deviation values of the medium block are calculated. The values are stored in the medium block standard deviation maximum value memory 35B and the medium block standard deviation minimum value memory 35C, respectively.

The above processing may be carried out every time the number of pieces of data forming the small block and the medium block is stored in the image memory 31, or if an inspection signal is obtained from the outside, it is based on the signal. You can go. When the number of medium blocks becomes equal to or larger than the pattern size PYS in the transport direction or based on an external signal, the pattern matching preprocessing is started.

A preset transport direction pattern size P
When YS is 30 mm, the conveyance direction resolution HB is 0.1.
Since it is mm, the number of pixels GXN in the transport direction is GXN =
It becomes 300 from PYS / HB. Therefore, the number of small blocks in the transport direction CYS is CYS = 1 + (GXN-SBY) / (S
From BY-1) to 74. Number of blocks in the transport direction CY
M becomes 14 from CYM = CYS / MBY. The pattern matching preprocessing is sequentially started when the number of middle blocks CYM in the transport direction becomes equal to or larger than the calculated value K4 (K4 is a coefficient). This coefficient K4 may be set within the range of 0.5 to 1 in consideration of the speed fluctuation of the transfer device, but here K4 =
Set to 0.7. Therefore, the pattern matching preprocessing is performed when the number of blocks CYM in the transport direction reaches 10 or more.

The number of elements (the number of pixels in the scanning direction) LB of the solid-state imaging device 2 is 1,024, the number of pixels in the small block scanning direction SBX = 5, and the number of small blocks in the medium block scanning direction MBX = 5. To the number of small blocks in the scanning direction CX
S is 204, and the number of blocks in the scanning direction CYM is 40. Although a supplemental number of pixels these residual may be closer to either one for may be equally divided into the starting end or the terminating end, or convenient.

In the pattern matching preprocessing, first, the data in the medium block standard deviation value memory 35A is converted into N values. The threshold for this N-value conversion is the medium block standard deviation maximum value memory 35.
B, the difference in the medium block standard deviation minimum value memory 35C is divided by N to obtain the difference. For example, when the value of the medium block standard deviation maximum value memory 35B is 105 and the value of the medium block standard deviation minimum value memory 35C is 0, the difference is 105 and N
When 3 is 3, the threshold values are 35 and 70. Based on the obtained threshold value, the medium block standard deviation value memory 35A
N-valued and stored in the medium block N-valued memory 36A and stored in the medium block conveyance direction projection memory 36B.
Stores the total value of middle blocks at the same position in the scanning direction for binarization.

Similarly, the central portion of the middle block at the same position is divided into three parts in the scanning direction starting projection memory 37A, scanning direction central projection memory 37B, and scanning direction ending projection memory 37C in the scanning direction. Stores the total value above. The reason why the projection in the scanning direction is divided into three is to prevent the defect of the conveyance direction pattern detection described later due to the defect data. Of course, if all the total values from the start end to the end in the scanning direction are divided into three, the visual field area other than the inspection target is included, so the portion excluding the specific areas on the start and end sides is divided into three. . In this embodiment, the scanning direction pattern size PXS is 80.
Since the field of view WS is 100 mm and the difference is 2
Whether 0 mm is on the start side or the end side in the scanning direction,
The portion of 20 to 80 mm where the pattern definitely exists is divided into three. That is, the scanning direction 20 to 40 mm is set to the scanning direction start projection memory 37A, the scanning direction 40 to 60 mm is set to the scanning direction central projection memory 37B, and the scanning direction is 60 to 80 mm.
Are stored in the scanning direction end projection memory 37C.

The pattern detection processing in the carrying direction is performed by the scanning direction start end projection memory 37A and the scanning direction center projection memory 37.
It starts when the total value of any one of B and the scanning direction end projection memory 37C becomes from 0 to a coefficient K5 or more. The coefficient K5 may be increased as the change in shade between the pattern and the plain portion is larger and the number of patterns PXN in the scanning direction is larger. As a matter of course, before the actual inspection target surface 1 reaches the visual field of the solid-state imaging device 2, random image data and patterns of the transport device are stored as image data and stored in the scanning direction start end projection memory 37A. K
It is natural that the pattern cannot be specified only by 5. This coefficient K5 is set as a trigger for the conveyance direction processing. Since the data before the coefficient K5 is not used in principle, there is an effect that unnecessary storage capacity is not used. However, in actuality, there is a possibility that the plain area size in the transport direction A may be used, so it is necessary to temporarily store a large amount of image data.

Scanning direction start end projection memory 37A, scanning direction center projection memory 37B, scanning direction end projection memory 37C.
The image data point where the total value of any one of the above becomes 0 or more and the coefficient K5 or more is set to the temporary transfer direction start point YDS of the first pattern.
Let P1. 2nd pattern conveyance direction start point YDSP2
Is defined as YDSP1 + PYS-K6 (PYS: carrying direction pattern size), and carrying direction pattern comparison is performed. Coefficient K
The value of 6 may be increased as the variation of the transport speed is increased. Although the number of pixels in the transport direction GXN is used as a reference here, the position resolution PW of the position detection device 4 is 0.01 m.
Since it is m, the number of pulses of the position detection device 4 may be measured starting from the YDSP 1 and the process may be performed when the conveyance direction pattern size PYS-K6 is reached.

In the carrying direction pattern comparison, first, the carrying direction starting point YDSP2 of the second pattern is extracted, and therefore, a portion where the total value of the scanning direction starting end projection memory 37A is from 0 to a coefficient K5 or more is extracted. If it can be extracted, each point is set as the starting point, and the transfer direction pattern size PYS
-K7 (coefficient K7 is set to a range of 0 to 0.5 PYS, 0 is set here) is set as a collation range, and the distribution of the N-valued data of the middle block included in the scanning direction start end projection memory 37A is set to the first value. The 1st pattern and the 2nd pattern are compared. This comparison is not performed for all N-valued patterns, only for the maximum value and the minimum value of N-valued, and the rest are not to be compared. If the distribution of the N-valued data does not match, the same process is performed for the scanning direction central projection memory 37B and the scanning direction end projection memory 37C, and if the distributions of the first pattern and the second pattern do not match, The temporary transport direction start point YDSP1 is discarded, the next start point is searched, and processing is performed.

When the first pattern and the second pattern match, the start and end points of the pattern in the carrying direction are calculated. First of all
The small block with the largest deviation value among the middle blocks on the same scanning line at the temporary start point of the pattern is extracted, and when the small block can be specified, projection processing is performed in the small block in the scanning direction, The portion changed from the pattern area portion to the pattern area portion is defined as the starting position of the first pattern in the transport direction. Next, from the temporary start point of the second pattern, the scanning direction start end projection memory 37
The total value of any one of A, the scanning direction central projection memory 37B, and the scanning direction end projection memory 37C is 0 to a coefficient K5.
The above image data points are used as the temporary transfer direction start point YDSP2 of the second pattern, and the small block having the largest deviation value among the middle blocks on the same scanning line of the temporary start point of the second pattern is extracted, When the block can be specified, projection processing is performed in the scanning direction B within the small block, and the portion changed from the plain area portion to the pattern area portion is set as the transporting direction start position of the second pattern.

When the starting point of the second pattern can be specified, the starting point from the front middle block toward the origin side in the carrying direction,
Image data points where the total value of the center and the end is from 0 to a coefficient K5 or more are extracted, and when a small block can be specified, projection processing is performed in the scanning direction B within the small block, and the solid area portion to the pattern area portion The portion changed to is the end position of the first pattern in the transport direction. Below, the number of patterns in the transport direction PYN
The above processing is repeatedly performed up to -1.

For the final pattern, the total value of the start end, the center and the end from the front middle block toward the origin side in the transfer direction from the transfer end point position + PYS + K7 (PYS: transfer direction pattern size) point of the previous pattern. Is 0
From the image data points having a coefficient of K5 or more, the small block can be identified.
Then, the projection process is performed on, and the portion changed from the plain area portion to the pattern area portion is set as the end position in the transport direction of the final pattern.

When the number of patterns PXN in the scanning direction is 1,
The scanning direction start edge detection processing is performed by fixed cutting based on the scanning direction start edge side position SSP and the scanning direction end edge position SEP which are input in advance, or when the scanning direction start edge side uncoated area SSW and the scanning direction end side uncoated area SEW are input. If it is present, change point detection processing is performed from the scanning direction start end side and the scanning direction end side, and the start end and the end end are cut out. For example, when both the scanning direction start end side uncoated region SSW and the scanning direction end side uncoated region SEW are designated as 5 mm, the scanning direction resolution WB is 0.1 mm and the small block scanning direction pixel number SBX = 5,
Since the number of small blocks in the medium block scanning direction MBX = 5, there should be at least a region of one or more medium blocks that is a solid region. Therefore, the medium block carrying direction projection memory 36B is scanned from the scanning direction start end side, A portion where the projection value increases from 0 is set as a block boundary portion at the start end. Similarly, the terminal side also calculates the block boundary portion during termination. Here, both the scanning direction start end side uncoated area SSW and the scanning direction end side uncoated area SEW are both set to 5 mm, but this value is 2 mm.
In the following cases, there is a possibility that the boundary portion cannot be detected in the medium block processing, so the processing is performed in small blocks, or the number of pixels in the small block scanning direction SBX = 5 and the number of small blocks in the medium block scanning direction MBX = 5 are reduced. Then, it is necessary to set such that at least a plain area having at least one medium block is obtained.

When the boundary part of the block in the starting end is found, the starting position of the scanning direction is detected. In the scanning direction start end position detection, the middle block having the largest deviation value is selected from among the middle blocks next to the start middle block boundary portion (portions in which the value of the middle block carrying direction projection memory 36B is 1 or more). If they have the same value, the one closer to the center is selected. When the medium block having the largest deviation value can be selected, the small block having the largest deviation value among all the small blocks included in the medium block is selected. When a small block can be specified, projection processing is performed in the carrying direction A within the small block, and the portion changed from the plain area portion to the pattern area portion is set as the scanning direction start position. Similarly, the processing is also performed on the end side to determine the end position in the scanning direction.

When the number of patterns PXN in the scanning direction is 2 or more, the scanning direction similar pattern detection processing is performed, and then the scanning direction start and end detection is performed for each pattern. When the number of scanning direction patterns PXN is 2, the medium block carrying direction projection memory 36B is divided by a value obtained by adding 1 to the number of scanning direction patterns PXN, and the last middle block carrying direction projection memory 36B.
The similar pattern is detected based on the pattern. In this embodiment, the number of pixels in the scanning direction (the same as the number of elements LB of the solid-state imaging device 2) is 1,024, and the number of pixels in the small block scanning direction SBX =
5, the number of small blocks in the medium block scanning direction MBX = 5 to the number of medium blocks in the scanning direction B CXM is CXM = ((LB-S
BX) / (SBX-1) +1) / MBX becomes 51. Here, since the number of patterns PXN in the scanning direction is 2, the end of the second pattern after the 34th position in the middle block is
Since the end of the first pattern should be included before the 16th, the end of the second pattern is detected by the following. The last side of the medium block transport direction projection memory 36B (5
From the 1st) to the 34th, the portion having the smallest value and the value of the medium block carrying direction projection memory 36B that does not increase thereafter is extracted, and the portion having the largest projection memory value is extracted. A medium block having a large deviation value is selected, and a small block having the maximum deviation value among all the small blocks included in the medium block is selected. When the small block can be identified, projection processing is performed in the carrying direction A within the small block, and the portion changed from the plain area portion to the pattern area portion is set as the scanning direction end position of the second pattern.

Similarly, since the start end of the first pattern is included in the 0th to 16th blocks of the medium block carrying direction projection memory 36B, the 16th from the first side (0th) of the medium block carrying direction projection memory 36B. The portion having the smallest value among the second and the portion in which the value of the medium block carrying direction projection memory 36B does not increase thereafter is extracted, and the largest deviation value among the portions having the larger projection memory value is extracted. The medium block is selected, and the small block having the maximum deviation value among all the small blocks included in the medium block is selected. When the small block can be specified, projection processing is performed in the small block in the carrying direction, and the portion changed from the plain area portion to the pattern area portion is set as the scanning direction start position of the first pattern.

Next, the boundary portion between the first pattern and the second pattern is obtained from the start position of the first pattern and the end position of the second pattern. Since there should be a boundary in the vicinity of the half position of the sum of the start end position of the first pattern and the end of the second pattern, a half position of the sum of the start end position of the first pattern and the end of the second pattern. The start end of the second pattern is detected on the end side in the scanning direction starting from the projection memory of the block of the middle block-1 including the. This start point detection also extracts a portion whose projection memory value is smaller than the projection memory value of the starting point block and whose projection memory value does not increase thereafter, and which has the largest deviation value among the portions whose projection memory value has increased. Is selected, and the small block having the maximum deviation value among all the small blocks included in the medium block is selected. When the small block can be identified, projection processing is performed in the small block in the carrying direction, and the portion changed from the plain area portion to the pattern area portion is set as the scanning direction start position of the second pattern.

Next, the end position of the first pattern in the scanning direction is detected. Since there should be a boundary near the half position of the total of the start position of the first pattern and the end position of the second pattern as in the start of the second pattern, the start position of the first pattern and the end of the second pattern The end of the first pattern is detected at the start end side in the scanning direction, starting from the projection memory of the block of the middle block +1 which includes the position of 1/2 of the total. This end detection also extracts a portion having a smaller value than the projection memory value of the block in the starting point and the projection memory value does not increase thereafter, and the middle block having the largest deviation value in the portion having the larger projection memory value. Is selected, and the small block having the maximum deviation value among all the small blocks included in the medium block is selected. When the small block can be specified, projection processing is performed in the carrying direction A within the small block, and the portion changed from the plain area portion to the pattern area portion is set as the scanning direction end position of the first pattern.

As described above, the start and end positions of the two patterns in the scanning direction are obtained. When the widths of the two obtained patterns are significantly different, the pattern detection size is abnormal and the process ends. In addition, the obtained start and end positions are, to be exact, the start and end of the pattern, and the area of the two patterns including the plain portion is the start interval between the two patterns or the end interval between the two patterns. Therefore, the first pattern (pattern) end position and the second pattern
2 with the middle point of the pattern (pattern) start position as the boundary point
Finalize the pattern position and size in one scan direction.

After the pattern sizes in the carrying direction A and the scanning direction B are determined, the pattern sizes are then normalized.
The normalization unifies the pattern size to be compared to a large size. There are various methods for unifying, such as simple insertion and average value insertion, and any method may be used. When the pattern sizes match, the difference between the patterns (difference between the image data) is calculated and stored in the difference memory 39.

Next, the judgment threshold value S is calculated. The determination threshold value S is calculated based on the contents of the medium block N-valued memory 36A that includes the pixel to be compared. If the two pixels to be compared are M and N, and the maximum value of the contents of the medium block N-valued memory 36A in which each pixel is included is the maximum value (= 2) in either case, the Peripheral pixel 5 adjacent to the target pixel
The maximum value of × 5 pixels + fixed threshold value coefficient K8 ”is used as the determination threshold value S. When both the maximum values of the contents of the medium block N-valued memory 36A including two pixels to be compared are the minimum values (= 0), the “fixed threshold coefficient K8” is set as the determination threshold S. If none of the above applies, "the maximum value of 3 × 3 neighboring pixels centering on the comparison target pixel + fixed threshold coefficient K8" is used as the determination threshold S.

The contents of the difference memory 39 are binarized by the judgment threshold value S obtained by the above processing. As the fixed threshold value coefficient K8, the maximum value of the standard deviation values of the blocks having the minimum value (= 0) in the medium block N-valued memory 36A may be applied, or simply the maximum value of the pixel level may be applied. 1/1
You may apply about 0.

After the binarization of the image data (that is, the contents of the difference memory 39) is completed, the obtained defect candidate portion (the difference stored in the difference memory 39 is the judgment threshold value S).
In a larger portion, the continuity of (obtained when binarizing the contents of the difference memory 39) is determined, and if at least K9 pixels or more are continuous in the transport direction A or the scanning direction B, it is determined and output as a defect. This judgment output is the defect signal E
May be output. The determination coefficient K9 is determined by the defect size ES to be detected, the carrying direction resolution HB, and the carrying direction decomposition coefficient K3. That is, in this embodiment, both the transport direction resolution coefficient K3 and the scanning direction resolution coefficient K2 are 10
Therefore, the value may be simply set to 10, and the defect candidate part of the maximum peripheral 7 × 7 pixels may have been deleted when the threshold value is determined. 5
The part where the × 5 process is performed may be changed to 5 and the remaining part may be changed to 10 such that the threshold process is different.

<< Second Embodiment >> The second embodiment of the present invention will be described below.
2 shows an embodiment of the present invention. When the transporting device always transports the surface 1 to be inspected at a constant speed, it is not necessary to perform the image memory writing control by the image memory writing control circuit 32 in FIG. All the digital image data signals VD from are stored (stored).

<< Third Embodiment >> The third embodiment of the present invention will be described below.
2 shows an embodiment of the present invention. If it is necessary to perform the inspection only when the transport device has a speed higher than a certain speed, the transport speed of the transport device is calculated from the cycle of the transport position detection signal PG, and only when the speed is higher than the designated transport speed. The image memory writing control circuit 32 may be configured to store (store) the digital image data signal VD.

In the third embodiment, the image memory writing control circuit 32 calculates the transport speed of the transport device from the cycle of the transport position detection signal PG, the transport speed computing unit, and the transport speed from this transport speed computing unit. It is configured to include a carrying speed comparison unit that compares the designated carrying speed and outputs a write control signal when the designated carrying speed is high.

<< Fourth Embodiment >> The fourth embodiment of the present invention will be described below.
2 shows an embodiment of the present invention. In the fourth embodiment, the transport device is provided with an inspection start signal section for generating an inspection start signal or an inspection start signal device for generating an inspection start signal in the vicinity of the inspection target surface 1. Therefore, it becomes possible to obtain the inspection start signal from the transport device and the like, and the image memory writing control circuit 32 is operated in synchronization with the inspection start signal.

[0067]

As described above, the present invention has the following effects. The first effect is that since the determination process does not depend on the pattern to be inspected, it is possible to easily inspect a wide variety of small amounts.
Furthermore, it is possible to perform inspection even when a plurality of types of inspection objects are supplied in a mixed state.

The second effect is that the inspection can be performed without generating an erroneous defect signal even if there is a speed change or a pattern deformation caused by the side of the transport device.

[Brief description of drawings]

FIG. 1 is a block diagram of the present invention.

FIG. 2 is a block diagram of an image processing apparatus.

FIG. 3 is an area configuration of small blocks.

[Explanation of symbols]

1 Inspection target surface 2 Solid-state imaging device 3 Image processing device 4 Position detection device 31 image memory 32 image memory writing control circuit 33 CPU 34 small block memory 34A small block average value memory 34B Small block standard deviation memory 34C Small block standard deviation maximum value memory 34D Small block standard deviation minimum value memory 35 Medium block memory 35A Medium block standard deviation value memory 35B Medium block standard deviation maximum value memory 35C Medium block standard deviation minimum value memory 36A Medium block N-valued memory 36B Medium block transport direction projection memory 37 Scanning direction projection memory 37A Scanning direction start projection memory 37B Scanning direction central projection memory 37C Scanning direction end projection memory 38 threshold memory 39 Difference memory PYN Number of patterns in transport direction PYS transport direction pattern size PXN number of patterns in scanning direction PXS scan direction pattern size SBX Number of pixels in small block scanning direction SBY Number of pixels in small block transport direction MBX Number of small blocks in medium block scanning direction MBY Medium block Number of small blocks in transport direction VD digital image data signal PG transport position detection signal HB transport direction resolution WB scanning direction resolution PW position resolution K1 coefficient K2 Scanning direction resolution coefficient K3 Transport direction decomposition factor K4 to K7 coefficient K8 fixed threshold coefficient K9 judgment coefficient

Continued front page    F term (reference) 2F065 AA49 AA56 BB02 BB18 CC02                       CC19 DD06 FF04 FF17 FF67                       JJ02 JJ03 JJ25 JJ26 MM03                       PP12 QQ24 QQ25 QQ31 RR09                       UU05                 2G051 AA51 AB02 CA04 DA06 EA08                       EA11 EA12 EA14 EA16 EB01                       EB02 EC01 ED04 ED08 ED22                 4M106 AA01 AA02 CA39 DB04 DJ11                 5B057 AA03 CA08 CA12 CA16 CC02                       CD05 CE09 CE12 CH11 DA03                       DB02 DB09 DC07 DC22 DC32                 5L096 AA06 BA03 CA02 EA13 EA35                       FA13 FA32 FA33 GA08 GA19                       HA07 LA05

Claims (9)

[Claims] 1. Data obtained from an image pickup device for picking up an object to be inspected having a similar pattern conveyed in the conveying direction in a scanning direction is divided into small blocks, and an average value and a standard deviation value of brightness values of the small blocks are divided. A small block memory that stores the standard deviation value of the lightness value of the middle block configured by the small blocks, and the difference between the standard deviation maximum value and the standard deviation minimum value of the middle block is N (N
Is a positive integer), and a medium block N-valued memory for N-valued and storing the contents of the medium block memory according to an N-valued threshold value obtained by division by N. A medium block carrying direction projection memory for storing the total value of the medium blocks at the same position and a central portion of the medium blocks at the same position in the scanning direction are divided into three, and the total value in each scanning direction is divided. A scanning direction projection memory for storing, detecting the start and end of the pattern from the small block and the middle block, normalizing the size of the pattern, calculating the difference between the patterns,
A determination threshold value is calculated from the medium block, the difference between the patterns is binarized using the determination threshold value, the continuity of defect candidate locations is determined, and the defect is output. A similar periodic pattern inspection device. 2. An image pickup device for picking up an image of an inspection target having a similar pattern in a scanning direction, an image processing device to which data from the image pickup device is input, and a transfer device for transferring the inspection target in the transfer direction. A position detection device that outputs a conveyance position detection signal, the image processing device stores an image memory that stores data from the imaging device, and data from the imaging device based on the conveyance position detection signal. And an image memory writing control circuit for storing the image memory in the image memory, and a small block memory for dividing the data stored in the image memory into small blocks and storing an average value and a standard deviation value of the brightness values of the small blocks. A medium block memory which is divided into medium blocks composed of the small blocks and stores a standard deviation value of the lightness values of the medium blocks, and a standard deviation maximum value of the medium blocks. And the standard deviation minimum difference is N
A medium block N-valued memory that N-values and stores the contents of the medium block memory according to an N-valued threshold value obtained by dividing by (N is a positive integer). A medium block transport direction projection memory that stores the total value of the middle blocks at the same position in the scanning direction and a central portion of the middle block at the same position in the scanning direction are divided into three, and the respective scanning directions are divided. A scanning direction projection memory that stores the above total value, a CPU that inputs the initial setting value of the pattern and performs processing in the image processing apparatus, and a determination calculated based on the contents of the medium block N-valued memory A similar periodic pattern inspecting apparatus, comprising: a threshold value memory for storing a threshold value for use in storage; and a difference memory for storing a difference between the patterns. 3. The analogy according to claim 1 or 2, wherein the small blocks are configured to overlap one pixel or more in the carrying direction and the scanning direction of the inspection object. Periodic pattern inspection device. 4. The small block memory, wherein the small block memory stores an average value of brightness values of the small blocks, and a small block storing a standard deviation value of brightness values in the small blocks for each small block. A deviation value memory, a small block standard deviation maximum value memory that stores the maximum standard deviation value of the brightness values of the small blocks, and a small block standard deviation that stores the minimum standard deviation value of the brightness values of the small blocks A maximum value memory, wherein the middle block memory stores a standard deviation value of the brightness value in the middle block for each middle block; and a standard deviation value of the brightness value of the middle block. A medium block standard deviation maximum value memory for storing a maximum value of the medium block and a medium block standard deviation maximum value memory for storing a minimum value of the standard deviation values of the lightness values of the medium blocks. A direction projection memory divides a central portion of the middle block at the same position in the scanning direction into three, and stores a total value in each of the scanning directions, a scanning direction start projection memory, a scanning direction central projection memory, 3. The similar periodic pattern inspection device according to claim 1, further comprising a scanning direction end projection memory. 5. The number of pulses of the transport position detection signal is measured, and the next data is written in the image memory only when the inspection target surface of the inspection target moves by the resolution in the transport direction. The similar periodic pattern inspection device according to claim 2. 6. The image memory writing control circuit does not control writing to the image memory when the conveying device conveys the inspection object surface of the inspection object at a constant speed. The similar periodic pattern inspection device according to claim 2, wherein the memory stores all the data from the imaging device. 7. If it is necessary to perform an inspection only when the transfer device has a speed higher than a certain speed, the transfer speed of the transfer device is calculated from the period of the pulse of the transfer position detection signal and designated. Only when the transport speed exceeds the specified
3. The image memory write control circuit is configured so that the image memory stores data.
The similar periodic pattern inspection device according to. 8. The transport device outputs an inspection start signal for starting inspection of the pattern, and operates the image memory writing control circuit in synchronization with the inspection start signal. 2. The similar periodic pattern inspection device according to 2. 9. An image memory storing step of storing data of an image of an inspection object having a similar pattern in an image memory, and dividing the data of the image memory into small blocks, and an average value and a standard deviation value of brightness values. A small block dividing step, a medium block dividing step of dividing into medium blocks composed of the small blocks to calculate a standard deviation value of the lightness value, and a threshold value obtained from the standard deviation value of the medium blocks. N-value conversion step of converting the data of the middle block into N-value (N is a positive integer), and the start end of the pattern in the small block of the middle blocks detected by projecting the N-valued data And a pattern detection step of performing end detection, a difference memory storing step of normalizing the size of the pattern, calculating a difference between the patterns and storing the difference memory, and the middle block. And a threshold value calculating step for calculating a threshold value for judgment, and binarizing the data in the difference memory by using the threshold value for judgment to judge the continuity of defect candidate points to detect defects. And a defect output step of outputting the similar defect pattern inspection method.