JP6287248B2 - Appearance inspection apparatus, appearance inspection method, and program - Google Patents

Description

  The present invention relates to an appearance inspection apparatus, and more particularly to an appearance inspection apparatus capable of improving inspection sensitivity and improving the efficiency of the visual inspection process by feeding back the determination contents in the visual inspection process.

  Conventionally, an appearance inspection apparatus for inspecting the appearance of an object to be inspected has been proposed and used in a manufacturing process (for example, Patent Document 1).

  Patent Document 1 discloses an image input unit that captures the surface of a non-inspection object, a feature extraction unit that extracts a plurality of feature amounts from the captured image, and a plurality of extracted feature amounts as explanatory variables by multivariate analysis. A storage means for storing a function for performing group prediction and a function stored in the storage means are selectively used to perform group prediction on the type of defect of the object to be inspected. Disclosed is an image determination apparatus having a group prediction unit that performs group prediction of size, and a determination unit that determines pass / fail of a defect and reinspection based on a combination of a defect type and a defect size obtained by the group prediction unit. ing.

JP 08-68764 A

  However, in the method of Patent Document 1, the accuracy of prediction is a function for classifying defect types and a function for performing group prediction of defect sizes corresponding to the results of group prediction of defect types. Depending on how the function is determined, there is a problem that the classification accuracy is lowered, and there is a possibility that erroneous detection of a defect and oversight may occur frequently.

  The present invention has been made in view of the above-described problems, and the object of the present invention is to improve defect classification accuracy and pass / fail judgment accuracy with the use of the apparatus, and to make the visual inspection process more efficient. It is to provide a possible appearance inspection device and the like.

In order to achieve the above-described object, the first invention is an appearance inspection apparatus for determining the type and pass / fail of a defect occurring in the appearance of an object to be inspected, and an image input means for inputting a photographed image of the object to be inspected. A defect candidate extraction unit that extracts a defect candidate from the captured image, a feature amount calculation unit that calculates a plurality of feature amounts of the defect candidate, and a type of the defect for each type of the defect First storage means for storing a first threshold value relating to the plurality of feature amounts, and a second threshold value relating to the plurality of feature amounts for determining pass / fail of the defect candidate for each type of defect. second storage means, a plurality of feature amounts of the defect candidate is compared with the first threshold value, and classifying means for identifying the type of the defect of the defect candidate, the type of the defect is specified by the classification means Good defect candidate And a determining quality determination means by comparing the plurality of features and the second threshold value of the defect candidate, when the type of the defect of the defect candidate can not be identified by said classifying means, the defect candidate The defect candidate when the defect candidate cannot be determined by the type input means for displaying the position information of the defect candidate and the defect type of the defect candidate is selected and input by the quality determination means , and when the defect candidate is determined to be defective And the first threshold value stored in the first storage means for the defect type selected and input by the type input means. and a classification threshold value updating means for updating based on a plurality of feature quantities of the defect candidate, said about the kind of the defect of the defect candidate entered the quality by the quality input means The second threshold value stored in the second storage means is the appearance inspection apparatus characterized by comprising: a quality determination threshold updating means for updating based on a plurality of feature quantities of the defect candidate .

The first threshold value is an upper limit value and a lower limit value determined for each of the plurality of feature amounts for each type of defect, and the classification unit determines that all feature amounts of the defect candidates are the first feature amount. In the range of the upper limit value and the lower limit value of the threshold value, it is desirable to specify that the defect candidate belongs to the defect type of the first threshold value.
In addition, the classification threshold update unit sets the feature amount of the defect candidate as the upper limit value of the first threshold when the feature amount of the defect candidate exceeds the upper limit value of the first threshold, When the feature amount of the defect candidate is lower than the lower limit value of the first threshold value, it is desirable to update the first threshold value using the feature amount of the defect candidate as the lower limit value of the first threshold value.

The second threshold value is an upper limit value and a lower limit value determined for each of the plurality of feature amounts for each type of the defect, and the pass / fail determination unit determines that the total feature amount of the defect candidate is the second feature amount. The defect candidate is determined to be defective when it is in the range of the upper limit value and the lower limit value of the threshold value, and the entire feature amount of the defect candidate is within the range of the upper limit value and the lower limit value of the second threshold value. If not, it is desirable to specify that the defect candidate is not defective.
When the feature value of the defect candidate exceeds the upper limit value of the second threshold value, the pass / fail judgment threshold update unit sets the feature value of the defect candidate as the upper limit value of the second threshold value, and the defect When the candidate feature amount falls below the lower limit value of the second threshold value, it is desirable to update the second threshold value using the feature amount of the defect candidate as the lower limit value of the second threshold value.

The feature amount includes the area of the defect candidate, ferret diameter, circumscribed rectangular main axis / sub axis length, main axis angle, circularity, acicularity, circumscribed circle / inscribed circle diameter, circumferential length, equivalent ellipse main axis / sub axis length It is desirable to include an equivalent ellipse major axis minor axis length ratio, an average luminance value, a maximum luminance value, and a brightness value.
In addition, it is desirable that the type of the defect includes dirt, a stain, a scratch, and a fiber.

The object to be inspected is, for example, a sheet-like product in the production line, and a photographed image of the object to be inspected is input by the image input means, and defect candidates are extracted from the photographed image by the defect candidate extracting means. To do. Next, the feature amount calculation means determines the ferret diameter, circumscribed rectangular main axis / sub axis length, main axis angle, circularity, acicularity, circumscribed circle / inscribed circle diameter, circumference, equivalent ellipse main axis / sub axis for defect candidates. Extracting feature quantities such as axial length, equivalent ellipse main axis / sub axis length ratio, average luminance value, maximum luminance value, brightness, etc., and comparing them with the first threshold value stored in the first storage means by the classification means To determine the type of defect.
The first threshold value is an upper limit value and a lower limit value of each feature amount provided for each type of defect. When all feature amounts of defect candidates are within the range of the first threshold value, The type of threshold value is the type of defect.

If the classification means does not determine the feature quantity of the defect candidate for any type of defect, the type input means displays the position of the defect candidate and allows the operator to visually observe the defect candidate of the object to be inspected. Enter the type of. The classification threshold value updating means updates the first threshold value for the inputted visual defect type based on the feature amount of the defect candidate. The first threshold is updated with the feature amount of the defect candidate not within the first threshold as the new first threshold.
As a result, by continuing to use the appearance inspection apparatus of the first invention, the first threshold value is updated, the defect classification accuracy can be improved, and the efficiency of the inspection work can be improved.

Further, when the defect type of the defect candidate is determined by the classification unit, the pass / fail determination unit determines the pass / fail of the defect candidate by comparing with the second threshold value stored in the second storage unit. .
The second threshold value is an upper limit value and a lower limit value of each feature amount provided for each type of defect. When all the feature amounts of the defect candidates are within the second threshold range, it is determined that the defect is defective. If all the candidate feature amounts are not within the range of the upper limit value and the lower limit value of the second threshold, it is determined that the feature is not defective.

If the defect candidate cannot be determined to be defective or not defective, the defect input unit displays the position of the defect candidate, allows the operator to visually check the defect candidate of the object to be inspected, and inputs the defect candidate. Let The pass / fail judgment threshold updating means updates the second threshold of the defect type for the defect candidate determined to be defective based on the feature amount of the defect candidate. That is, the second threshold value is updated with the feature amount of the defect candidate not within the second threshold value as the new second threshold value.
Accordingly, by continuing to use the appearance inspection apparatus according to the first aspect of the invention, the second threshold value is updated, and it is possible to improve the accuracy of defect determination and increase the efficiency of inspection work.

According to a second aspect of the present invention, there is provided a first storage unit that stores a first threshold value related to a plurality of feature amounts for specifying the defect type for each defect type, and a defect candidate for each defect type. Determining the type and pass / fail of a defect generated in the appearance of the object to be inspected, which is executed by an appearance inspection apparatus that includes a second storage unit that stores a second threshold value related to the plurality of feature amounts for determining pass / fail. An appearance inspection method, an image input step for inputting a captured image of an object to be inspected, a defect candidate extraction step for extracting a defect candidate from the captured image, and a feature amount calculation for calculating a plurality of feature amounts of the defect candidate steps and, a plurality of feature quantities of the defect candidate is compared with the first threshold value, a classification step of identifying a type of the defect of the defect candidate, the defect type of the defect is identified by the classification step The quality of the complement, and quality determination step of determining by comparing the plurality of features and the second threshold value of the defect candidate, when the type of the defect of the defect candidate can not be identified by the classification step, When the position information of the defect candidate is displayed, the type input step for selectively inputting the type of the defect of the defect candidate, and the case where the quality of the defect candidate cannot be determined by the quality determination step and when it is determined as defective, The defect input step for displaying the position information of the defect candidate and inputting the pass / fail of the defect candidate, and the first threshold value for the defect type selected and input by the type input step are set as a plurality of defect candidates. a classification threshold value updating step of updating based on the feature quantity of the type of the defect of the defect candidate entered the quality by the quality input step Nitsu The second threshold value Te, is an external inspection method characterized by comprising: a threshold updating step for judging acceptability of updating based on a plurality of feature quantities of the defect candidate.

A third invention is a program for causing a computer to function as an appearance inspection apparatus for determining the type of defect occurring in the appearance of an object to be inspected and the acceptance / rejection of the image. The computer inputs a photographed image of the object to be inspected. Image input means, defect candidate extraction means for extracting defect candidates from the captured image, feature quantity calculation means for calculating a plurality of feature quantities of the defect candidates, and for specifying the defect type for each defect type First storage means for storing a first threshold value relating to the plurality of feature amounts, and storing a second threshold value relating to the plurality of feature amounts for determining pass / fail of the defect candidate for each type of defect. second storage means, comparison with the plurality of feature quantity the first threshold value of the defect candidate, classifying means for identifying the type of the defect of the defect candidate, the type of the defect by the classification means The quality of the constant has been the defect candidate, quality determination unit for determining by comparing the plurality of features and the second threshold value of the defect candidate, the type of the defect of the defect candidate is specified by the classification means If it is not possible, the position information of the defect candidate is displayed, the type input means for selectively inputting the defect type of the defect candidate, and the case where the quality of the defect candidate cannot be judged by the quality judgment means and the case where the defect candidate is judged as defective In this case, the defect information selected by the quality input means for displaying the position information of the defect candidates and inputting the quality of the defect candidates and the type of the defect selected and input by the type input means are stored in the first storage means. the defect of the first threshold value, the classification threshold value updating means for updating based on a plurality of feature quantities of the defect candidate, the defect candidate entered the quality by the quality input means The second threshold value stored in the second storage unit about the type, a program to function as, quality determination threshold updating means for updating based on a plurality of feature quantities of the defect candidate.

  According to the present invention, there is provided an appearance inspection apparatus and the like capable of improving defect classification accuracy and pass / fail determination accuracy with the use of the apparatus and improving the efficiency of the visual inspection process.

The figure which shows the outline | summary of the external appearance inspection apparatus 1 which concerns on embodiment of this invention. Block diagram showing a hardware configuration example of the appearance inspection apparatus 1 The block diagram which shows the function structural example of the external appearance inspection apparatus 1 Illustration of examples of defect feature values The figure explaining defect candidate feature-value 41 The figure explaining the threshold value for classification 43 and the threshold value for determination 45 The flowchart which shows the flow of a process of the defect classification process part 25. The flowchart which shows the flow of a process of the quality determination process part 26. The flowchart which shows the flow of a process of the visual determination input part 27. The figure which shows the example of the visual determination input screen 50 The flowchart which shows the flow of a process of the threshold value update process part 28 for classification | category. The flowchart which shows the flow of a process of the threshold value update process part 29 for a determination.

  Hereinafter, preferred embodiments of an appearance inspection apparatus 1 and the like according to the present invention will be described in detail with reference to the accompanying drawings. In the following description and the accompanying drawings, the same reference numerals are given to components having substantially the same functional configuration, and redundant description will be omitted.

FIG. 1 is a diagram showing an outline of the appearance inspection apparatus 1.
A camera 5 is connected to the appearance inspection apparatus 1, the appearance of the inspection sheet 3 is photographed by the camera 5, a photographed image is taken into the appearance inspection apparatus 1, and is processed by the appearance inspection apparatus 1 of the present embodiment. The position and type of the defect 31 on the inspection sheet 3 and the quality of the defect are output.

  The sheet to be inspected 3 is, for example, a roll-shaped film sheet, and the appearance inspection apparatus 1 has a defect (for example, contamination, stains, scratches, fibers mixed) generated in the production line on the sheet to be inspected 3. Is detected, and the type of defect and the quality of the defect are determined.

FIG. 2 is a block diagram illustrating a hardware configuration example of the appearance inspection apparatus 1.
The appearance inspection apparatus 1 can be configured by a computer device such as a personal computer, for example.

  The appearance inspection apparatus 1 includes, for example, a control unit 11, a storage unit 12, a media input / output unit 13, a communication control unit 14, an input unit 15, a display unit 16, a peripheral device I / F (interface) unit 17, and the like through a bus 18. It is a connected configuration.

  The control unit 11 includes a CPU (Central Processing Unit), a ROM (Read Only Memory), a RAM (Random Access Memory), and the like.

The CPU calls and executes a program stored in the ROM, the storage unit 12 and the like in a work memory area on the RAM, drives and controls each device connected via the bus 18, and realizes processing performed by the computer.
The ROM is a non-volatile memory and permanently holds a computer boot program, a program such as BIOS, data, and the like.
The RAM is a volatile memory, and temporarily holds a program, data, and the like loaded from the storage unit 12, ROM, storage medium, and the like, and includes a work area used by the control unit 11 for performing various processes.

The storage unit 12 stores a program executed by the control unit 11, data necessary for program execution, an OS (Operating System), and the like. The program is a processing program or the like of the appearance inspection apparatus 1 according to the embodiment of the present invention, and includes an image processing program, a defect candidate feature amount calculation program, a defect classification processing program for determining the type of defect candidate, and defect candidate quality. And a classification threshold value used in the defect classification process and the quality determination process, and a determination threshold update process program.
These programs will be described later.
The program code is read by the control unit 11 as necessary, transferred to the RAM, and executed by the CPU as various means.

The media input / output unit 13 (drive device) inputs / outputs data, for example, media such as a CD drive (-ROM, -R, -RW, etc.), DVD drive (-ROM, -R, -RW, etc.) Has input / output devices.
The communication control unit 14 includes a communication control device, a communication port, and the like, and performs communication control with other devices via a network. The network may be wired or wireless.

The input unit 15 inputs data and includes, for example, a keyboard, a pointing device such as a mouse, and an input device such as a numeric keypad. An operator of the appearance inspection apparatus 1 inputs the type of defect visually and information on pass / fail from the input unit 15.
The display unit 16 is a display device such as a CRT monitor or a liquid crystal panel. The display unit 16 displays captured images of defect candidates, selection inputs for visual input, and the like.

The peripheral device I / F (interface) unit 17 is a port for connecting peripheral devices, and is configured by USB, IEEE1394, RS-232C, or the like, and the connection form may be wired or wireless. The captured image captured by the camera 5 is input to the appearance inspection apparatus 1 via the peripheral device I / F unit 17 and stored in the storage unit 12.
The bus 18 is a path that mediates transmission / reception of control signals, data signals, and the like between the devices.

FIG. 3 is a block diagram illustrating a functional configuration example of the appearance inspection apparatus 1 according to the present embodiment.
The appearance inspection apparatus 1 includes an image input unit 21, a preprocessing unit 22, a defect candidate extraction unit 23, a feature amount extraction unit 24, a defect classification processing unit 25, a pass / fail determination processing unit 26, a visual determination input unit 27, and a classification threshold update. The processing unit 28, the determination threshold update processing unit 29, and the like, and a defect candidate feature database (DB) 41, a classification threshold database 43, and a database storing the determination threshold database 45 are configured.

  The image input unit 21 captures the captured image of the inspection sheet 3 captured by the camera 5 into the appearance inspection apparatus 1. The captured image is stored in the storage unit 12, and each process described below is executed.

  The preprocessing unit 22 performs shading correction processing, filtering processing such as differentiation, contrast enhancement, and smoothing, and binarization processing on the captured image in order to facilitate the extraction of the defective portion.

  The defect candidate extraction unit 23 executes a process of extracting a defect candidate part from the image after the binarization processing by the preprocessing unit 22. A defect candidate region is extracted by providing a threshold for the gradation of the image and extracting a portion higher than the threshold. Since the defect candidate extraction processing is also performed in the conventional appearance inspection apparatus 1 (see, for example, Patent Document 1), detailed description thereof is omitted here.

The feature amount calculation unit 24 calculates a feature amount for the defect candidate extracted by the defect candidate extraction unit 23 and stores the feature amount in the defect candidate feature amount database 41.
The feature amount is, for example, area, ferret diameter (X direction and Y direction), circumscribed rectangular main axis length, circumscribed rectangular minor axis length, main axis angle, circularity, acicularity, circumscribed circle diameter, inscribed circle diameter, circumference Equivalent ellipse principal axis length, equivalent ellipse minor axis length, equivalent ellipse principal axis minor axis ratio, average luminance value, maximum luminance value, brightness and darkness, and the like. These feature amounts can be calculated by a general image processing program.

  The defect classification processing unit 25 compares the feature amount extracted by the feature amount calculation unit 24 and stored in the defect candidate feature amount database 41 with the classification threshold value stored in the classification threshold value database 43, thereby obtaining a defect candidate. Execute the process to specify the type. Details will be described later.

  When the type of defect candidate is specified by the defect classification processing unit 25, the defect determination processing unit 26 then determines whether the defect candidate part is minor and has no problem (OK), defective (NG), or Processing for determining whether or not determination is possible (unknown) is executed. The pass / fail determination processing unit 26 determines pass / fail by comparing the feature amount of the defect candidate with a determination threshold for the type of defect candidate stored in the determination threshold database 45. Details will be described later.

  When the defect classification processing unit 25 cannot identify the type of the defect candidate (unknown), when the non-defective product determination processing unit 26 determines that the defect candidate is NG, and when the quality is not determined (unknown) In this case, the visual judgment input unit 27 causes the worker to visually recognize the defect candidate, and executes a process of inputting the type and quality of the defect candidate from the visual result. Details will be described later.

The classification threshold update processing unit 28 executes processing for updating the classification threshold value of the type of defect candidate visually input by the visual determination input unit 27 with the value of the feature amount of the defect candidate visually observed. . Details will be described later.
The determination threshold update processing unit 29 is a feature of the defect candidate obtained by visually determining the determination threshold for the type of the defect candidate based on the determination result of the defect candidate by the visual input by the visual determination input unit 27. The process of updating with the value of the quantity is executed. Details will be described later.

  As described above, by continuously updating the classification threshold and the determination threshold by the classification threshold update processing unit 28 and the determination threshold update processing unit 29, the classification accuracy of the defect classification processing unit 25, and It is possible to improve the pass / fail judgment accuracy of the pass / fail judgment processing unit 26.

  In the initial stage of operation, since there is no threshold for classification and threshold for determination, or it is created from the types and quality determination results based on the visual results of a small number of defect candidates, the defect classification processing unit 25 or the quality determination processing unit 26 There are a large number of defect candidates whose types are unknown or which cannot be determined to be acceptable (unknown), and for these defect candidates, the classification threshold update processing unit 28 and the determination threshold update process based on the visual type and the quality determination result By updating the threshold for classification and the threshold for determination by the unit 29, it is possible to improve the classification accuracy and the pass / fail determination accuracy.

FIG. 4 shows an example of the types of defect candidate feature amounts calculated by the feature amount calculation unit 24 and the trend of feature amounts depending on the types of defects.
As described above, the feature amount includes area, ferret diameter (X direction and Y direction), circumscribed rectangular main axis length, circumscribed rectangular minor axis length, main axis angle, circularity, acicularity, circumscribed circle diameter, inscribed circle diameter. , Circumference, equivalent ellipse principal axis length, equivalent ellipse minor axis length, equivalent ellipse principal axis minor axis ratio, average luminance value, maximum luminance value, brightness and the like.

The types of defects include “dirt”, “stain”, “scratch”, “fiber” and the like.
“Dirty” occurs, for example, when liquid or the like falls on the production line and adheres to it. In many cases, the shape is “round”, so the “circularity” of the feature amount is close to 1 (the perfect circle is 1). In addition, the acicularity (ratio between the main axis length and the sub axis length) also tends to take a value close to 1 (square, perfect circle is 1).

In addition, “stain” often has a light color, and the “average luminance value” and the “maximum luminance value” of the feature amount tend to take a low value.
In addition, “scratches” are often vertically long and linear, and the “circularity” value of the feature amount tends to be low and the “needleness” value tends to be large. Further, under the production conditions that occur along the conveyance direction of the inspection sheet, the “spindle angle” tends to be 90 degrees.
“Fiber” is a defect that occurs when fibers, hair, or the like adheres to the sheet to be inspected. The value of “circularity” is low, and the values of “needleness” and “perimeter” tend to be large. There is.

  The type of defect is not limited to this, and several types such as “dirt 1”, “dirt 2”,... Good.

FIG. 5 is a diagram illustrating an example of defect candidate feature amount data stored in the defect candidate feature amount database 41.
As shown in FIG. 5A, the defect candidate feature quantity database 41 includes a defect number for identifying each defect candidate. 411, position information 412 of the defect candidate on the inspection sheet, and the values of the feature quantities 413 shown in FIG. 4 are stored.

As shown in FIG. 5B, the N feature amounts 413P (i) of the i-th defect candidate are set to (p _i1 , p _i2 ,... P _iN ).

  FIG. 6 is a diagram illustrating an example of each feature amount stored in the classification threshold value database 43 and the determination threshold value database 45. FIG. 6A shows a classification threshold, and FIG. 6B shows a determination threshold.

The classification threshold is a threshold for specifying the type of defect candidate, and includes a lower limit value and an upper limit value for each feature quantity 413.
The determination threshold is a threshold for determining whether or not the defect candidate is a defect that needs to be corrected when the type of defect candidate is specified. The amount 413 includes a lower limit value and an upper limit value.

  Hereinafter, each process of the defect classification process 25, the pass / fail determination process 26, the visual determination input process 27, the classification threshold update process 28, and the determination threshold update process 29 will be described in detail.

FIG. 7 is a flowchart showing a process flow of the defect classification process 25.
First, the control unit 11 of the appearance inspection apparatus 1 reads the defect candidate feature amount P stored in the defect candidate feature amount database 41 (step 101).
The feature amount P is N numerical values p ₁ , p ₂ ,... P _N.

  Next, the control unit 11 substitutes 1 for the defect classification number m (step 102) and 1 for the feature number n (step 103). The defect classification number m indicates a defect type number, and there are M types. The feature quantity number n indicates the feature quantity number, and there are N feature quantities.

Next, the control unit 11 determines whether the feature value p _n is in the range of lower-upper limit of the classification threshold 43 of the n-th feature value of defect classification number m (Step 104).
When it is in the range of the lower limit / upper limit value (Yes in Step 104), the control unit 11 increments the feature quantity number n by 1 (Step 105), and when the feature quantity number n is N or less (Yes in Step 106). Returning to step 104, similarly, a process for determining whether or not the next feature quantity number n is in the range of the classification threshold value 43 is performed.

  When the feature quantity number n exceeds N in step 106 (No in step 106), all the feature quantities of the defect candidate feature quantity P are within the lower and upper limits of the classification threshold 43 of the defect classification number m. The control unit 11 identifies the defect candidate P as the defect classification m (step 107), and ends the process.

On the other hand, when the feature quantity _pn is not within the lower limit / upper limit range of the classification threshold value 43 of the nth feature quantity of the defect classification number m in step 104 (No in step 104), the defect candidate is the defect classification number m. The controller 11 increments the defect classification number m by 1 (step 111). If the defect classification number m is M or less (Yes in step 111), the control unit 11 returns to step 103. The feature amount number n is returned to 1, and processing for determining whether or not the feature amount P is within the lower limit / upper limit range of the classification threshold value 43 is performed for the next defect type (steps 104 to 107, step 111). 112).

  In step 112, when the defect classification number m exceeds M (No in step 112), it indicates that the feature amount P of the defect candidate is not in the range of the threshold value 43 for classification of any defect type, and the control unit 11 Sets the defect candidate P as unknown (step 113) and ends the process.

  As described above, when all the feature amounts P of the defect candidate are within the lower limit / upper limit value of the classification threshold 43 for a certain defect type by the defect classification process 25, the defect candidate is Identified as a type.

  In the initial stage of the operation of the appearance inspection apparatus 1, there is no classification threshold 43 or it is formed from a small amount of feature amount data, so the type of defect candidate is not specified, and classification may be unknown in step 113. Many can occur. As shown in FIG. 3, the visual determination input process 27, the classification threshold update process 28, and the determination threshold update process 29 are executed for the defect candidates whose classification is unknown in step 113.

On the other hand, if the type of defect candidate is identified in step 107, the control unit 11 then executes a pass / fail judgment process 26.
FIG. 8 is a flowchart showing the flow of the quality determination process 26.
In this case, the defect candidate is assumed to have the defect type identified by the defect classification process 25 as the defect classification number m.

First, the control unit 11 reads the feature amount P of the defect candidate identified as the defect classification number m from the defect candidate feature amount database 41 (step 201).
The feature number n is initialized to 1 and the determination variable K is initialized to 0 (step 202).

Next, the control unit 11 determines whether the feature value p _n is in the range of lower-upper limit of the determination threshold 45 of the n-th feature value of defect classification number m (Step 203).
When it is in the range of the lower limit / upper limit value (Yes in Step 203), the control unit 11 increments the determination variable K by 1 (Step 204). On the other hand, when not in the range of the lower limit / upper limit value (No in Step 203), the control unit 11 does not execute Step 204.

Next, the control unit 11 increments the feature number n by 1 (step 205).
Next, when the feature number n is N or less (Yes in Step 206), the control unit 11 returns to Step 203, and when it is larger than N (No in Step 206), the control unit 11 proceeds to the next.

  Next, the control unit 11 determines whether the determination variable K is equal to N (step 207). If they are equal (Yes in step 207), it is indicated that all the feature quantities of the defect candidate feature quantity P are within the lower limit / upper limit value of the determination threshold 45 of the defect classification number m. Determines that the defect candidate P is defective (NG) (step 208), and ends the process.

  On the other hand, when the determination variable K and N are not equal (No in Step 207), the control unit 11 determines whether the determination variable K is equal to 0 (Step 209). If they are equal (Yes in step 209), it indicates that all the feature quantities of the defect candidate feature quantity P are not within the lower limit / upper limit value of the determination threshold 45 of the defect classification number m, and the control unit 11 Determines that the defect candidate P is not defective (OK) (step 210), and ends the process.

On the other hand, if the determination variable K is not equal to 0 (No in step 209), that is, if 0 <K <N, the element p _i (1 ≦ i ≦ N) of the defect candidate feature amount P is The control unit 11 indicates that there are some that are not within the lower and upper limits of the determination threshold 45 of the defect classification number m, and the control unit 11 determines that the pass / fail determination of the defect candidate P is unknown (step 211). ), The process is terminated.

  If all the feature amounts P of the defect candidate with the defect classification number m are all within the lower limit / upper limit value of the determination threshold value 45 with the defect classification number m, the defect candidate is defective ( NG), when all the feature quantities P of the defect candidate of the defect classification number m are not within the lower limit / upper limit value of the determination threshold 45 of the defect classification number m, the defect candidate is not defective (OK), Otherwise, it is determined that the pass / fail determination of the defect candidate is unknown.

  By preparing the determination threshold value 45 used for the non-defective product determination process 26 separately from the classification threshold value 43, even if the defect candidate is classified as a defect in the defect classification process 25, a short-length scratch or a shading is light. In the case of a flaw, it is possible to determine OK in the pass / fail determination process 26. For example, the lower limit threshold value of the ferret diameter, which is a characteristic amount peculiar to a defect (NG), or the lower limit threshold value of the average density value related to the density of the scratch may be set to an appropriate value.

In the pass / fail judgment process 26, when the defect candidate is NG or the judgment is unknown, a visual judgment input process 27, a classification threshold update process 28, and a judgment threshold update process 29 are executed as shown in FIG. The
If it is determined that the defect candidate is not defective (OK), it is not subject to visual inspection by the operator. However, the feature quantity as a defect candidate is recorded in the database as a defect candidate feature quantity 41.

  When it is determined in the defect classification process 25 that the defect candidate is unknown, and in the pass / fail determination process 26, if the defect candidate is NG or unknown, the defect candidate is subject to visual inspection. Next, the control unit 11 executes a visual determination input process 27.

FIG. 9 is a flowchart showing a process flow of the visual determination input process 27.
When it is determined in the defect classification process 25 that the defect candidate is unknown, and in the pass / fail determination process 26, the control unit 11 displays a defect on the display unit 16 when the defect candidate is NG or the determination is unknown. A visual determination input screen 50 including a map 51, a defect image 53, a defect position 55, etc. is displayed (step 301).
Further, the control unit 11 displays the defect classification name option 57 and the visual determination option 59 on the display unit 16 (step 302).

FIG. 10 is a diagram illustrating an example of the visual determination input screen 50.
The visual determination input screen 50 includes, for example, a defect map 51, a defect image 53, a defect position 55, a defect classification name option 57, a visual determination option 59, and the like.

The defect map 51 indicates the positions of defect candidates on the inspection sheet 3.
In the defect map 51, some mark (for example, x mark), a number assigned to the defect, and the like are displayed at the position of the defect candidate.
In the defect image 53, one photographed image of one or more defect candidates shown on the defect map 51 is displayed together with the defect number N.
The defect position 55 is a position coordinate on the inspection sheet 3 of the defect candidate.

The operator can find a defect candidate on the inspected sheet 3 from the information of the defect map 51 and the defect position 55 on the visual determination input screen 50 and visually check the defect candidate. At this time, it is also possible to confirm the defect candidate by comparing with the captured image of the defect image 53.
The operator can determine the defect type of the defect candidate and whether it is defective or not by visually observing the defect candidate on the inspection sheet 3.

  The defect classification name option 57 is an option for allowing the operator to input the type of defect. The operator can input the defect classification name as a classification result by placing the cursor on the defect classification name (type) selected in the defect classification name option 57 with a mouse or the like and clicking.

  The visual determination option 59 is an option for allowing the operator to input whether the defect candidate is defective (NG) or not defective (OK). When the operator puts the cursor on OK or NG of the visual determination option 59 with a mouse or the like and clicks, OK (NG) / defective (NG) of the defect candidate can be input as the visual determination result.

Returning to FIG. 9, the control unit 11 causes the operator to input the defect classification name and the visual determination result (step 303).
When a defect type is selected from the defect classification name option 57 on the visual determination input screen 50, a classification threshold update process 28 is executed.

FIG. 11 is a flowchart showing the process flow of the classification threshold update process 28.
First, the control unit 11 reads the feature amount P of the target defect candidate visually input by the operator on the visual determination input screen 50 from the defect candidate feature amount database 41 (step 401).
Next, the feature quantity number n is set to an initial value 1 (step 402).
Next, the control unit 11 reads the classification threshold value 43 of the nth feature amount for the type (classification name) selected by the defect classification name option 57 (step 403).
_Let the lower limit value and the upper limit value of the read classification threshold value 43 be C _{n, min} and C _{n, max} , respectively.

Next, the feature value P of the defect candidate is compared with the classification threshold value 43, and when the feature value P of the defect candidate is not within the range of the classification threshold value 43, processing for updating the threshold value is performed (steps 404 to 412). ).
That is, when the feature quantity _pn is less than the lower limit value C _{n, min} (Yes in step 404), the feature quantity _pn is set to the lower limit value C _{n, min} of the classification threshold 43 (step 405), and the feature quantity p _{When n} exceeds the upper limit value C _{n, max} (Yes in step 411), the value of the feature quantity _pn is set as the upper limit value C _{n, max} of the classification threshold value 43 (step 412).
If step 411 is No, the feature value _pn is within the range between the lower limit value C _{n, min} and the upper limit value C _{n, max} of the classification threshold value 43, and therefore the classification threshold value 43 is not updated.

After the processing in step 405 and step 412, and in the case of No in step 411, the feature amount number n is incremented by 1, and if the value of n is less than or equal to the number N of feature amounts, steps 403, 404, 405, 406, If the value of the feature number n exceeds N, the processing ends.
With the above processing, the value of the classification threshold 43 is updated with the feature value of the defect candidate visually observed.

When the visual determination option 59 is selected as defective (NG) on the visual determination input screen 50, a determination threshold update process 29 is executed.
FIG. 12 is a flowchart showing the flow of the determination threshold update process 29.

First, the control unit 11 reads the feature amount P of the target defect candidate visually input by the worker on the visual determination input screen 50 from the defect candidate feature amount database 41 (step 501).
Next, the feature quantity number n is set to an initial value 1 (step 502).
Next, the control unit 11 reads the determination threshold value 45 of the nth feature amount for the type (classification name) selected by the defect classification name option 57 (step 503).
The lower limit value and the upper limit value of the read determination threshold value 45 are set to D _{n, min} and D _{n, max} , respectively.

Next, the feature value P of the defect candidate is compared with the determination threshold value 45, and when the feature value P of the defect candidate is not within the range of the determination threshold value 45, processing for updating the threshold value is performed (steps 504 to 512). ).
That is, when the feature quantity _pn is lower than the lower limit value D _{n, min} (Yes in step 504), the feature quantity _pn is set to the lower limit value D _{n, min} of the determination threshold 45 (step 505), and the feature quantity p _{When n} exceeds the upper limit value D _{n, max} (Yes in step 511), the value of the feature quantity _pn is set as the upper limit value D _{n, max} of the determination threshold 45 (step 512).
When step 511 is No, the feature value _pn is within the range between the lower limit value D _{n, min} and the upper limit value D _{n, max} of the determination threshold value 45, and therefore the determination threshold value 45 is not updated.

After the processing in step 505 and step 512 and in the case of No in step 511, if the feature quantity number n is incremented by 1 and the value of n is equal to or less than the number N of feature quantities, steps 503, 504, 505, 506, The processing of 511 and 512 is repeated, and when the value of the feature number n exceeds N, the processing is terminated.
Through the above processing, the value of the determination threshold 45 is updated with the feature value of the defect candidate visually observed.

As described above, the visual determination input process 27, the classification threshold update process 28, and the determination threshold update process 29 are executed, and the classification threshold 43 and the determination threshold 45 are updated and fed back to the appearance inspection apparatus 1. From the next time, it becomes possible to automate the type determination and pass / fail determination for the same defect candidates.
Further, by repeating the update of the classification threshold 43 and the determination threshold 45, it is possible to improve the classification accuracy and the pass / fail determination accuracy on a daily basis, and to improve the efficiency of the appearance inspection work.

In the classification threshold update process 28 (FIG. 11) and the determination threshold update process 29 described above, it has been described that when the threshold value is updated, the value of the feature quantity p is used as the update value, but extreme updating is avoided. Therefore, a correction coefficient k (0 <k <1) may be provided.
For example, when the feature amount is p, the lower limit value of the threshold value is q1, and the lower limit value q1 is updated to obtain q2, the updated lower limit value q2 = q1- (q1-p) · k is updated. Thus, it is possible to suppress extreme updating.

  Further, in the operation of the appearance inspection apparatus 1 according to the present embodiment, the classification accuracy and the pass / fail judgment accuracy may not be stable in the initial stage. Therefore, when automating the appearance inspection apparatus 1, each defect is detected. It is preferable that the number of defect candidates to which the type belongs is equal to or more than an appropriate number.

  Further, in order to stably maintain the classification accuracy and the pass / fail judgment accuracy, in the defect classification process 25 and the pass / fail judgment process 26, when comparing the defect candidate feature amount P with the classification threshold 43 or the determination threshold 45, Except for the vicinity of the lower limit value and the upper limit value, for example, when the feature amount P is in the range of 80% between the lower limit value and the upper limit value, the type may be specified or the defect may be determined.

  As mentioned above, although embodiment of this invention was described referring an accompanying drawing, the technical scope of this invention is not influenced by embodiment mentioned above. It is obvious for those skilled in the art that various modifications or modifications can be conceived within the scope of the technical idea described in the claims, and these are naturally within the technical scope of the present invention. It is understood that it belongs.

DESCRIPTION OF SYMBOLS 1 ......... Appearance inspection apparatus 3 ......... Inspection sheet 5 ......... Camera 21 ......... Image input part 22 ......... Preprocessing part 23 ......... Defect candidate extraction part 24 ......... Feature amount calculation part 25 ………… Defect classification processing unit 26 ………… Quality determination processing unit 27 ………… Visual determination input unit 28 ………… Classification threshold update processing unit 29 ………… Determination threshold update processing unit 41 ………… Defect candidate feature Quantity database 43 ......... Classification threshold database 45 ......... Determination threshold database 50 ......... Visual decision input screen 51 ......... Defect map 53 ......... Defect image 55 ......... Defect position 57 ......... Defect classification Name option 59 ……… Visual judgment option

Claims (9)

An appearance inspection apparatus that determines the type of defect that has occurred in the appearance of an object to be inspected and whether it is acceptable or not
Image input means for inputting a captured image of the object to be inspected;
Defect candidate extraction means for extracting defect candidates from the captured image;
Feature quantity calculating means for calculating a plurality of feature quantities of the defect candidates;
First storage means for storing, for each defect type, a first threshold value related to the plurality of feature amounts for specifying the defect type;
Second storage means for storing a second threshold value for the plurality of feature amounts for determining the quality of the defect candidate for each type of defect;
A classification means for comparing a plurality of feature quantities of the defect candidate with the first threshold and identifying the type of the defect candidate;
Pass / fail judgment means for judging pass / fail of the defect candidate whose defect type is specified by the classifying means by comparing a plurality of feature quantities of the defect candidate and the second threshold;
When the classification means cannot identify the defect type of the defect candidate, the type input means for displaying position information of the defect candidate and selectively inputting the defect type of the defect candidate;
Pass / fail input means for displaying the position information of the defect candidate and inputting the pass / fail of the defect candidate when the defect candidate cannot be determined by the pass / fail determination means and when the defect candidate is determined to be defective.
Classification threshold value updating means for updating the first threshold value stored in the first storage means for the defect type selected and input by the type input means based on a plurality of feature quantities of the defect candidates. When,
PASS / FAIL to update the second threshold value stored in the second storage means for the defect type of the defect candidate whose pass / fail is input by the pass / fail input means based on a plurality of feature quantities of the defect candidate. A threshold updating means for determination;
A visual inspection apparatus comprising: The first threshold value is an upper limit value and a lower limit value determined for each of the plurality of feature amounts for each type of defect,
The classification means determines that the defect candidate belongs to the defect type of the first threshold value when all the feature values of the defect candidate are in the range of the upper limit value and the lower limit value of the first threshold value. The visual inspection apparatus according to claim 1, wherein the visual inspection apparatus is specified. When the feature value of the defect candidate exceeds the upper limit value of the first threshold, the classification threshold update unit sets the feature value of the defect candidate as the upper limit value of the first threshold, and the defect candidate when the feature amount of is below the lower limit of the first threshold value, 請 Motomeko the feature amounts of the defect candidate and updates the first threshold value as the lower limit value of the first threshold value 2. An appearance inspection apparatus according to 2. The second threshold value is an upper limit value and a lower limit value determined for each of the plurality of feature amounts for each type of the defect,
The pass / fail judgment means specifies that the defect candidate is defective when all the feature values of the defect candidate are in the range of the upper limit value and the lower limit value of the second threshold, feature value, if not in the range of the upper limit value and the lower limit value of the second threshold value, the appearance inspection apparatus according to claim 1, wherein the defect candidate is specified as not defective. When the feature value of the defect candidate exceeds the upper limit value of the second threshold value, the pass / fail judgment threshold update unit sets the feature value of the defect candidate as the upper limit value of the second threshold value, and the defect when the characteristic amount candidates is below the lower limit of the second threshold value, billed for and updates the second threshold value characteristic of the defect candidate as the lower limit value of the second threshold value Item 5. An appearance inspection apparatus according to Item 4 . The feature amount includes the area of the defect candidate, ferret diameter, circumscribed rectangular main axis / sub axis length, main axis angle, circularity, acicularity, circumscribed circle / inscribed circle diameter, circumferential length, equivalent ellipse main axis / sub axis length , equivalent ellipse major axis minor axis length ratio, the average luminance value, maximum luminance value, and, the appearance inspection apparatus according to any one of claims 1 to 5, characterized in that it comprises the value of brightness. The appearance inspection apparatus according to any one of claims 1 to 6 , wherein the types of defects include dirt, spots, scratches, and fibers. For each defect type, a first storage unit that stores a first threshold value related to a plurality of feature amounts for specifying the defect type, and for determining whether the defect candidate is good or bad for each defect type An appearance inspection method for determining a type and a pass / fail of a defect generated in an appearance of an object to be inspected, which is executed by an appearance inspection apparatus including a second storage unit that stores a second threshold relating to the plurality of feature amounts. ,
An image input step for inputting a captured image of the inspected object;
A defect candidate extraction step for extracting defect candidates from the captured image;
A feature amount calculating step for calculating a plurality of feature amounts of the defect candidates;
A step of comparing a plurality of feature quantities of the defect candidate with the first threshold and identifying the type of the defect candidate;
A pass / fail determination step for determining pass / fail of the defect candidate in which the type of the defect is specified by the classification step by comparing a plurality of feature amounts of the defect candidate with the second threshold;
A type input step of displaying the position information of the defect candidate and selectively inputting the type of the defect of the defect candidate when the defect type of the defect candidate cannot be specified by the classification step;
The pass / fail input step of displaying the position information of the defect candidate and inputting the pass / fail of the defect candidate when the pass / fail determination step cannot determine the pass / fail of the defect candidate and when it is determined to be defective.
A classification threshold updating step for updating the first threshold for the type of defect selected and input in the type input step based on a plurality of feature amounts of the defect candidates;
A pass / fail judgment threshold updating step for updating the second threshold for the defect type of the defect candidate whose pass / fail is input by the pass / fail input step based on a plurality of feature quantities of the defect candidate;
An appearance inspection method comprising: A program for causing a computer to function as an appearance inspection apparatus that determines the type of defect that has occurred in the appearance of an object to be inspected and whether it has passed or failed,
The computer,
An image input means for inputting a photographed image of the inspected object;
Defect candidate extraction means for extracting defect candidates from the captured image;
Feature amount calculating means for calculating a plurality of feature amounts of the defect candidates;
First storage means for storing, for each defect type, a first threshold value related to the plurality of feature amounts for specifying the defect type;
A second storage unit that stores a second threshold value for the plurality of feature amounts for determining the quality of the defect candidate for each type of defect;
Classification means for comparing a plurality of feature quantities of the defect candidate with the first threshold and identifying the type of the defect candidate.
Pass / fail judgment means for judging pass / fail of the defect candidate whose defect type is specified by the classifying means by comparing a plurality of feature quantities of the defect candidate with the second threshold;
Type input means for displaying position information of the defect candidate and selectively inputting the defect type of the defect candidate when the defect type of the defect candidate cannot be specified by the classification means;
A pass / fail input means for displaying the position information of the defect candidate and inputting the pass / fail of the defect candidate when the pass / fail determination means cannot determine the pass / fail of the defect candidate and when it is determined to be defective.
Classification threshold value updating means for updating the first threshold value stored in the first storage means for the defect type selected and input by the type input means based on a plurality of feature quantities of the defect candidates. ,
PASS / FAIL to update the second threshold value stored in the second storage means for the defect type of the defect candidate whose pass / fail is input by the pass / fail input means based on a plurality of feature quantities of the defect candidate. A threshold updating means for determination;
Program to function as.

Images