JP2008139262A - Detect inspection method and inspection apparatus using the same - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To prevent variations in pattern forming positions and pattern sizes from affecting detection precision of defects even if those variations are present. <P>SOLUTION: Local regions are set in an image of a work in which linear circuit patterns are formed in a restricted direction to created a histogram on a concentration gradient direction of edge-constituting pixels in the regions. By analyzing whether the histogram is matched with a histogram of the circuit patterns in the restricted direction, it is determined whether defects are included in the local regions. <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

  The present invention relates to a method of image processing, in which a workpiece on which a predetermined number of linear patterns having a limited orientation are formed is to be inspected, and whether or not defects due to scratches, dirt, or the like have occurred on the surface on which the linear pattern is formed. The present invention relates to a method of inspecting using the method and an inspection apparatus for executing the method.

  As a technique often used when performing defect inspection on a workpiece on which a pattern having a predetermined shape is formed, an image obtained by imaging the workpiece to be inspected (hereinafter referred to as an “inspection target image”) and a preliminarily used method. There are a method of extracting a difference between the two by a difference calculation process with a registered reference image, and a method of obtaining the similarity of the inspection target image with respect to the reference image by a normalized correlation calculation.

  In addition, in order to accurately detect defects occurring in the vicinity of the edge, the inspection apparatus previously developed by the applicant expands and contracts the reference image and extracts the difference between the images to be inspected with respect to these images. (See Patent Document 1).

JP 2002-140695 A

  The reference image used in the various methods listed above is obtained by imaging a non-defective product model having no workpiece defect. When a large number of workpieces are inspected, it is necessary to register a model for each workpiece. Therefore, it is necessary to prepare a memory with a large capacity, which increases the cost.

  Also, in the inspection using the reference image, when a minute defect is detected for a work having a fine pattern, such as a printed circuit board on which a circuit pattern is formed, the pattern formation position and size are slightly Even if there is only a variation, the variation may be erroneously detected as a defect.

  The present invention pays attention to the above-mentioned problems, makes it possible to detect a defect without having to register a reference image, and even if there is a variation in the pattern formation position and size, the influence is detected. The issue is to prevent the accuracy from being affected.

  In the first defect inspection method for solving the above-described problem, a local region of a predetermined size is set in each of a plurality of locations of an image generated by imaging a surface to be inspected, and edge constituent pixels in the local region The process of creating a histogram for the angle data determined by the density gradient direction and whether or not the angle data corresponding to the peak appearing in the histogram conforms to the limited direction of the linear pattern is checked for the presence or absence of defects. The process of discriminating is executed for each set local area.

  The above method can be applied when the workpiece to be inspected is imaged in a certain direction and the angle corresponding to the direction of the linear pattern on the image is fixed. If the workpiece is in good condition, the edge component pixels included in the image to be inspected are all considered to form edges along the limited direction of the linear pattern, so even in the histogram, the limited direction of the linear pattern It is considered that a peak occurs near the angle corresponding to. On the other hand, since it is considered that the inspection target image of a workpiece having a defect includes various edges facing in a direction different from the direction of the linear pattern, the histogram created in the local region including the defect is There is a high possibility that a peak occurs at a position other than the angle corresponding to the limited direction of the linear pattern. That is, it can be considered that a defect is included in the local region when the angle data appearing in the histogram does not match the limited orientation of the linear pattern.

  Next, in the second defect inspection method, a local region of a predetermined size is set at each of a plurality of locations of an image generated by imaging the surface to be inspected, and the density gradient of the edge constituent pixels included in the local region Determining the presence or absence of defects by collating the process of creating a histogram with respect to angle data determined by the direction, and whether or not the degree of variation in frequency in the created histogram matches a histogram with a linear pattern with a limited orientation. The process is executed for each set local area.

  The above method can be applied when it is difficult to specify the direction of the linear pattern in the image, such as when the orientation of the workpiece during imaging varies.

  In one aspect of the second method, among the frequency data in the histogram, the number of frequency data exceeding a predetermined threshold is compared with a reference value based on a histogram with a linear pattern of limited orientation, It is determined that there is a defect when the number of frequency data exceeds a reference value. In another aspect, the number of peaks in the histogram is compared with a reference value suitable for the number of orientations limited for the linear pattern, and a defect is determined when the number of peaks exceeds the reference value.

  In addition, the “direction” of the linear pattern to be inspected in the present invention is not limited to a specific direction, and may be limited to a plurality of directions. For example, with regard to the circuit pattern of the printed circuit board, there are cases where the pattern is limited to three types: a linear pattern extending in the horizontal direction, a linear pattern extending in the vertical direction, and a linear pattern extending in the direction of 45 degrees obliquely.

  An inspection apparatus according to the present invention includes an imaging unit for imaging a workpiece in which a predetermined number of linear patterns having a limited orientation are formed, and processes an image generated by the imaging unit to form a linear pattern. Inspection execution means for inspecting whether or not the surface is defective. Further, the inspection execution means extracts edge constituent pixels from the image generated by the imaging means, calculates edge data for the extracted edge constituent pixels based on the density gradient direction, and a plurality of images. With respect to angle data determined by the density gradient direction of the edge constituent pixels in the local region set by the region setting unit, using the processing result of the region setting unit and the edge processing unit, each of which sets a local region of a predetermined size in each place, Histogram creation means for creating a histogram, a memory for registering data necessary for inspection, a discrimination means for discriminating the presence / absence of a defect using the histogram created by the histogram creation means and the data registered in the memory, For each local area set by the area setting means, the histogram creation means and the judgment Run the process by means, and control means for outputting a discrimination result by the discrimination unit is provided.

  When the first defect inspection method is executed, an angle corresponding to a limited direction of the linear pattern is registered in the memory. The determining means determines whether or not there is a defect by checking whether or not the angle data corresponding to the peak appearing in the histogram matches the angle registered in the memory.

  When executing the second defect inspection method, the memory has a predetermined parameter indicating the degree of variation in the frequency of the histogram (for example, the number of frequency data exceeding a predetermined threshold, the number of peaks in the histogram), A reference value based on a histogram with a linear pattern in a limited direction is registered. The discriminating unit discriminates the presence or absence of a defect by comparing the value of the pattern in the histogram created by the histogram creating unit with a reference value registered in the memory.

  According to the present invention, it is not necessary to register a reference image prior to inspection if the work has a limited linear pattern orientation. In addition, since a local region having a size corresponding to the size of the defect to be detected is set and an image in this region is processed, even a minute defect can be detected.

  Furthermore, since the presence or absence of defects is determined based on whether or not there are many edge constituent pixels having different density gradient directions from the edge constituent pixels of the linear pattern, there is a variation in the formation position and size of the linear pattern. However, the defect can be detected with high accuracy by adjusting the setting position and the number of local regions.

FIG. 1 shows the configuration of an inspection apparatus according to an embodiment of the present invention.
This inspection device targets a workpiece printed with a linear circuit pattern with a limited orientation, such as a printed circuit board or a molded product, and causes defects such as scratches, dirt, and defective printing on the circuit pattern forming surface. It comprises a camera 1 (imaging means) and a controller 2 (inspection execution means) for imaging a workpiece to be inspected.

  The controller 2 includes an image input unit 3, an edge extraction unit 4, an image memory 5, an input / output interface 6, an image output unit 7, a monitor 8, a control unit 10, and the like. The image input unit 3 includes a camera interface and an A / D conversion circuit. The image input unit 3 inputs a grayscale image signal from the camera 1 and digitally converts it. The converted digital image is stored in the image memory 5 and given to the edge extraction unit 4.

  The edge extraction unit 4 includes a circuit for differentiating the digital image and calculating the magnitude of the density gradient for each pixel, and a circuit for calculating the angle data indicating the direction of the density gradient for each pixel. That is, the magnitude of the density gradient and the density gradient direction are obtained for all the pixels constituting the input image. These calculation results are edited into pseudo images respectively associated with the coordinates of the pixels.

  In the following, the magnitude of the density gradient is referred to as “edge strength”, the angle data indicating the density gradient direction is referred to as “edge code”, and the pseudo image based on these is referred to as “edge strength image” and “edge code image”. In this embodiment, each pseudo image is generated by a dedicated circuit for speeding up the processing. However, the present invention is not limited to this, and each image may be generated by software processing of the control unit 10.

  The image memory 5 stores a digitally converted inspection target image, and an edge strength image and an edge code image generated from the inspection target image. Further, a defect image set in the inspection described later is also stored in the image memory 5.

  The input / output interface 6 is connected to an input unit (keyboard, mouse, etc.) and an external device (not shown). The image output unit 7 includes a D / A conversion circuit, an interface circuit to the monitor 8, and the like. The image output unit 7 can also read out a predetermined type of image data from the image memory 5 in accordance with a command from the control unit 10 and display it on the monitor 7.

  The control unit 10 includes a CPU 11, a ROM 12, a RAM 13, and a hard disk 14. The hard disk 14 stores programs for executing various processes relating to inspection (including programs that embody the flowcharts of FIGS. 3 to 7 described later) and setting data. The CPU 11 inspects the workpieces conveyed in the imaging target area of the camera 1 in order using these.

  Here, before describing the details of the inspection, a method of calculating the edge strength and the edge code will be described. In this embodiment, the concentration gradient strengths EX (x, y) and EY in the X and Y axis directions are executed by performing edge extraction calculation such as Sobel calculation while paying attention to each pixel of the image to be inspected in order. Calculate (x, y). Then, using these values EX (x, y) and EY (x, y), the edge strength Ei (x, y) is calculated by the following equation (A). Further, the edge code Ec (x, y) is calculated by executing the equation (B) or (C) based on the magnitude relationship between EX (x, y) and EY (x, y). Note that (x, y) is the coordinates of the pixel of interest (the same applies to the following). For convenience of data processing, Ei (x, y) and Ec (x, y) are both integers rounded to one decimal place. Represent as

Expression (A): Ei (x, y) = | EX (x, y) | + | EY (x, y) |
(B) Formula: Ec (x, y) = arctan (EX (x, y) / EY (x, y))
When EX (x, y) ≦ EY (x, y) (C) Expression: Ec (x, y) = arctan (EY (x, y) / EX (x, y))
When EX (x, y)> EY (x, y)

  According to the above formulas (B) and (C), the edge code is represented as an integer in the range from 0 degrees to 45 degrees. In other words, of the four directions defined by the positive and negative of the horizontal axis (x axis) and the vertical axis (y axis), the angle formed by the direction closest to the vector indicating the concentration gradient direction and the vector of the concentration gradient direction , An edge code will be represented.

  However, the display method of the edge code is not limited to this. For example, the edge code may be represented by a numerical value in the range of 0 to 90 degrees as an angle in the clockwise direction from one of the positive and negative axes of x and y. Or you may represent by the numerical value of the range to 0 to 180 degree | times as a clockwise and counterclockwise angle on the basis of the positive direction of the x-axis. Further, instead of a vector indicating the density gradient direction, an edge code having an angle indicating a direction orthogonal to the vector may be used.

  In the inspection apparatus having the above-described configuration, a local area of a predetermined size is scanned along the x and y axis directions in the edge intensity image and the edge code image, and is described later using the edge intensity and edge code in the local area. By performing the process, it is determined whether or not there is a pixel representing a defect in the local region.

  The determination result is expressed as the defect image described above. The defect image is a binary image having the same size as the original grayscale image, and the pixel indicating the defect (hereinafter referred to as “defective pixel”) is “1”, and the other pixels are “0”. Each is represented by data. In the final stage of the inspection, the number of defective pixels in the defective image is counted, and the count value is compared with a predetermined reference value to determine whether the work to be inspected is good or bad.

Here, the image processing executed in the local region and the processing for determining the presence / absence of a defect will be described in detail.
In this example, pixel data at the same coordinates as each pixel in the local region is read from the edge intensity image or edge code image, and a histogram indicating the distribution state of the edge code of each pixel (hereinafter referred to as “edge code histogram”). Or simply referred to as “histogram”) and analyzing this histogram to determine whether or not a defect is included in the local region.

  FIG. 2 shows an example of setting a local region for an inspection target image and a specific example of the above edge code histogram. The upper part of the drawing shows an image of a workpiece having a defect (hereinafter referred to as “defect image”) and a histogram created from a local region set in a defect portion of the defect image. In the lower row, the same circuit pattern is formed, but the image is created from a defect-free workpiece image (hereinafter referred to as “defect-free image”) and a local region set at the same position as the upper image. Indicates. As shown in this example, the histogram created in the local region including the defect is significantly different from the histogram when there is no defect.

  FIG. 3 is a flowchart showing a procedure for creating the edge code histogram. In the flowcharts of FIG. 3 and subsequent figures, θ is an angle constituting the horizontal axis of the histogram, and varies in the range from 0 to 45. H (θ) is a frequency corresponding to the angle θ in the histogram. Moreover, x and y are the coordinates of the pixel of interest in the flowchart (FIG. 7) of the overall procedure for inspection described later, and are set as the center pixel of the local region. Further, the width of the local region in the x-axis direction is Bx + 1, the width in the y-axis direction is By + 1, and the x and y coordinates of each pixel in the local region are specified by i and j counters. “ST” in each flowchart is an abbreviation of step.

  The first loop of ST501 and ST502 to 504 in FIG. 3 means a process of resetting each frequency of the histogram to zero. Next ST505-512 means the process which completes an edge code histogram using the edge code and edge intensity | strength of each pixel in a local area | region. Specifically, while paying attention to each pixel in the local region in order, the edge code Ec (i, j) and edge intensity Ei (i, j) of the target pixel are read (ST507), and the edge code Ec (i, j, The frequency data is updated by adding the edge strength Ei ((i, j) to the current value of the frequency data H (Ec (i, j)) corresponding to j) (ST508). By setting the weighted frequency, even if there is some noise in the original grayscale image, the frequency of the edge code corresponding to the noise is not so high, and the edge code corresponding to the true edge constituent pixel is high The frequency can be set.

  In FIG. 3 described above, the entire image in the local region on the edge intensity image or the edge code image is processed instead of processing the inspection target image itself generated by the camera 1 (imaging means). However, these pseudo images are all derived from the inspection target image, and since the frequency weighted by the edge strength is set in the histogram, a local region is set in the inspection target image, It is considered that substantially the same processing as that in the case of creating a histogram using only the edge codes of the edge constituent pixels in the region is executed.

  As shown in each image of FIG. 2, the circuit pattern of the printed circuit board is generally formed so as to extend along any one of the horizontal direction, the vertical direction, and the oblique 45-degree direction. Therefore, according to the expressions (B) and (C) described above, it is considered that the theoretical edge code is 0 degree or 45 degrees in the constituent pixels of the edge of the circuit pattern. Therefore, when the processing of FIG. 3 is performed on a local area where no defect exists, the range in which the frequency is high is the above even if the work or circuit pattern in the image has some rotational deviation. It is thought that it is limited to around the theoretical value of. In the example of FIG. 2 as well, in the edge code histogram of a defect-free image, the frequency is in the range of 0 to 5 degrees, and the frequency corresponding to an angle of 15 degrees or more is zero or a number close to zero. ing.

  On the other hand, the direction of the edge corresponding to the defect does not have the regularity as described above, and there are edges that face in various directions. Therefore, when the process of FIG. 3 is performed on a local region including a defect, the frequency is considered to be higher at a much larger angle than the histogram when there is no defect. In the example of FIG. 2 as well, in the histogram of the upper defective image, the frequency difference between the angles is small, and the frequency of a certain value varies throughout.

  Based on the histogram characteristics as described above, in this embodiment, the histogram is analyzed by one of the processes shown in FIGS. 4 to 6 to determine whether or not a defect is included in the local region. Note that the determination result is represented by a defect flag d_flg.

The process of FIG. 4 is to determine the presence or absence of a defect by a process of checking whether or not the angle corresponding to the maximum value of the frequency in the histogram matches the limited direction of the circuit pattern. Loop initial ST101 and ST102~106 the maximum value _{H max} of the frequency in the histogram, and the angle _{T max} corresponding to the maximum value _{H max,} is for obtaining.

When the above process ends, in ST107, the maximum frequency value H _max is compared with a predetermined threshold value Hth. If H _max is equal to or greater than the threshold value Hth, the process proceeds to ST108, and the angle T _max is compared with the reference angles T1, T2 (T1 <T2). Here, when T _max is located between T1 and T2, ST108 is “YES”, the process proceeds to ST109, and the defect flag d_flg is turned on (d_flg is set to “1” for data processing). .) On the other hand, if T _max is equal to or less than T1 or equal to or greater than T2, ST108 is “NO”, the process proceeds to ST110, and the defect flag d_flg is turned off (d_flg is set to “0” for data processing). .

  For T1 and T2, an angle corresponding to the boundary between an angle range in which a frequency equal to or higher than a predetermined value is obtained in a histogram without defects and a low frequency range by using a histogram of samples with no defects in advance. Is set. For example, in the sample histogram, the angle represents a limited direction of the circuit pattern, and the maximum value is within a range of about 5 degrees with respect to the angle (0 degree or 45 degrees) that should theoretically have the maximum degree. If it is recognized, T1 = 5 and T2 = 40.

According to the processing of FIG. 4 above, when the angle T _max corresponding to the maximum value H _{max of the} histogram is in the range from T1 to T2, this angle T _max is suitable for the limited orientation of the circuit pattern. However, it is determined that there is a defect in the local region. On the other hand, if the angle T _max is equal to or smaller than T1 or equal to or larger than T2, the angle T _max conforms to the limited direction of the circuit pattern, and it is determined that there is no defect in the local region. .

Also, in the absence of the most edge constituent pixels in the local region, because the maximum value _{H max} calculated in loop ST102~106 smaller than the threshold value Hth, the ST110 ST 107 becomes "NO" move on. Therefore, also in this case, it is determined that there is no defect in the local region.

  In the above processing, only the maximum value in the histogram is compared with the reference angles T1 and T2. However, the present invention is not limited to this, and the maximum value in the histogram has a value equal to or larger than a predetermined reference value. Each of the processes of ST107 may be executed. In this case, if any one of the maximum values reaches a value between T1 and T2, the defect flag d_flg may be turned on.

  Next, the processing of FIG. 5 determines the presence or absence of a defect by obtaining the number C of frequency data whose values are larger than the preceding and succeeding data in the histogram and comparing the value of C with a predetermined reference value CA. is there. As shown in FIG. 2, in the defect-free image histogram, except for the vicinity of the maximum value, there is no noticeable undulation, and the value tends to decrease as the distance from the maximum value increases. On the other hand, in the histogram of an image with a defect, a large change in undulation occurs even in an angle range that does not match the circuit pattern. In other words, the range of undulation changes is large.

  The above C is considered to represent the degree of undulation of the histogram. Further, when the degree of occurrence of undulation is large, it is considered that the variation in frequency is large, and therefore the value of C may be regarded as a parameter representing the degree of variation in the frequency of the histogram.

  In FIG. 5, in the first ST201 and ST202 to 206, the value of C is obtained for each angle of 1 to 44 degrees, and in the next ST207, the value of C is compared with the reference value CA. If C is greater than CA, ST207 is “YES” and the process proceeds to ST208, where the defect flag d_flg is turned on. On the other hand, if the number C of maximum values is less than or equal to the reference value CA, the process proceeds to ST209, and the defect flag d_flg is turned off. The reference value CA is desirably set in accordance with the value of C obtained from the histogram of the sample with no defect, similarly to T1 and T2 in the process of FIG.

  According to the above processing, if the count value C is larger than the reference value CA, the degree of variation in the histogram frequency does not match the histogram based on the limited direction of the circuit pattern, and the local region has a defect. It will be judged to be included. On the other hand, when the count value C is less than the reference value CA, it is determined that the degree of variation in the histogram frequency conforms to the histogram based on the limited direction of the circuit pattern, and no defect is included in the local region. Is done.

Next, in the process of FIG. 6, after obtaining the maximum value H _max in the histogram in ST 301 to 305 by the same process as ST 101 to 106 of FIG. 4, this H _max is equal to or greater than the reference value Hth. in, run a half-maximal value _{H max (H max} / 2) or more values processing for obtaining the number C1 of frequency data taking (ST307~311).

In ST312, the count value C1 is compared with a predetermined reference value CB. Here, when C1 exceeds CB, ST312 becomes “YES”, the process proceeds to ST313, and the defect flag d_flg is turned on. Meanwhile, C1 is the case when it is less CB, or the maximum value _{H max} is equal to or less than the reference value Hth, the process proceeds to ST 314, to turn off the fault flag D_flg.

  When no defect is included in the local region, as shown in the lower histogram of FIG. 2, the frequency data indicating a value of half or more of the maximum value of the histogram is limited to the vicinity of the maximum value. On the other hand, in the histogram in the case where the defect is included in the local region, the frequency is generally increased, so that the value of C1 is significantly increased as compared with the case where there is no defect. Therefore, it can be considered that the count value C1 functions as a parameter indicating the degree of variation in the frequency of the histogram.

The threshold value for determining the frequency data to be counted is not limited to the above H _max / 2, but may be a value obtained by multiplying the area of the local region by a predetermined coefficient, for example.

  According to the processing of FIG. 6, if the count value C1 is equal to or greater than the reference value CB, the degree of variation in histogram frequency does not conform to the histogram based on the limited direction of the circuit pattern, and the local region is defective. Is determined to be included. On the other hand, when the count value C1 is lower than the reference value CB, the degree of variation of the histogram frequency is adapted to the histogram based on the limited direction of the circuit pattern, and there is no defect in the local region. To be judged.

  The above three analysis processes are selected according to circumstances such as the imaging state of the workpiece and the formation state of the circuit pattern. For example, when the orientation of the workpiece during imaging is constant and the rotation deviation of the circuit pattern is small, the angle corresponding to the limited orientation of the circuit pattern corresponds to a substantially constant position on the histogram. The process of FIG. 4 can be selected. On the other hand, when it is difficult to specify the angle corresponding to the limited orientation of the circuit pattern on the histogram because the workpiece at the time of imaging is not facing a specific direction or the rotation deviation of the circuit pattern is large It is desirable to select the processing of FIG. 5 or FIG.

  The histogram analysis process is not limited to the above three processes. For example, the peak apex (maximum value) included in the histogram may be extracted, and the number of extractions may be compared with a reference number (1 or 2) corresponding to the number of orientations of the circuit pattern limited. In this case, the orientation of the circuit pattern included in the local area is obtained in advance for each local area using an image of a non-defective product model or CAD data, and a reference number corresponding to the result is set. Is desirable.

FIG. 7 is a flowchart showing the procedure of the entire inspection process.
The work (printed circuit board) to be inspected is transported and positioned in the imaging target area of the camera by a transport mechanism (not shown). The process of FIG. 7 is started in accordance with the introduction of the work. In the first ST1, the value of each pixel of the defect image is set to the initial value “0”.

  In ST2, the introduced work is imaged by the camera 1. The image generated by this imaging is digitally converted by the image input unit 3 and an edge strength image and an edge code image are generated by the edge extraction unit 4.

  The next ST3 to ST15 pay attention to a plurality of pixels in a rectangular area specified by a preset upper left vertex (x1, y1) and lower right vertex (x2, y2) in order, and a local center around that pixel. For each region, processing for creating an edge code histogram and processing for determining the presence or absence of a defect are executed.

  Note that sx in ST13 and sy in ST15 are set to a value obtained by adding 1 to the number of unfocused pixels (interval of thinning processing) (when sx = sy = 1, no thinning is performed and the inspection region is set) The processing is performed on all the pixels in the target). The values of sx and sy can be appropriately changed according to the size of the defect to be detected, similarly to the size of the local region.

  The specific processing in the above loop will be described in order. After the local region is set in ST5, the sum BA of the edge intensities of the respective pixels included in the local region is calculated in ST6. Is compared with a predetermined reference value. The reference value in this case is set based on the edge strength caused by noise such as density unevenness. If BA is smaller than the reference value, the following process is skipped and the process proceeds to ST13. Therefore, the target pixel on the defective image in this case remains the initial value “0”.

  If the above BA is greater than or equal to the reference value, the process proceeds to ST8 and a histogram in the local region is created. In ST8, the series of processing shown in FIG. 3 is executed. In the next ST9, a smoothing process for adjusting the waveform of the created histogram is performed. Specifically, for each frequency data, an average value of a total of three data, that data and the frequency data before and after that data, is obtained, and the original frequency data is replaced with the average value.

  In ST10, a histogram analysis process according to the procedure shown in any of FIGS. When the defect flag d_flg is set to ON in this analysis processing, ST11 becomes “YES” and the process proceeds to ST12, and the data D (x, y) of the pixel of interest in the defect image is set to the initial value “0”. By changing from 1 to “1”, the target pixel is set as a defective pixel.

  When the above-described loop of ST3 to 14 ends, a defect image reflecting the hourly histogram analysis processing result is generated. In ST17, a process for generating a corrected image in which defective pixels that exist locally are deleted is executed for the defective image.

The image correction process will be briefly described with reference to FIG.
First, in addition to the original defect image, a storage area for the corrected defect image is set in the image memory 5, and each pixel of the image is set to an initial value “0”. Then, attention is sequentially paid to defective pixels in the original defective image, and it is determined whether or not the value of this pixel is maintained even in the new defective image.

  In the above determination processing, the pixels g1 to g4 that are located sx in the x direction or sy in the y direction from the target pixel g0, and the pixels that are located in the x and y directions with a distance of sx and sy, respectively. The data in the original defect image is read for a total of 8 pixels g5 to g8. These pixels g1 to g8 correspond to the central pixel of the local region set next to the local region centered on the target pixel g0. In this embodiment, only when these eight pixels g1 to g8 are all defective pixels, the value of the target pixel g′0 of the corrected defect image is maintained at the same value “1” as before the correction (FIG. 8). (Refer to (1).) If even one of the eight pixels is not a defective pixel, the target pixel g′0 of the corrected image is set to the initial value “0” ((2) in FIG. 8). reference.). By such processing, only pixels corresponding to defects having a size greater than a certain value are maintained as defective pixels.

  Returning to FIG. 7, when the above correction processing is completed, in ST18, the number SD of defective pixels in the corrected defect image is counted. In the next ST19, the value of this SD is compared with a predetermined reference value Sth. If SD is equal to or greater than the reference value Sth, ST19 is “YES”, the process proceeds to ST20, and the workpiece to be inspected is determined to be “defective”. If SD is smaller than the reference value Sth, the process proceeds to ST21 and it is determined that the workpiece to be inspected is a non-defective product.

  Thereafter, the process proceeds to ST22, and the above determination result is output to the external device or the monitor 8. When the next workpiece to be inspected is further conveyed, ST23 becomes “YES” and the process returns to ST1 and the above procedure is repeated. On the other hand, if the next workpiece has not been transported even after the predetermined time has elapsed, or if an inspection end command has been input, ST23 is “YES” and the inspection is terminated.

It is a block diagram which shows the structure of an inspection apparatus. It is explanatory drawing which matches and shows the image with a defect, an image without a defect, and the histogram produced from each image. It is a flowchart which shows the preparation processing procedure of a histogram. It is a flowchart which shows the procedure of the analysis process of a histogram. It is a flowchart which shows the procedure of the analysis process of a histogram. It is a flowchart which shows the procedure of the analysis process of a histogram. It is a flowchart which shows the procedure of the whole process of a test | inspection. It is explanatory drawing which shows the correction processing method of a defect image.

Explanation of symbols

1 Camera 2 Controller 4 Edge Extraction Unit 10 Control Unit 11 CPU
14 Hard disk

Claims (6)

A method for inspecting whether or not a defect has occurred on the formation surface of the linear pattern, with a workpiece in which a predetermined number of linear patterns having a limited orientation are formed as inspection targets,
Processing for setting a local area of a predetermined size in each of a plurality of locations of an image generated by imaging the surface to be inspected, and creating a histogram for angle data determined by the density gradient direction of the edge constituent pixels in the local area; A process for determining whether or not there is a defect by checking whether or not the angle data corresponding to the peak appearing in the histogram matches the limited direction of the linear pattern is executed for each set local region. A defect inspection method characterized by: A method for inspecting whether or not a defect has occurred on the formation surface of the linear pattern, with a workpiece in which a predetermined number of linear patterns having a limited orientation are formed as inspection targets,
A process of setting a local area of a predetermined size in each of a plurality of locations of an image generated by imaging the surface to be inspected, and creating a histogram for angle data determined by the density gradient direction of the edge constituent pixels included in the local area And the process of determining the presence or absence of a defect by comparing whether or not the degree of variation in frequency in the created histogram matches the histogram of the linear pattern in the limited direction, the set local region A defect inspection method characterized by being executed every time.   In the process of determining the presence / absence of the defect, the number of frequency data exceeding a predetermined threshold among the frequency data in the histogram is compared with a reference value based on the histogram based on the linear pattern of the limited orientation. The defect inspection method according to claim 2, wherein a defect is determined when the number of frequency data exceeds the reference value.   In the process of determining the presence or absence of the defect, the number of peaks in the histogram is checked against a reference value suitable for the number of orientations limited for the linear pattern, and the defect is detected when the number of peaks exceeds the reference value. The defect inspection method according to claim 2, wherein it is determined that there is a defect. An imaging unit for imaging a workpiece in which a predetermined number of linear patterns with a limited orientation are formed, and an image generated by the imaging unit is processed so that no defects are generated on the surface on which the linear pattern is formed. An inspection execution means for inspecting whether or not,
The inspection execution means includes
Edge processing means for extracting edge constituent pixels from the image generated by the imaging means and calculating angle data determined by the density gradient direction for the extracted edge constituent pixels;
Area setting means for setting a local area of a predetermined size at each of a plurality of locations of the image;
Using the processing result of the edge processing means, a histogram creation means for creating a histogram for angle data determined by the density gradient direction of the edge constituent pixels in the local area set by the area setting means;
A memory for registering an angle corresponding to a limited orientation of the linear pattern;
A discriminating means for discriminating whether or not there is a defect by checking whether or not the angle data corresponding to the peak appearing in the histogram matches the angle registered in the memory;
An inspection apparatus comprising: control means for executing processing by the histogram creation means and discrimination means for each local region set by the area setting means and outputting a discrimination result by the discrimination means. An imaging unit for imaging a workpiece in which a predetermined number of linear patterns with a limited orientation are formed, and an image generated by the imaging unit is processed so that no defects are generated on the surface on which the linear pattern is formed. An inspection execution means for inspecting whether or not,
The inspection execution means includes
Edge processing means for extracting edge constituent pixels from the image generated by the imaging means and calculating angle data determined by the density gradient direction for the extracted edge constituent pixels;
Area setting means for setting a local area of a predetermined size at each of a plurality of locations of the image;
Using the processing result of the edge processing means, a histogram creation means for creating a histogram for angle data determined by the density gradient direction of the edge constituent pixels in the local area set by the area setting means;
A memory for registering a reference value based on a histogram based on the linear pattern of the limited orientation for a predetermined parameter indicating the degree of variation in the frequency of the histogram;
A discriminating means for discriminating the presence or absence of a defect by comparing the value of the parameter in the histogram created by the histogram creating means with a reference value registered in the memory;
An inspection apparatus comprising: control means for executing processing by the histogram creation means and discrimination means for each local region set by the area setting means and outputting a discrimination result by the discrimination means.

Images