JP2010071876A - Defect inspecting method and defect inspecting device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a technology that does not incorrectly recognize a part other than a defective part as defective even when the shape of a surface varies irregularly during imaging. <P>SOLUTION: (A) When a flexible sheet-like article or the like is imaged while being conveyed, concentration irregularity can be generated in an image by deformation during conveyance. The whole image is divided into two inspection regions A1 and A2, the most frequent value (150) of the concentration is determined in an inspection region A1, a pixel of a concentration lower than the most frequent value is extracted, and adjacent pixels are coupled to each other. The coupled pixels are set as a processing area AD, and the maximum value (100) of the concentration among them is determined. (B) Then, the concentration in the processing area AD is gradation-corrected with the ratio (150/100) of the most frequent value and the maximum value, the influence of shape variation is removed, and then a defect is inspected. <P>COPYRIGHT: (C)2010,JPO&INPIT

Description

  The present invention relates to a defect inspection method and a defect inspection apparatus capable of inspecting a surface defect using an image obtained by imaging an inspection object.

  Conventionally, with respect to this type of defect inspection method or defect inspection apparatus, for example, even when the luminance gradation changes depending on the location, such as an inspection object having an approximate congruent pattern on the surface, the change in luminance of the surface pattern There is known a prior art that performs gradation correction in order to enable detection of a defect with a low luminance change level (see, for example, Patent Document 1).

In other words, this prior art averages luminance data values in a small area in an image obtained by imaging an object to be inspected, and uses the average value while shifting the small area by several pixels to cross-correlate the small areas. By performing data processing, the gradation of the entire image is corrected. Therefore, according to the prior art, the cross-correlation coefficient data processing eliminates the luminance change based on the approximate congruent pattern reflected in the captured image, and instead the luminance change smaller than the luminance change based on the approximate congruent pattern It appears as a deviation between numerical data. Therefore, in principle, it is possible to determine a part where this deviation appears as a defective part.
Japanese Patent Laid-Open No. 2006-9401

  However, in the above-described prior art, since the entire image is comprehensively processed while shifting the small area finely (for example, one pixel at a time), a long processing time is required and the load on the hardware is large. There is a problem of becoming.

  In addition, for an inspected object having flexibility (for example, a sheet-like article), the density of a part of the image may be greatly changed in an irregular shape due to the influence of slackness or undulation during imaging. However, in the prior art method, gradation correction is performed only on a small area of a rectangular pattern corresponding to a known regular surface pattern, so gradation correction is performed on an area where the density changes in an irregular shape. As a result, there is a possibility that the density change region caused by the deformation of the inspection object may be erroneously recognized as a defect.

  Therefore, the present invention can shorten the time required for arithmetic processing, and does not misrecognize other than a defective portion as a defect even when the surface shape irregularly changes during imaging. Is the issue.

  In order to solve the above problems, the present invention employs the following solutions.

First, the present invention is a defect inspection method comprising the following steps.
(1) Acquisition Step This step is a step of acquiring an image obtained by imaging the inspection object. The acquired image may have been captured and stored before this step, or may be captured in this step.

(2) Calculation Step In this step, a representative density value is obtained from the pixels constituting the image acquired in the acquisition step (1). For this representative value, for example, a mode value calculated through statistical processing can be used.

(3) Forming step In this step, pixels having a density lower than the representative value are extracted from the acquired image, and adjacent pixels are combined to form a processing region. For example, when the surface shape of the object to be inspected is changed at the time of imaging, partial density unevenness may be formed in the image even though it is not a defect. Such density unevenness appears as a region where the density is lower than the representative density value in the image. Therefore, by extracting pixels having a density lower than the representative value and connecting adjacent pixels together, it is possible to specify a region where density unevenness occurs in the image.

(4) Correction Step In this step, when the processing region formed in the formation step (3) is larger than the specified area, the gradation of all the pixels included in the processing region is corrected for gradation. Here, the “specified area” is assumed to be an area larger than the area when a defect occurring on the surface is imaged. If the area of the processing region is larger than the specified area, it is due to a change in the surface shape, not the defect portion. Can be determined. Therefore, for the processing region larger than the specified area, the influence of the change in the surface shape can be removed by correcting the density of all pixels. On the other hand, when there is only an area smaller than the specified area, the processing region is not caused by a change in the surface shape, and it can be determined that it is purely a defective portion of the surface, so tone correction is not performed here.

(5) Determination Step In this step, the presence or absence of a defect is determined based on the density after correction for the processing area in the acquired image, and the presence or absence of a defect is determined based on the original density for the other areas. As a result, regarding the portion where the density change has occurred due to the change in the surface shape, the presence / absence of a defect is determined after performing gradation correction using this as a processing region, so that it is not erroneously recognized as a defective portion.

Secondly, the defect inspection method of the present invention can add the following contents in each step.
That is, in the correction step (4), gradation correction can be performed using the ratio between the representative value obtained in the calculation step (2) and the maximum value of the density of pixels included in the processing region as a correction coefficient.

  If the processing area is due to a change in the surface shape, the gradation in the processing area is generally raised based on at least the pixel with the highest density, thereby reducing the effect of the density reduction due to the change in the surface shape. Can be removed. Thereby, the density unevenness of the entire image including the processing region can be smoothed, and erroneous detection can be reliably prevented.

  Thirdly, the defect inspection method of the present invention further includes a dividing step of dividing the image acquired in the acquisition step (1) into a plurality of inspection regions. In this case, in the calculation step (2), the representative value is obtained from the pixels constituting any one of the inspection areas divided in the previous division step in the acquired image. Further, in the forming step (3), pixels extracted from the inspection area are combined to form a processing area. In the determination step (5) above, the presence or absence of a defect is determined by comparing a predetermined threshold value with the density of each pixel. At that time, the pixels in the processing area are represented in the inspection area. It is assumed that the threshold value is corrected using the ratio between the value and the representative value of the density obtained from all the pixels constituting the acquired image as a correction coefficient.

  For example, when an influence due to a complex surface shape change occurs in the entire image, density unevenness may appear in a plurality of stages (for example, three stages) in the entire image. That is, it is assumed that there are a relatively high density area and a low density area in the image as a whole, and there is a further low density area in a part of the low density area. In such a case, the entire image is divided into an area with a comparatively low density and an area with a low density, and an inspection area is set. Then, a processing area is formed for each inspection area, and more accurate gradation correction is possible. It becomes.

  On the other hand, when the inspection area is divided, the density of the defective portion is also raised overall in the area where the density is relatively low, so that in some cases, the gradation correction is excessive and the defect threshold may be exceeded. . Therefore, in the present invention, the threshold value is corrected using the ratio of the representative value of density in the inspection region and the representative value of density seen in the entire image. As a result, it is possible to prevent the defective portion from being overlooked by lowering the inspection hurdle for the processing region subjected to the gradation correction.

  Therefore, fourthly, the defect inspection method of the present invention repeatedly executes the calculation process to the determination process for each of the inspection areas divided in the division process.

  Accordingly, it is possible to perform an appropriate defect inspection on the entire acquired image while performing gradation correction of the processing region for each of the plurality of inspection regions.

  Fifth, the present invention provides a defect inspection apparatus. This defect inspection apparatus can execute the first to third defect inspection methods.

  That is, the defect inspection apparatus includes an image generation unit that captures an image of an inspection object to generate an image, a calculation unit that calculates a representative value of density from an image that forms the image generated by the image generation unit, and a representative from among the images. Extracting pixels having a density lower than the value and combining adjacent pixels together to form a processing area, and if the processing area is larger than a specified area, the density of all the pixels contained therein Correction means for correcting gradation, and determination means for determining the presence / absence of a defect based on the density after correction for the processing area of the image and determining the presence / absence of a defect based on the original density for the other areas. Is.

  According to the defect inspection apparatus of the present invention, for example, while imaging a large number of inspected objects in order to generate an image, the presence or absence of defects is determined by performing arithmetic processing in order using the generated image, or A procedure for continuously capturing images of a long inspected object, generating an image, and sequentially performing arithmetic processing using the appropriately divided images to determine the presence or absence of a defect can be performed as a flow operation. Therefore, for example, it is useful for improving the quality control by improving the efficiency of the work of sequentially inspecting objects to be mass-produced on the production line or continuously inspecting long sheet-like articles. Can be obtained.

  As described above, according to the defect inspection method and the defect inspection apparatus, even if an uneven density portion due to a change in the surface shape of the object to be inspected is included in the image, it is highly accurate without erroneously recognizing this as a defective portion. A defective portion can be detected. In addition, since it is not necessary to divide the entire image and perform correction processing, it is possible to reduce the time required for calculation processing and to reduce the hardware processing load.

Hereinafter, embodiments of the present invention will be described with reference to the drawings.
FIG. 1 is a diagram schematically illustrating a configuration example of the defect inspection apparatus 10. The defect inspection method can be realized by using this defect inspection apparatus 10. Here, first, a configuration row of the defect inspection apparatus 10 will be described.

  The defect inspection apparatus 10 includes a belt conveyor 12 as a transport mechanism, for example. The belt conveyor 12 is an object to be inspected (hereinafter referred to as “work”) placed on the conveyance surface of the endless conveyance belt 12a while traveling in a certain direction along with the rotation of the conveyor roller 12b. W is conveyed in the running direction of the conveyor belt 12a. The belt conveyor 12 includes a motor (not shown) as its power.

  Here, the workpiece W to be inspected (inspected) is, for example, a highly flexible sheet-like article, and the workpiece W is fed from a roll and conveyed on the belt conveyor 12 in the longitudinal direction. . For this reason, the belt conveyor 12 is provided with pinch rollers 12c on the conveyor rollers 12b at both ends, and the work W is fed out from the roll while being sandwiched between the pinch rollers 12c and the conveyor rollers 12, and It is conveyed on the belt conveyor 12. At this time, the workpiece W is held in a state where it is spread horizontally by being supported by the conveyor belt 12a. Note that, for example, a guide plate (not shown) for keeping the workpiece W horizontal together with the conveyor belt 12a may be disposed at the inner peripheral position of the conveyor belt 12a.

  Each conveyor roller 12b receives power from a drive source such as a motor (not shown) and rotates in the same direction (clockwise as viewed in FIG. 1) at a constant speed. The workpiece W sent from the end of the belt conveyor 12 is wound up in a roll again at a winding position (not shown), for example. In addition, although the elongate sheet-like workpiece | work W is mentioned here, the workpiece | work W may be a sheet-like sheet | seat article.

  In the belt conveyor 12, for example, a rotary encoder 14 is connected to one conveyor roller 12b. The rotary encoder 14 is driven by the rotation of the conveyor roller 12b via, for example, the transmission belt 16, and generates a pulse signal at every fixed rotation angle.

  The defect inspection apparatus 10 includes a line sensor 18 as an imaging device, and the line sensor 18 is installed above the belt conveyor 12. The line sensor 18 incorporates a large number of imaging elements arranged in a straight line with a certain width. For the image sensing device used for the line sensor 18, an electron tube type imaging tube or the like can be used in addition to a semiconductor type such as a CCD image sensor or a CMOS image sensor. In addition, a lens 20 is installed on the front surface of the line sensor 18 (in front of the light receiving surface) so that the reflected light from the surface of the work W is imaged on the image sensor.

  In addition, the defect inspection apparatus 10 includes a lighting fixture 22 as a light source that illuminates the surface of the workpiece W. The lighting fixture 22 is also installed above the belt conveyor 12, and the lighting fixture 22, the line sensor 18 and the lens 20 are positioned opposite to each other across the imaging position (main scanning line) in FIG. Yes. Various kinds of illumination lamps such as a cold cathode tube, a discharge tube, and an incandescent lamp can be used for the illumination fixture 22.

  The defect inspection apparatus 10 includes an arithmetic processing unit 24 in addition to these mechanical elements. The arithmetic processing unit 24 is configured using an electronic computer having a central processing unit (CPU), a storage device (ROM, RAM), an input / output interface (I / O), and the like. Examples of electronic computers include general-purpose personal computers and workstations. The arithmetic processing unit 24 receives a pulse signal from the rotary encoder 14 and an image signal from the line sensor 18. The arithmetic processing unit 24 performs various processes on the image data based on these input signals. The processing performed by the arithmetic processing unit 24 will be further described later.

  FIG. 2 is a perspective view illustrating a process in which the workpiece W is imaged by the line sensor 18. The main scanning line L of the line sensor 18 is positioned in the width direction (transverse direction) of the surface of the workpiece W, and the luminaire 22 emits illumination light centering on the main scanning line L. As the work W is transported, the light reflected from the surface of the work W at the position of the main scanning line L passes through the lens 20 and forms an image on the image sensor of the line sensor 18.

  FIG. 3 is a block diagram showing the configuration of the arithmetic processing unit 24 in more detail. The arithmetic processing unit 24 includes an input unit 30, a data processing unit 32, a data storage unit 34, a defect detection unit 36, and an output unit 38 as its constituent elements, and a CPU 40 that controls these operations. Hereinafter, each component will be described. In this example, the data processing unit 32 and the defect detection unit 36 are provided separately from the CPU 40. However, the functions of the data processing unit 32 and the defect detection unit 36 may be realized using the resources of the CPU 40. Good. In addition, the configuration described here with reference to the drawing is merely a preferable example, and another configuration may be adopted for the arithmetic processing unit 24.

  An image signal from the line sensor 18 and a pulse signal from the rotary encoder 14 are input to the input unit 30 via the input interface. The input unit 30 has a function of performing attenuation adjustment on an image signal (analog signal) as necessary while referring to an input pulse signal, A / D converting, and transmitting it to the data processing unit 32.

  The data processing unit 32 processes the transmitted image signal in the form of image data (raster data) for one line. At this time, for example, gray scale density values (0 to 255 gradations) are assigned to each pixel in one line. The data processing unit 32 assigns a column address to each pixel included in one line and assigns a row address for each line.

  The image data for each line processed by the data processing unit 32 is transferred to the data storage unit 34 via the data bus 42. A fixed storage area is secured in the data storage unit 34, and memory addresses respectively corresponding to the row address and the column address assigned to the image data are allocated to this storage area. The transferred image data (density value of 0 to 255) is written in the memory cell (for 1 byte) at the corresponding memory address.

  The defect detection unit 36 accesses the data storage unit 34 via the data bus 42, reads image data from the storage area, and performs defect inspection processing. The defect detection unit 36 writes defect inspection result data in the data storage unit 34 and transfers the inspection result data to the output unit 38 as necessary. A specific aspect of the inspection process performed by the defect detection unit 36 will be further described later.

  In addition to outputting the inspection result data as described above, the output unit 38 outputs imaged image data (image of the workpiece W) as necessary. The data output destination is, for example, an image display device (not shown), another general-purpose computer, a recording medium, or the like. Accordingly, the inspection result can be displayed on the image display device so that the operator can visually recognize the defect on the screen, or the surface image of the workpiece W can be displayed on the screen and the operator can visually recognize the image. .

  In addition, although not particularly illustrated, the arithmetic processing unit 24 may include a data correction unit. The data correction unit reads the image data stored in the data storage unit 34 and performs the correction process. The correction process is performed, for example, for the purpose of correcting unevenness in density due to a periodic change in the conveyance speed by the belt conveyor 12, or correcting changes in density difference due to flickering in room lighting. Such a function of the data correction unit may be realized using resources of the CPU 40.

[Generation of image data]
In the arithmetic processing unit 24, image data is generated through the following process, for example. In other words, when a pulse signal is input from the rotary encoder 14, the arithmetic processing unit 24 performs imaging (imaging) for one line using the line sensor 18. Thereby, in the line sensor 18, each image sensor is activated (exposure state), and an image signal is generated based on the intensity of reflected light from the main scanning line L. Next, when a pulse signal is input, the arithmetic processing unit 24 ends the imaging for one line. Through the processing so far, image data for one line is generated.

  When the image data generation process for one line is completed once, the arithmetic processing unit 24 repeats the next image data generation process again and accumulates them for a plurality of lines to generate one unit of image data. When the workpiece W is a long product as in the present embodiment, the image data generation process may be performed, for example, by dividing the workpiece W in the longitudinal direction.

  That is, registration marks (not shown) are attached to one side edge of the workpiece W at regular intervals, for example, and the registration marks not shown in FIGS. It has a sensor. When the registration mark is detected by the registration mark sensor immediately before the main scanning line L (upstream position in the transport direction), the arithmetic processing unit 24 starts the image data process. When the registration mark sensor detects the next registration mark, the arithmetic processing unit 24 ends the image data generation process for one unit, and then starts the next image data generation process. The image data for each unit generated at this time is recorded in the data storage unit 34 in association with the registration mark number, for example.

[Mode of concentration change]
FIG. 4 is a perspective view showing an aspect in which the image density changes due to various factors during the image forming process. Hereinafter, the aspect of density change will be described.

[Effects of workpiece deformation]
In FIG. 4, (A): For example, it is assumed that the workpiece W is deformed (swelled) so as to wave in the conveying direction. In this case, from the relationship between the illumination angle and the imaging angle, a bright region R1 that strongly reflects light toward the line sensor 18 on the main scanning line L occurs, or a dark region R2 that reduces the amount of reflected light conversely occurs. There are things to do. Further, the end of the workpiece W may be partially loosened, and there may be a dark region R3 where the amount of reflected light falls particularly in the dark region R2.

  In FIG. 4B: In addition to the above example, even if the workpiece W is not deformed in the transport direction, a dark region R3 may also be generated due to slack at the end portion.

[Influence of surface pattern]
In FIG. 4, (C): Even when the workpiece W itself is not deformed, for example, a case is assumed in which the pattern PT is attached to the surface of the workpiece W. In such a case, a bright and dark region due to the deformation of the workpiece W does not occur, but a difference occurs in the density of the entire image depending on the presence or absence of the pattern PT.

  In any case, the image density generally increases in the bright area R1, and conversely the image density decreases in the dark areas R2 and R3 and the design pattern PT. Further, when deformation such as slack occurs, the dark region R3 is generated in an irregular pattern. The defect inspection apparatus 10 of the present embodiment performs the following defect inspection processing in order to avoid misrecognizing this as a defective portion even when irregular density changes occur in the image data. In addition, by applying the defect inspection process of the present embodiment, even when a defect exists in the dark regions R2 and R3 or the pattern PT, the defect can be detected with high accuracy. Become. Details of the defect inspection process will be described below. In addition, the contents of each process for realizing the defect inspection method will be clarified by the following description.

[Defect inspection method]
FIG. 5 is a flowchart showing a procedure example of a defect inspection process executed in the arithmetic processing unit 24. Hereinafter, the outline of the defect inspection process will be described step by step.

  Step S10: First, the defect detection unit 36 acquires (data capture) image data from the data storage unit 34 (acquisition process). The image data acquired at this time is, for example, for all the pixels of the two-dimensional image corresponding to a certain length of the workpiece W as described above.

  Step S11: The defect detection unit 36 divides the entire acquired image into a plurality of (for example, two to three) inspection areas (dividing step). The inspection area can be divided mainly in the image sub-scanning direction (work W conveyance direction). An example of dividing the inspection area will be described later with reference to another drawing.

Step S12: The defect detection unit 36 calculates the mode value (representative value) D _freq of density from the pixels in the inspection area to be inspected this time (calculation step). The division of the inspection area and the calculation of the mode value D _freq will be described later with a specific example.

Step S14: Next, the defect detection unit 36 calculates the maximum density value D _fmax from the pixels in the inspection area which is the current inspection target. The calculation of the maximum value D _fmax will be described later with a specific example.

Step S16: The defect detection unit 36 extracts a pixel having a density value D lower than the mode value _Dfreq obtained in the previous step S12 for the pixels included in the current inspection area. Note that pixel extraction is performed, for example, by acquiring pixel coordinates (for example, an XY address) in the memory space.

  Step S18: The defect detection unit 36 combines adjacent pixels among the extracted pixels to form a combined pixel (AreaD. Note that the direction recognized as the adjacent pixels is, for example, the vertical direction (Y direction) ), The horizontal direction (X direction) and the diagonal direction, and the combined pixels formed at this time serve as a processing area (formation step), and are subject to gradation correction when the following conditions are satisfied.

  Step S20: The defect detection unit 36 confirms whether or not the area of the combined pixel satisfies a condition larger than the specified defect area. The prescribed defect area is assumed to be an area sufficiently larger than the area when a normal surface defect is imaged, and is experimentally determined based on the size of various defects that may exist in the workpiece W in advance, for example. I can keep it. Further, the comparison of the areas can be performed simply by comparing the number of pixels.

Step S22: When the formed combined pixel satisfies the above condition (Step S20: Yes), the defect detection unit 36 treats the combined pixel as the current processing area, and calculates the maximum density value D _max from the pixel in the processing area. calculate. The calculation of the maximum value _Dmax will be described later with a specific example.

Step S24: The defect detection unit 36 performs gradation correction on the density of the pixels in the processing area (correction process). At this time, the defect detection unit 36 applies the ratio (= D _freq / D _max ) between the mode value D _freq and the maximum value D _max as the gradation correction coefficient. Details of the correction will be described later with a specific example.

  When the above procedure is executed or when it is determined that the condition is not satisfied in the previous step S20 (step S20: No), the defect detection unit 36 executes step S26.

  Step S26: The defect detection unit 36 confirms whether or not there are remaining combined pixels in the inspection area. As a result, if other combined pixels remain (Yes), the defect detection unit 36 returns to step S16 and repeats the procedure from step S16 to step S24. At this time, if the combined pixel does not satisfy the condition of step S20, the defect detection unit 36 executes step S26 without executing steps S22 and S24.

  When the defect detection unit 36 performs the above procedure for all the combined pixels in the inspection area, the gradation correction (steps S22 and S24) for all the processing areas is completed. Then, when the defect detection unit 36 confirms that there is no remaining combined pixel (step S26: No), the defect detection unit 36 proceeds to the next step S28.

Step S28: The defect detection unit 36 performs setting to apply the correction coefficient to the defect threshold in the processing area. At this time, the defect detection unit 36 applies the ratio (= D _freq / D _all ) between the mode value D _freq and the mode value D _{all of the} entire image as a threshold correction coefficient. Note that when there is no processing area in the inspection area (when the area of all coupled pixels is equal to or less than the specified defect area), the setting for applying the threshold correction coefficient is not performed. The application of the correction coefficient will be described later with a specific example.

  Step S30: The defect detection unit 36 executes a defined defect detection process. In this defect detection process, for example, the density of a pixel included in the inspection area is compared with a predetermined defect threshold value, and a pixel whose density is less than the threshold value is determined as a defective portion (determination step). At this time, the pixels in the processing area AD are determined using the density after gradation correction, and the other pixels are determined using the original density (original density). Here, as described above, the correction coefficient is applied to the defect threshold for the pixels in the processing area.

  Step S32: When returning from the defect detection process, the defect detection unit 36 confirms whether or not to continue the inspection for the current image. For example, if there is an uninspected inspection area among the divided inspection areas, the defect detection unit 36 determines that the inspection should be continued (Yes), and then executes step S34.

Step S34: The defect detection unit 36 changes the inspection area to be inspected next.
Thereafter, the defect detection unit 36 returns to Step S12 and repeats the procedure from Step S12 to Step S30.

  When the inspection is completed for all inspection areas in the acquired image, the defect detection unit 36 determines that it is not necessary to continue the inspection (step S32: No), and ends the defect inspection processing.

[Example of inspection area division]
FIG. 6 is a conceptual diagram showing an example of processing for dividing the image data into a plurality of inspection areas in step S11.

  In FIG. 6, (A): When the defect detection unit 36 acquires image data as described above (step S10), the image is transferred in the transport direction (sub-scanning direction of the image) due to waviness or the like generated when the work W is transported. A large density difference may occur. At this time, the acquired image data includes a total of n line images in the main scanning direction.

  In FIG. 6, (B): The defect detection unit 36 for each line image has a density average value Ave (1), Ave (2), Ave (3), Ave (4),. (N) is calculated.

  In FIG. 6, (C): Next, the defect detection unit 36 determines the differences Deff (1), Deff (2), Deff (3), Deff (4) for the density average values Ave (1) to Ave (n) for each line. , ..., Deff (n-1) is calculated. For example, the difference Deff (1) is a difference between the density average values Ave (1) and Ave (2).

  In FIG. 6, (D): When the calculated difference is plotted on the memory space, for example, a difference peak (two peaks in this example) appears at the boundary position of the image where the density (brightness and darkness) changes greatly in the transport direction. . In the figure, the vertical axis represents the transport direction, and the horizontal axis represents the difference value.

  6E: Therefore, the defect detection unit 36 can divide the image into three inspection areas (for example, inspection areas 1 to 3) with the position where each difference peak appears as a boundary.

  In particular, when the difference peak as described above does not appear in the acquired image, the defect detection unit 36 can execute the processing from step S12 onward using the entire image as one inspection area. In this case, the defect detection unit 36 ends the inspection without changing the inspection area (step S34).

[Specific examples of defect inspection]
FIG. 7 is a diagram showing the image data acquired in the defect inspection process and the density distribution for each pixel in a simplified manner. In FIG. 7, for convenience, the number of pixels in the image data is adjusted to be smaller than the actual number. Hereinafter, the defect inspection process will be described with a specific example.

[Division of inspection area]
In FIG. 7, (A): As described above, when the defect detection unit 36 acquires image data (step S10), the entire image is divided into, for example, two inspection areas A1 and A2 (step S11). Here, for convenience, it is assumed that the image is divided into two inspection areas at the center position in the transport direction (sub-scanning direction), for example.

[Calculation of mode (representative value) and maximum value]
Assuming that the current target is the examination area A1, the mode value D _freq and the maximum value D _fmax are obtained as the following values (step S12, step S14). Here, the larger the value, the higher the density.
Mode D _freq : 150
Maximum value D _fmax : 150

[Formation of treatment area]
In the inspection area A1, pixels having a density lower than the mode D _freq are extracted, and adjacent pixels are connected to form a combined pixel (processing area AD) shown in FIG. 7 (step S18). . Note that even if the pixel PS1 has a lower density than the mode D _{freq, the} pixels distributed independently are not combined and are not formed as the processing area AD. Here, for convenience, it is assumed that the processing area AD is larger than the specified defect area.

Then, the maximum density value D _max in the processing area AD is obtained as follows (step S22).
Maximum value D _max : 100

[Tone correction for processing area]
FIG. 7B shows the density distribution when gradation correction (step S24) is performed by applying the gradation correction coefficient D _fmax / D _max (= _150/100 ) for the processing area AD. By this gradation correction, the pixel density in the processing area AD is increased as a whole (100 → 150, 20 → 30). As a result, even when a dark area is generated in the image due to slackness of the workpiece W or the like, gradation correction can be performed using only this dark area as a processing area. Note that no correction is performed for the single pixel PS1.

[Application of threshold correction coefficient]
The entire image of the mode _{D all} and threshold correction coefficient _D freq _{/ D all} is determined as follows (step S28).
Mode D _all : 200
Threshold correction coefficient D _freq / D _all : 0.75 (= _150/200 )

[Defect detection processing]
In the defect detection process (step S30), a standard defect threshold is applied to pixels other than the process area AD in the inspection area A1 to determine the presence or absence of a defect. On the other hand, regarding the pixels in the processing area AD, the presence / absence of a defect is determined by multiplying the standard defect threshold by the threshold correction coefficient D _freq / D _all (determination step). As a result, even if the defective portion has been tone-corrected in the processing area AD, the corrected defect threshold is applied to the defective portion, so that the defective portion is not missed.

When the inspection of the inspection area A1 in the image shown in FIG. 7 is finished, the mode is set as the remaining inspection area A2 (step S34), and the mode value _Dfreq and the maximum value _Dfmax are calculated in the same manner, but the inspection area A2 There is only a low-density pixel PS2 alone, and there is no pixel to be combined (step S20: No and step S26: No). In this case, since the processing area is not formed, without particularly applying a threshold correction coefficient D _{freq /} D _all the inspection area A2, the defect detection process is performed by applying a standard defect threshold to all the pixels.

  As described above, according to the defect inspection method performed using the defect inspection apparatus 10 of the present embodiment, even when a dark region having an irregular shape exists in an image, only such region is subjected to gradation correction. Therefore, the entire dark region is not erroneously detected as a defective portion, and only the original defective portion can be detected with high accuracy. In addition, even when a defective portion exists in the dark region (processing area), the threshold value correction coefficient is applied to the dark region (processing area), and thus the tone-corrected defective portion is detected. There is no leakage.

[Other Embodiments]
The present invention can be implemented with various modifications without being limited to the above-described embodiment. In one embodiment, the entire image is divided into two to three inspection areas in the conveyance direction (sub-scanning direction), but other division patterns may be adopted. For example, when the entire area of the image is large, the inspection area may be divided not only in the transport direction but also in the main scanning direction.

  Further, in the embodiment, the defect inspection in the case where the waviness and the slack in the conveyance direction of the workpiece W occur in combination is taken as an example. When there is no middle (A) dark region R3), when there is a difference in density due to slack only (for example, (B) in FIG. 4), or there is no deformation in the workpiece W itself, and only the difference in density due to the pattern on the surface. Even when it occurs (for example, (C) in FIG. 4), the defect inspection method of the present embodiment is applied to perform gradation correction of the dark region (processing area), and only the original defect portion is detected with high accuracy. Is possible.

  Further, the density given together with the illustration is merely an example, and the specific value varies depending on the object to be inspected and the imaging environment. Therefore, an appropriate threshold value may be set each time to detect a defect.

  In one embodiment, the workpiece W is conveyed and imaged in the sub-scanning direction. However, the workpiece W is fixed, and on the contrary, the main scanning line L is imaged while moving in the sub-scanning direction with respect to the workpiece W. Also good. Even in this case, the influence of the density change caused by the deformation of the workpiece W can be corrected. In one embodiment, grayscale image data is handled, but the image data may be processed as color.

  In addition, the arrangement of various devices described in the embodiment is a preferable example, and it is needless to say that the arrangement of the devices can be appropriately changed when the present invention is implemented.

It is the figure which showed schematically the example of a structure of the defect inspection apparatus. It is the perspective view which showed the process in which a workpiece | work is imaged with a line sensor. It is a block diagram which shows the structure of an arithmetic processing part in detail. FIG. 5 is a perspective view illustrating a mode in which a change in image density occurs due to various factors during an image forming process. It is the flowchart which showed the example of the procedure of the defect inspection process. It is the conceptual diagram which showed the example of the process which divides | segments image data into a some test | inspection area. It is the figure which simplified and represented distribution of image data and the density for every pixel.

Explanation of symbols

DESCRIPTION OF SYMBOLS 10 Defect inspection apparatus 12 Belt conveyor 14 Rotary encoder 18 Line sensor 20 Lens 22 Lighting fixture 24 Arithmetic processing part 30 Input part 32 Data processing part 34 Data storage part 36 Defect detection part 38 Output part 40 CPU
A1, A2 Inspection area (inspection area)
AD processing area (processing area)

Claims (5)

An acquisition step of acquiring an image of the inspected object;
A calculation step of obtaining a representative value of density from the pixels constituting the acquired image;
Forming a processing region by extracting pixels having a density lower than the representative value from the image and combining adjacent pixels;
If the formed processing region is larger than the prescribed area, a correction step for correcting the gradation of the density of all the pixels included therein,
A defect inspection method comprising: a determination step of determining presence / absence of a defect based on a density after correction for the processing region of the image, and determining a presence / absence of a defect based on an original density for other regions. The defect inspection method according to claim 1,
In the correction step,
A defect inspection method, wherein gradation correction is performed using a ratio between a representative value obtained in the calculation step and a maximum value of density of pixels included in the processing region as a correction coefficient. In the defect inspection method according to claim 1 or 2,
Further comprising a dividing step of dividing the image acquired in the acquiring step into a plurality of inspection regions;
In the calculation step,
Among the acquired images, a representative value is obtained from pixels constituting any one of the inspection regions divided into a plurality in the dividing step,
In the forming step,
Combine the pixels extracted from the inspection area to form the processing area,
In the determination step,
The presence or absence of a defect is determined by comparing a predetermined threshold value with the density of each pixel. At that time, the pixels in the processing area are acquired by the representative value in the inspection area and the acquisition step. A defect inspection method, wherein a threshold value is corrected using a ratio with a representative value of density obtained from all pixels constituting an image as a correction coefficient. The defect inspection method according to claim 3,
A defect inspection method comprising repeatedly performing the calculation step to the determination step for each of the inspection regions divided in the division step. Image generating means for capturing an image of the inspection object and generating an image;
Calculating means for obtaining a representative value of density from an image constituting the image generated by the image generating means;
Forming means for extracting pixels having a density lower than the representative value from the image and combining adjacent pixels to form a processing region;
When the processing region is larger than the specified area, a correction unit that performs gradation correction on the density of all pixels included in the processing region;
A defect inspection apparatus comprising: a determination unit configured to determine presence / absence of a defect based on a density after correction for the processing region of the image, and to determine presence / absence of a defect based on an original density for other regions.

Images