JP2008051558A - Visual inspection method and device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a visual inspection method and a visual inspection device capable of evaluating quantitatively a visual defect generated, for example, on a cabinet of a liquid crystal display. <P>SOLUTION: This visual inspection device 1 calculates a brightness level to an imaging surface as a continuous curved surface based on image data showing the brightness level of an image acquired by imaging an inspection object 50, and inspects quantitatively the appearance of the inspection object 50 based on an evaluation value calculated by multiplying each parameter by each weight coefficient corresponding to each parameter, by using a plurality of parameters such as the minimum curvature value with respect to the calculated continuous curved surface, the area of a crossing domain crossing with a virtual plane whose brightness level is a prescribed value, a volume of a closed space formed by the crossing domain and the continuous curved surface, or the maximum line segment length on the crossing domain, and also calculates the weight coefficient by a learning method of a neural network or the like. <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

  The present invention relates to an appearance inspection method and an appearance inspection apparatus for performing an appearance inspection of an inspection object from image data obtained by photographing the inspection object.

  A display device such as a liquid crystal display uses a plastic cabinet for its appearance. The cabinet is manufactured by injection molding, and defects in appearance such as a weld phenomenon occur in the appearance of the cabinet due to manufacturing variations. In view of this, an inspector performs a visual sensitivity inspection to determine whether or not it is defective.

  However, since the visual sensitive inspection is determined by the inspector's intuition, there is a problem that the inspection result varies depending on the inspector. In addition, even for the same inspector, there is a problem that the inspection result varies depending on the inspection date and time. As described above, it is very difficult to make a quantitative determination by visual inspection.

By the way, in recent years, a defect level calculation unit that calculates a defect level of a pixel to be inspected based on a video signal picked up by a television camera and a video inspection having a configuration in which a pass / fail is determined based on a predetermined reference level. An apparatus is disclosed (for example, refer to Patent Document 1).
JP 2001-112029 A

  However, the technique disclosed in Patent Document 1 relates to an inspection technique for inspecting a pixel defect, and cannot be applied to an appearance inspection for inspecting an appearance defect such as a weld phenomenon.

  The present invention has been made in view of such circumstances, and based on an image obtained by imaging an inspection object, the luminance level is expressed as a continuous curved surface, and the evaluation value calculated based on the expressed continuous curved surface An object of the present invention is to provide an appearance inspection method and an appearance inspection apparatus capable of evaluating defects in appearance such as a weld phenomenon by evaluating the appearance of an inspection object.

  The present invention also provides an appearance inspection method and an appearance inspection apparatus capable of improving the accuracy of evaluation by updating a weighting factor relating to a mathematical expression for calculating an evaluation value through learning using a neural network. The purpose.

  The present invention relates to an appearance inspection method for performing an appearance inspection of an inspection object based on image data indicating a luminance level of at least one of a plurality of colors of an image obtained by imaging the inspection object. A continuous curved surface calculating step of interpolating a luminance level at a position between each imaging pixel based on a luminance level at the position of each imaging pixel and calculating a continuous curved surface indicating a luminance level with respect to the imaging surface; And an evaluation value calculating step for calculating an evaluation value based on the evaluation value, and a determination step for performing a quantitative determination of the inspection object based on the calculated evaluation value.

  In the present invention, the evaluation value includes a plurality of parameters, and the evaluation value calculation step includes multiplying each parameter by a weighting factor corresponding to the parameter, and adding the multiplied value to the inspection value. The evaluation value of the object is calculated.

  The present invention further includes a weighting factor updating step of updating a weighting factor corresponding to each parameter based on learning by a neural network using a plurality of standard samples having preset evaluation values.

  In the present invention, the evaluation value calculating step includes a minimum curvature value calculating step of calculating a minimum curvature value on the continuous curved surface calculated in the continuous curved surface calculating step, and using the calculated minimum curvature value as one of the parameters. It is characterized by including.

  According to the present invention, the evaluation value calculation step assumes a virtual plane having a predetermined luminance level, and derives an intersection region deriving step for deriving an intersection region where the virtual plane intersects the continuous curved surface, and the derived intersection region And an area calculating step using the calculated area as one of the parameters.

  According to the present invention, the evaluation value calculation step assumes a virtual plane having a predetermined luminance level, and derives an intersection region deriving step for deriving an intersection region where the virtual plane intersects the continuous curved surface, and the derived intersection region And a volume calculating step of calculating a volume surrounded by the continuous curved surface and using the calculated volume as one of the parameters.

  According to the present invention, the evaluation value calculation step assumes a virtual plane having a predetermined luminance level, and derives an intersection region deriving step for deriving an intersection region where the virtual plane intersects the continuous curved surface, and the derived intersection region And calculating a maximum line segment length in step (a), and calculating a line segment length using the calculated maximum line segment length as one of the parameters.

  The present invention relates to an appearance inspection apparatus that performs an appearance inspection of an inspection object based on image data indicating a luminance level of at least one of a plurality of colors of an image obtained by imaging the inspection object, and is included in the image data A continuous curved surface calculating means for interpolating a luminance level at a position between each imaging pixel based on a luminance level at the position of each imaging pixel and calculating a continuous curved surface indicating a luminance level with respect to the imaging surface; and a calculated continuous curved surface An evaluation value calculating unit that calculates an evaluation value based on the evaluation value and a determination unit that quantitatively determines the inspection object based on the calculated evaluation value are provided.

  In the present invention, the evaluation value includes a plurality of parameters, and the evaluation value calculation means multiplies each parameter by a weighting factor corresponding to the parameter, and adds the multiplied value to the inspection value. An evaluation value of the object is calculated.

  The present invention is characterized by further comprising weight coefficient updating means for updating a weight coefficient corresponding to each parameter based on learning by a neural network using a plurality of standard samples having preset evaluation values.

  In the present invention, the inspection target is imaged, and image data indicating the luminance level of at least one of the plurality of colors of the captured image is acquired. The image data includes the luminance level at the position of each imaging pixel, and the luminance level at the position between the imaging pixels is interpolated based on the luminance level at the position of each imaging pixel included in the image data. Then, based on the luminance level at each imaging pixel and the position between each imaging pixel, a continuous curved surface indicating the luminance level with respect to the imaging surface is calculated. Thus, since the image data is discontinuous with respect to the imaging surface, interpolation is performed between the imaging pixels to form a continuous curved surface. Then, an evaluation value is calculated based on the calculated continuous curved surface, and a quantitative determination of the inspection object is performed based on the calculated evaluation value.

  In the present invention, the evaluation value is composed of a plurality of parameters, each parameter is multiplied by a weighting factor corresponding to the parameter, and the quantitative determination of the inspection object is performed based on the added value obtained by adding the multiplied values. Do.

  In the present invention, the weighting coefficient corresponding to each parameter is updated by learning using a neural network using a plurality of standard samples of evaluation values set in advance.

  In the present invention, the minimum curvature value in the continuous curved surface is calculated, and the calculated minimum curvature value is set as one of the evaluation value parameters. Since the minimum curvature value is an index indicating the magnitude of the change amount of the luminance level, it is possible to find a problem with a large change in the luminance level by using the minimum curvature value as a parameter of the evaluation value.

  In the present invention, a virtual plane having a predetermined luminance level is assumed, an intersection area where the assumed virtual plane and the continuous curved surface intersect is derived, and the area of the derived intersection area is calculated. Then, the calculated area of the intersecting region is set as one of evaluation value parameters. Since the area of the intersection region is, for example, the area of the region where the luminance level exceeds a predetermined level, it is possible to find a problem with a large area.

  In the present invention, assuming a virtual plane having a predetermined luminance level, a crossing area where the assumed virtual plane and the continuous curved surface intersect is derived, and the volume surrounded by the derived crossing region and the continuous curved surface Is calculated. The calculated volume is set as one of evaluation value parameters. The volume enclosed by the intersecting area and the continuous curved surface is, for example, an index combining the area and the excess degree of the area where the luminance level exceeds the predetermined level, so the area is large and the excess degree is large. Can be found.

  In the present invention, a virtual plane having a predetermined luminance level is assumed, an intersection area where the assumed virtual plane and the continuous curved surface intersect is derived, and the maximum line segment length in the derived intersection area is calculated. . The calculated maximum line length is set as one of the evaluation value parameters. Since the maximum line segment length in the intersection region is a length where the luminance level exceeds a predetermined level, it is possible to find a defect with a long length.

  The appearance inspection method and the appearance inspection apparatus according to the present invention are based on image data indicating a luminance level of at least one of a plurality of colors of an image obtained by imaging an inspection object, and between each imaging pixel included in the image data. The luminance level at the position is interpolated based on the luminance level at the position of each imaging pixel, the luminance level for the imaging surface is calculated as a continuous curved surface, the evaluation value is calculated based on the calculated continuous curved surface, and the calculated evaluation value Based on this, a quantitative determination of the inspection object is performed.

  With this configuration, in the present invention, the appearance of the inspection object can be expressed as a continuous curved surface based on the luminance level, and the features of the continuous curved surface can be quantified. There is an excellent effect such that it is possible to evaluate quantitatively based on a certain standard without depending on intuition.

  In addition, the appearance inspection method of the present invention is characterized in that, as a characteristic of a continuous curved surface that is quantified, a minimum curvature value, an area of a crossing region that intersects a virtual plane having a predetermined luminance level, a crossing region, and a closed curved surface formed by a continuous curved surface. By using multiple parameters such as the volume of space and the maximum line segment length on the intersecting area, multiplying each parameter by the weighting coefficient corresponding to each parameter, and adding the evaluation value to calculate the appearance There are excellent effects such as being able to properly evaluate various characteristics of the above defects according to their weight.

  Furthermore, the appearance inspection method of the present invention has an excellent effect that the weighting factor can be optimized by updating the weighting factor based on learning by a neural network.

  Hereinafter, the present invention will be described in detail with reference to the drawings illustrating embodiments thereof.

(Embodiment 1)
FIG. 1 is a block diagram showing a configuration example of an appearance inspection apparatus according to Embodiment 1 of the present invention. In FIG. 1, reference numeral 1 denotes an appearance inspection apparatus according to Embodiment 1 of the present invention, and the appearance inspection apparatus 1 includes a control unit 10 composed of a CPU. The control unit 10 is connected to the ROM 11, the RAM 12, the imaging unit 13, the image memory 14, the image processing unit 15, the determination unit 16, the input unit 17, and the output unit 18, and controls these units and is stored in the ROM 11 in advance. Various functions are executed according to the computer program. The appearance inspection apparatus 1 is an apparatus for inspecting the appearance of an inspection object 50 such as a plastic cabinet used for a liquid crystal display, and inspects whether there is a defect such as a weld phenomenon on a cabinet surface manufactured by injection molding. Is possible.

  The RAM 12 is configured using a memory such as a DRAM or an SDRAM, and stores temporary data generated when the computer program is executed by the control unit 10.

  The imaging unit 13 includes an imaging element such as a CCD, and images the inspection target 50 to generate image data. The generated image data is stored in the image memory 14, and various processes are performed according to instructions from the control unit 10. FIG. 2 is an explanatory diagram showing an example of image data generated by the appearance inspection apparatus 1 according to Embodiment 1 of the present invention. FIG. 2 shows the image data stored in the image memory 14, and the image data is L11, L21,..., Lmn indicating the luminance level at each position corresponding to each imaging element of the imaging unit 13. Is included as a discrete value. These luminance data indicate the luminance level of at least one of a plurality of colors such as the L * value, R (red) component, G (Green) component, and B (Blue) component at the position of each imaging pixel. In the figure, the X axis represents the horizontal axis of the captured image, and the Y axis represents the vertical axis of the captured image.

  The image processing unit 15 includes a continuous curved surface calculation unit 15a, a minimum curvature value calculation unit 15b, an intersection region derivation unit 15c, an area calculation unit 15d, a volume calculation unit 15e, and a line segment length calculation unit 15f.

  The continuous curved surface calculation unit 15a calculates the luminance level at the position between the respective imaging pixels by interpolating the luminance level at the position of each imaging pixel included in the image data captured by the imaging unit 13 and stored in the image memory 14. Then, based on the luminance level at each imaging pixel and the position between each imaging pixel, a continuous curved surface indicating the luminance level with respect to the imaging surface is calculated. FIG. 3 is a three-dimensional graph schematically showing an example of a continuous curved surface calculated by the appearance inspection apparatus 1 according to Embodiment 1 of the present invention. As described with reference to FIG. 2, the image data is discontinuous data indicating the luminance level at discrete positions corresponding to each imaging element of the imaging unit 13. By calculating and interpolating, a continuous curved surface S which is a set of continuous data as shown in FIG. 3 is calculated. Note that the X axis and Y axis in FIG. 3 correspond to the X axis and Y axis in FIG. 2, and the Z axis indicates the luminance level. That is, the continuous curved surface S is a virtual surface that represents the brightness level of each position on the imaging surface as high and low. As the interpolation process, a known technique such as a least square method or a spline method is used.

  FIG. 4 is a graph showing an example of the change in the X-axis direction of the luminance level calculated by the appearance inspection apparatus 1 according to Embodiment 1 of the present invention. The X axis and Z axis in FIG. 4 correspond to the X axis and Z axis in FIG. As shown in FIG. 4, the minimum curvature value calculation unit 15a calculates a minimum curvature value A by performing a process such as second-order differentiation on the continuous curved surface S calculated by the continuous curved surface calculation unit 15a. Note that the position of the minimum curvature value A indicates the maximum change point of the luminance level constituting the image data. Since the minimum curvature value A is an index indicating the magnitude of the change amount of the luminance level, it is possible to detect a defect reflected as the magnitude of the change in luminance level based on the minimum curvature value A.

  FIG. 5 is a graph illustrating an example of changes in the X-axis direction and the Y-axis direction of the luminance level calculated by the appearance inspection apparatus 1 according to Embodiment 1 of the present invention. The X axis, Y axis, and Z axis in FIG. 5 correspond to the X axis, Y axis, and Z axis in FIG. As shown in FIG. 5, the intersection area deriving unit 15c assumes a virtual plane P having a predetermined luminance level, and intersects the virtual plane P and the continuous curved surface S calculated by the continuous curved surface calculation unit 15a. R is derived. The predetermined value of the luminance level is set based on, for example, statistical values such as an average value, deviation, and variance of the luminance level change.

  As shown in FIG. 5, the area deriving unit 15d calculates the area B of the intersecting region R derived by the intersecting region deriving unit 15c. Further, as shown in FIG. 5, the volume calculation unit 15 e is a closed space surrounded by the intersection region R derived by the intersection region deriving unit 15 c and the continuous curved surface S calculated by the continuous curved surface calculation unit 15 a. Volume C is calculated. Further, as shown in FIG. 5, the line segment length calculation unit 15f calculates the maximum line segment length D in the intersection region R derived by the intersection region deriving unit 15c. Since the area B of the intersection region R is an index indicating the area of the region where the luminance level exceeds a predetermined value, it is possible to detect a problem with a large area based on the area B. The volume C of the closed space surrounded by the intersecting region R and the continuous curved surface S is an index that combines the area of the region where the luminance level exceeds a predetermined value and the degree of excess, so the area C is based on the volume C. It is possible to detect a problem with a large and large degree. Since the maximum line segment length D in the intersection region R is an index indicating the length of the luminance level exceeding a predetermined value, it is possible to detect a long defect based on the line segment length D.

  The determination unit 16 includes weights indicated as ω1, ω2, ω3, and ω4 corresponding to the parameters of the minimum curvature value A, the area B, the volume C, and the line segment length D calculated by the 15 calculation units of the image processing unit. An evaluation value that sets a coefficient, multiplies each parameter by a corresponding weighting coefficient, and adds the obtained values to obtain an evaluation value H (H = ω1 × A + ω2 × B + ω3 × C + ω4 × D) A calculation unit 16a is provided. The weighting factor is set in advance by a user or the like, for example, and is stored in a memory (not shown) built in the determination unit 16.

  Moreover, the determination part 16 performs the quantitative determination of the test object 50 based on the evaluation value H calculated by the evaluation value calculation part 16a. For example, the threshold value Hth for discriminating between good / bad is stored in the above-described memory, and the good / bad of the inspection object 50 is determined by comparing the calculated evaluation value H with the threshold Hth.

  The input unit 17 is a device that accepts external inputs such as a keyboard, push buttons, and a mouse. The output unit 18 is a device that performs output to the outside, such as a monitor, an LED, and a speaker.

  Next, the operation of the appearance inspection apparatus 1 according to the first embodiment of the present invention configured as described above will be described. FIG. 6 is a flowchart showing an example of inspection processing of the appearance inspection apparatus 1 according to Embodiment 1 of the present invention.

  First, the control unit 10 included in the appearance inspection apparatus 1 controls the imaging unit 13 to image the inspection target 50 and generate image data indicating a luminance level (step S1). The generated image data is stored in the image memory 14. Then, an area (inspection area) to be inspected in the image data is designated (step S2). By setting an area for designating the inspection area in advance, the process of S2 can be omitted.

  Next, the control unit 10 controls the image processing unit 15, interpolates the luminance level at the position of each imaging pixel included in the image data, calculates the luminance level at the position between the imaging pixels, Based on the luminance level at the position between the imaging pixels, a continuous curved surface S indicating the luminance level with respect to the imaging surface is calculated (step S3). Data indicating the calculated continuous curved surface S is stored in the RAM 12. And the control part 10 controls the image process part 15, and calculates the minimum curvature value A of the continuous curved surface S memorize | stored in RAM12 (step S4).

  Further, the control unit 10 controls the image processing unit 15 to assume a virtual plane P having a predetermined luminance level (step S5), and derive an intersection region R where the virtual plane P and the continuous curved surface S intersect (step S5). S6). Then, the area B of the intersecting region R is calculated (step S7), and the volume C of the closed space surrounded by the intersecting region R and the continuous curved surface S is calculated (step S8). Furthermore, the maximum line segment length D in the intersection region R is calculated (step S9).

  Next, the calculated parameters (minimum curvature value A, area B, volume C, line segment length D) are respectively multiplied by weighting factors (ω1, ω2, ω3, ω4) corresponding to each parameter, and multiplication is performed. An added value obtained by adding the obtained values is calculated as an evaluation value H (step S10). Then, the calculated evaluation value H and the threshold value Hth are compared to determine whether the inspection object 50 is good or bad (step S11), and the determination result is output from the output unit 18. For example, when the evaluation value H is larger than the threshold value Hth, it is determined that the inspection target 50 is defective.

  As described above, according to the first embodiment, it is possible to quantitatively determine good / bad based on the image data obtained by imaging the inspection object 50. There is no fear that the inspection results differ depending on the inspector and the inspection date and time. Therefore, it is possible to suppress an error that a defective product is determined as a non-defective product and shipped, or an error that a non-defective product is determined as a defective product and discarded.

(Embodiment 2)
In the first embodiment, the weighting coefficients (ω1, ω2, ω3, ω4) are set in advance by, for example, the user and stored in a memory (not shown) built in the determination unit 16; The plurality of standard samples may be used to update the weighting coefficients (ω1, ω2, ω3, ω4) corresponding to each parameter by a learning method such as a neural network, and such is the second embodiment. .

  FIG. 7 is a block diagram showing a configuration example of an appearance inspection apparatus according to Embodiment 2 of the present invention. The appearance inspection apparatus according to the second embodiment of the present invention has a configuration in which a weight coefficient updating unit 20 is added to the appearance inspection apparatus 1 shown in the first embodiment. Since other configurations are the same as those in the first embodiment, the same reference numerals as those in the first embodiment are given and the description thereof is omitted.

  The weight coefficient updating unit 20 has a function of updating a weight coefficient corresponding to each parameter using a learning method such as a neural network using a plurality of standard samples. The standard sample is a sample whose evaluation value H is determined in advance by one or a plurality of inspectors who make highly reliable determinations such as experienced inspectors and inspection managers. As standard samples, samples having different evaluation values H and samples having the same evaluation value H but different parameters (in this example, minimum curvature value A, area B, volume C, line segment length D) are included. Is used.

  Then, by the same method as the inspection process of the inspection object 50 described in the first embodiment, the minimum curvature value A, area B, volume C, Parameters such as line segment length D are calculated. Since the evaluation value H of each standard sample is determined in advance, a weighting coefficient is calculated by a learning method such as a neural network so as to take a reasonable value as compared with the determined evaluation value H, and is calculated. The update processing is performed so that the weighting factor thus obtained becomes the weighting factor at the time of inspection.

  Next, the operation of the weight coefficient update process of the appearance inspection apparatus 1 according to the second embodiment of the present invention configured as described above will be described. The process for calculating various parameters for the inspection object 50 is the same as the process described with reference to the flowchart of the first embodiment. FIG. 8 is a flowchart showing an example of the weighting factor update process of the appearance inspection apparatus 1 according to Embodiment 2 of the present invention.

  First, the control unit 10 included in the appearance inspection apparatus 1 controls the imaging unit 13 to capture an image of a standard sample and generate image data indicating a luminance level (step S21). The generated image data is stored in the image memory 14. Then, an area (inspection area) to be inspected in the image data is designated (step S22). By setting an area for designating the inspection area in advance, the process of S22 can be omitted.

  Next, the control unit 10 controls the image processing unit 15, interpolates the luminance level at the position of each imaging pixel included in the image data, calculates the luminance level at the position between the imaging pixels, Based on the luminance level at the position between the imaging pixels, a continuous curved surface S indicating the luminance level with respect to the imaging surface is calculated (step S23). Data indicating the calculated continuous curved surface S is stored in the RAM 12. And the control part 10 controls the image process part 15, and calculates the minimum curvature value A of the continuous curved surface S memorize | stored in RAM12 (step S24).

  Further, the control unit 10 controls the image processing unit 15 and assumes a virtual plane P having a predetermined luminance level (step S25), and derives an intersection region R where the virtual plane P and the continuous curved surface S intersect (step S25). S26). Then, the area B of the intersecting region R is calculated (step S27), and the volume C surrounded by the intersecting region R and the continuous curved surface S is calculated (step S28). Further, the maximum line segment length D in the intersection region R is calculated (step S29).

  Next, the control unit 10 controls the image processing unit 15, and sets each parameter based on each parameter calculated in the processes of steps S <b> 21 to S <b> 29 and the evaluation value H determined in advance for the standard sample. The corresponding weighting factor is calculated by the neural network (step S30), and each stored weighting factor is updated based on the calculated weighting factor (step S31). The evaluation value H used in step S30 is sequentially received from the input unit 17, or the value received in advance and stored in the RAM 12 is used. Then, using the evaluation value H as a teacher signal, importance and relevance of each parameter are calculated by a learning method such as a neural network, and a weighting coefficient is calculated.

  And the process of step S21-S31 is performed with respect to all the standard samples, and a process is complete | finished.

  As described above, according to the second embodiment, each parameter is calculated based on image data obtained by imaging a plurality of standard samples in which evaluation values are set in advance, and the evaluation value of each standard sample is valid. Since the weighting coefficient is calculated by a learning method such as a neural network so that the value becomes a new value and the calculated weighting coefficient is updated to be the weighting coefficient at the time of inspection, the user needs to set the weighting coefficient for each parameter. Disappears. In other words, by giving an evaluation value as a criterion for determining good / bad, the appearance inspection apparatus automatically calculates and updates the weighting coefficient so that the evaluation value of each standard sample is valid. Therefore, the burden on the user can be greatly reduced.

  In the first and second embodiments, the mode using the minimum curvature value, area, volume, and line segment length as parameters is shown, but the present invention is not limited to this, and various parameters can be set. In addition, when using more parameters, it is possible to use various models that are more complex than the simple model described above, and it is important to set the optimal weighting factor using the learning method, and further, the connection weight, etc. It becomes. For example, a model represented by the following mathematical formula is used for setting various weighting factors.

ak = f (ΣPi ωki)
H = f (Σai Vi)
Where f (): sigmoid function
Pi: Parameter
ωki: Weight coefficient
Vi: Join weight
H: Evaluation value

  In the first and second embodiments, the weld phenomenon that occurs in the appearance of the cabinet of the display device is an inspection target. However, the inspection target is not limited to this. The inspection object may be (due to variations in liquid crystal gap, variations in alignment film drying, etc.). In this case, non-uniformity is revealed by performing gray scale display, and the state is captured by the imaging unit to acquire image data. Each parameter constituting the evaluation value is an example in the case of evaluating the weld phenomenon, and the evaluation of the weld phenomenon may be performed using other parameters. When evaluating another phenomenon, each parameter is appropriately determined so that the phenomenon can be quantitatively evaluated.

  Furthermore, the present invention can be developed in a form of comprehensively inspecting a plurality of luminance levels indicating stimulus values of a plurality of color components.

It is a block diagram which shows the structural example of the external appearance inspection apparatus which concerns on Embodiment 1 of this invention. It is explanatory drawing which shows an example of the image data which the external appearance inspection apparatus which concerns on Embodiment 1 of this invention produces | generates. It is a three-dimensional graph which shows typically an example of the continuous curved surface calculated with the external appearance inspection apparatus which concerns on Embodiment 1 of this invention. It is a graph which shows an example of the change of the X-axis direction of the luminance level computed with the appearance inspection apparatus which concerns on Embodiment 1 of this invention. It is a graph which shows an example of the change of the X-axis direction of the brightness | luminance level calculated with the external appearance inspection apparatus which concerns on Embodiment 1 of this invention, and a Y-axis direction. It is a flowchart which shows an example of the test | inspection process of the external appearance inspection apparatus which concerns on Embodiment 1 of this invention. It is a block diagram which shows the structural example of the external appearance inspection apparatus which concerns on Embodiment 2 of this invention. It is a flowchart which shows an example of the weighting coefficient update process of the external appearance inspection apparatus which concerns on Embodiment 2 of this invention.

Explanation of symbols

DESCRIPTION OF SYMBOLS 10 Control part 13 Imaging part 14 Image memory 15 Image processing part 15a Continuous curved surface calculation part 15b Minimum curvature value calculation part 15c Crossing area derivation part 15d Area calculation part 15e Volume calculation part 15f Line segment length calculation part 16 Judgment part 16a Evaluation value calculation 20 Weight coefficient update unit

Claims (10)

In an appearance inspection method for performing an appearance inspection of the inspection object based on image data indicating a luminance level of at least one of a plurality of colors of an image obtained by imaging the inspection object,
A continuous curved surface calculating step of interpolating a luminance level at a position between each imaging pixel included in the image data based on a luminance level at a position of each imaging pixel and calculating a continuous curved surface indicating a luminance level with respect to the imaging surface;
An evaluation value calculating step for calculating an evaluation value based on the calculated continuous curved surface;
An appearance inspection method comprising: a determination step of performing a quantitative determination of the inspection object based on the calculated evaluation value. The evaluation value includes a plurality of parameters,
The evaluation value calculating step includes:
The appearance inspection method according to claim 1, wherein each parameter is multiplied by a weighting factor corresponding to the parameter, and an evaluation value of the inspection object is calculated based on an addition value obtained by adding the multiplied values. . The weight coefficient update step of updating a weight coefficient corresponding to each parameter based on learning by a neural network using a plurality of standard samples having a preset evaluation value. Visual inspection method. The evaluation value calculating step includes:
The minimum curvature value calculation step which calculates the minimum curvature value in the continuous curved surface calculated in the continuous curved surface calculation step and uses the calculated minimum curvature value as one of the parameters is included. Item 3. The appearance inspection method according to Item 3. The evaluation value calculating step includes:
Assuming a virtual plane whose luminance level is a predetermined value, an intersecting area deriving step for deriving an intersecting area where the virtual plane intersects the continuous curved surface;
The visual inspection according to claim 2, further comprising: calculating an area of the derived intersection region, and calculating an area using the calculated area as one of the parameters. Method. The evaluation value calculating step includes:
Assuming a virtual plane whose luminance level is a predetermined value, an intersecting area deriving step for deriving an intersecting area where the virtual plane intersects the continuous curved surface;
5. A volume calculation step of calculating a volume surrounded by the derived intersection area and the continuous curved surface, and using the calculated volume as one of the parameters. 5. The visual inspection method according to claim 1. The evaluation value calculating step includes:
Assuming a virtual plane whose luminance level is a predetermined value, an intersecting area deriving step for deriving an intersecting area where the virtual plane intersects the continuous curved surface;
5. A line segment length calculating step of calculating a maximum line segment length in the derived intersection region and using the calculated maximum line segment length as one of the parameters. An appearance inspection method according to any one of the above. In an appearance inspection apparatus that performs an appearance inspection of the inspection object based on image data indicating a luminance level of at least one of a plurality of colors of an image obtained by imaging the inspection object,
Continuous curved surface calculating means for interpolating the luminance level at the position between each imaging pixel included in the image data based on the luminance level at the position of each imaging pixel, and calculating a continuous curved surface indicating the luminance level with respect to the imaging surface;
Evaluation value calculating means for calculating an evaluation value based on the calculated continuous curved surface;
An appearance inspection apparatus comprising: a determination unit configured to quantitatively determine the inspection target based on the calculated evaluation value. The evaluation value includes a plurality of parameters,
The evaluation value calculation means includes
9. The evaluation value of the inspection object is calculated based on an addition value obtained by multiplying each parameter by a weighting factor corresponding to the parameter and adding the multiplied values. Visual inspection equipment. The weight coefficient update means for updating the weight coefficient corresponding to each parameter based on learning by a neural network using a plurality of standard samples having preset evaluation values. Visual inspection equipment.

Images