JPH0572139A - Inspection of surface state through bright and dark illumination - Google Patents

Abstract

PURPOSE:To properly inspect surface state, as for the surface state inspection method for inspecting the surface state by the comparison between the brightness/darkness data and the judgement value for detecting defect which are obtained from the detection of the reflection light supplied from the inspected surface. CONSTITUTION:The bright and dark light whose brightness gradually changes slong a prescribed direction is radiated onto an inspected surface and a light receiving image is formed by catching the reflection light supplied from the inspected surface. A determining rule for determining the setting tendency of the judgement value for detecting defect in correspondence with the gradient of the brightness and darkness and the magnitude of the defect is previously set, and the degree of fittiness of the judgement value for detecting defect is calculated on the basis of the determining rule and the magnitude of the defect which is calculated from the data and the gradient of brightness and darkness which is calculated from the brightnes and darkness data obtained through the scanning of the light receiving image, and the judgement value for detecting defect is varied according to the degee of fittingness.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method for inspecting a surface condition, and in particular to irradiating a surface to be inspected with light by means of a light irradiator, and capturing light reflected from the surface to be inspected by an image pickup means to create a light receiving image, The present invention relates to a surface state inspection method for inspecting the surface state of a surface to be inspected based on the brightness difference in the received light image.

[0002]

2. Description of the Related Art As a technique for inspecting the surface condition of the painted surface of an automobile body, for example, Japanese Patent Laid-Open No. 62-233710.
Japanese Patent Laid-Open Publication No. 2003-242242 discloses a technique for irradiating light on a surface to be inspected of an inspection object, projecting the reflected light on a screen, and automatically detecting the surface state of the surface to be inspected from the sharpness of the projected image. Is disclosed.

By the way, as an example of a technique for detecting a surface defect by irradiating the surface to be inspected with light, capturing the reflected light and performing image processing as described above, a predetermined gradient suitable for the detection of the surface defect. Illuminates the surface to be inspected with bright and dark light whose intensity changes gradually, and the reflected light reflected from the surface to be inspected is CC
It is considered that a light receiving image is created by being captured by an image pickup means such as a D camera, and that a surface defect is detected based on a change in lightness of a light receiving amount output by horizontal scanning of the light receiving image. In that case, for example, it is determined whether or not the change amount per predetermined pitch of the video signal corresponding to the received light amount output from the image pickup unit has a value that exceeds a predetermined defect detection determination value. When it is performed and the change amount of the signal shows a change exceeding the determination value, it is determined to be a surface defect.

[0004]

However, when determining a surface defect based on the video signal output from the image pickup means as described above, the problem is how to set the determination value for defect detection. Becomes That is, even if the surface to be inspected is illuminated with a constant gradient of light and dark light, ambient light is incident on the image pickup means in addition to the regular reflected light. It does not always accurately reflect the change in the bright and dark light that is actually applied to the surface. In that case, even if the size of the defective portion on the surface to be inspected is constant, as the brightness gradients A, B, and C seen at the signal level of the video signal are larger as shown in FIG. Height difference of the video signals shown A _H , B _H ,
C _H also tends to increase.

Therefore, if the defect detection judgment value is set to a large value in correspondence with the bright / dark gradient C having a large gradient, the difference in height A _H is small when the bright / dark gradient A has a small gradient, and therefore the defect may be overlooked. There is a problem in terms of sex. If the determination value for defect detection is set small in order to cope with this and the detection sensitivity is increased, the height difference of the video signal is enlarged even for a minute uneven portion that does not affect the quality when the brightness gradient is large. Therefore, if the height difference is larger than the defect detection determination value, it is determined as a defect, which causes a problem in determination accuracy.

In particular, the above-mentioned problem becomes remarkable in a curved portion where the intensity of reflected light is not constant.

The present invention addresses the above-mentioned problems in the surface state inspection method for inspecting the surface state by comparing the light and dark data obtained by detecting the reflected light from the surface to be inspected and the judgment value for defect detection. The purpose is to be able to properly inspect.

[0008]

That is, the surface condition inspection method using bright and dark illumination according to the invention of claim 1 of the present application (hereinafter referred to as the first invention) is a bright and dark light whose brightness gradually changes along a predetermined direction. Is irradiated to the surface to be inspected, the reflected light from the surface to be inspected is captured to create a light-receiving image, and this light-receiving image is scanned in the direction of light-dark change, and the detected light-dark data and predetermined defect detection judgment In a surface state inspection method for inspecting a surface state on a surface to be inspected by comparing the value with a value, the defect detection determination value is changed according to light and dark data obtained by scanning the light-receiving image. ..

The surface condition inspection method using bright and dark illumination according to the second aspect of the present invention (hereinafter referred to as the second invention) irradiates the surface to be inspected with bright and dark light whose brightness gradually changes along a predetermined direction. Then, the reflected light from the surface to be inspected is captured to create a light-receiving image, the light-receiving image is scanned in the direction of light-dark change, and the detected light-dark data is compared with a predetermined judgment value for defect detection. In the surface state inspection method for inspecting the surface state on the surface to be inspected, for the defect detection according to the light-dark gradient calculated from the light-dark data obtained by scanning the light-receiving image and the size of the defect calculated from the data. The feature is that the judgment value is changed.

The invention of claim 3 of the present application (hereinafter, referred to as the third
According to the invention), a surface state inspection method using bright and dark illumination irradiates a surface to be inspected with light and dark light whose brightness gradually changes along a predetermined direction, and captures reflected light from the surface to be inspected to create a light-receiving image. At the same time, the light-receiving image is scanned in the direction of light-dark change, and the surface state inspection method for inspecting the surface state of the surface to be inspected by comparing the detected light-darkness data with a predetermined defect detection determination value, Setting a decision rule in which the tendency of setting the judgment value for defect detection with respect to the light and dark data is set in advance, and adapting the judgment value for defect detection based on the light and dark data obtained by scanning the light-receiving image and the above decision rule It is characterized in that the degree is calculated and the defect detection determination value is changed according to the degree of conformity.

Further, according to the invention of claim 4 of the present application (hereinafter referred to as the fourth invention), the surface condition inspection method by the light and dark illumination irradiates the surface to be inspected with light and dark light whose brightness gradually changes along a predetermined direction. Then, the reflected light from the surface to be inspected is captured to create a light-receiving image, the light-receiving image is scanned in the direction of light-dark change, and the detected light-dark data is compared with a predetermined judgment value for defect detection. In the surface condition inspection method for inspecting the surface condition on the surface to be inspected by the above, a decision rule in which the tendency of setting the determination value for defect detection is determined corresponding to the brightness gradient and the size of the defect is set in advance, The degree of conformity of the defect detection determination value is calculated based on the lightness / darkness gradient calculated from the lightness / darkness data obtained by scanning the image, the size of the defect calculated from the data, and the determination rule, and Flip and wherein the changing the defect detection determination value.

According to the invention of claim 5 of the present application (hereinafter referred to as the fifth invention), the surface condition inspection method using the light and dark illumination irradiates the surface to be inspected with light and dark light whose brightness repeatedly changes along a predetermined direction. Then, the reflected light from the surface to be inspected is captured to create a light-receiving image, the light-receiving image is scanned in the direction of light-dark change, and the detected light-dark data is compared with a predetermined judgment value for defect detection. In the surface condition inspection method for inspecting the surface condition on the surface to be inspected, the defect detection determination value is changed according to the light and dark data obtained by scanning the light-receiving image.

[0013]

According to the first aspect of the invention, the defect detection determination value is changed by, for example, the brightness gradient, so that the surface inspection can be performed accurately even if the brightness gradient changes.

According to the second aspect of the invention, since the defect detection determination value is changed depending on the brightness / darkness gradient and the size of the defect, it is not affected by the disturbance caused by the curvature of the surface to be inspected. The surface inspection will be performed with high accuracy.

Further, according to the third and fourth aspects of the present invention, since it is sufficient to store a simple decision rule indicating the tendency of setting the defect detection judgment value with respect to the brightness gradient, the memory capacity is saved, and at the same time, It is possible to set the determination value with higher accuracy than in the case where the approximation calculation of the determination value for defect detection is performed using a regression equation represented by a straight line or the like.

Further, according to the fifth aspect, even if the gradient of the bright and dark light changes, an appropriate defect detection determination value is set, and the detection accuracy is improved.

[0017]

Embodiments of the present invention will be described below with reference to the drawings.

First, a surface condition inspection apparatus configured to inspect a vehicle body painted surface by the surface condition inspection method according to the present invention will be described. As shown in FIG.
In the vicinity of the vehicle body 1 conveyed to the coating inspection station S, a surface state inspection device 2 for inspecting the coating surface 1a of the vehicle body 1 to detect the presence or absence of coating defects is arranged. , A robot apparatus 4 mounted on the pedestal 3, and a light source 5 and a C
A CD camera 6 is attached via a support fitting 7, and these light source 5 and CCD camera 6 trace the surface of the vehicle body 1 carried into the coating station S, that is, the coating surface 1a of the vehicle body 1, At this time, the light emitted from the light source 5 is reflected by the coating film surface 1a of the vehicle body 1 and received by the CCD camera 6.

In the coating defect inspection by the light source 5 and the CCD camera 6 as described above, the robot controller 9 is controlled by a command given by the host computer 8, and the signal of the robot controller 9 is sent to the robot apparatus 4. Then, a predetermined actuator (not shown) built in the robot apparatus 4 is driven, whereby the robot apparatus 4 drives the light source 5 and the CCD camera 6 so that they trace the coating surface 1a of the vehicle body 1. While moving, the received light image obtained by the CCD camera 6 is sent to the image processor 10. Then, the image processor 10 amplifies the video signal from the CCD camera 6 and then performs image processing by identifying the level difference of the brightness of the received light image indicated by the signal, and the brightness data is processed by the host computer 8 It is configured to detect the presence or absence of a coating defect on the coating film surface 1a of the vehicle body 1, the coordinates of the defective portion, the shape of the coating defect, and the size thereof.

Next, the structure of the light source 5 will be described. As shown in FIG. 2, the light source 5 has a plurality of fluorescent lamps (particularly limited to fluorescent lamps) provided in a box 11 having one side open. 12), an optical filter 13 provided on the front surface of the fluorescent lamp 12 and closing one side surface of the box 11, and a diffusion screen 14. Then, the optical filter 13 is configured such that the light transmittance of each fluorescent lamp 12 is different depending on the transmitting place so that the light emitted from each fluorescent lamp 12 can be converted into, for example, bright and dark light whose brightness gradually changes from dark to bright. Has been done. That is, in this embodiment, the transmittance is repeatedly changed from small to large only in the x direction of the x and y directions orthogonal to each other shown in FIG. As shown in FIG. 5, bright and dark light whose brightness changes with a predetermined light and dark gradient is formed along the x direction, and this bright and dark light is applied to the coating surface 1a of the vehicle body 1.

The diffusing screen 14 diffuses the light transmitted through the optical filter 13 to diffuse the light through the coating surface 1a of the vehicle body 1.
It is for evenly irradiating bright and dark light.

In the surface condition inspection device 2 configured as described above, the defect inspection work is started as the painted vehicle body 1 is carried into the coating station S.
That is, the robot device 4 is controlled by the robot controller 9 so that the light source 5 and the CCD camera 6 are kept at a constant interval, and both are kept at an appropriate distance from the coating film surface 1a. Then, it is moved along the coating film surface 1a. Then, at that time, as shown in FIG. 4, with the light source 5, bright and dark light whose brightness gradually changes from dark to bright in the x direction with respect to a relatively wide coating surface 1a covering the field of view F of the CCD camera 6. 5a is irradiated. Therefore, a light irradiation region in which the brightness gradually changes from dark to bright is formed on the coating film surface 1a. Then, as shown in FIG. 5, a light-receiving image 1 of the CCD camera 6 that receives the reflected light from the light irradiation region.
In 5, the lightness gradually changes from dark to light in the direction indicated by arrow X corresponding to the direction x of change of the dark-light 5a from dark to light. Note that FIG. 5 shows that the lighter the line density, the higher the lightness, and the denser the line, the lighter the lightness.

Therefore, if the light irradiation area on the coating surface 1a of the vehicle body 1 has a convex defect z as shown in FIG. 4, for example, a convex defect portion having a lightness different from that of the surroundings at the corresponding position in the received light image 15 is obtained. Z will be generated.

Then, the CCD camera 6 sequentially scans the received light image 15 in the Y direction for each scanning line along the X direction, and outputs a video signal which changes according to the change in the brightness to the image processor 10. To do. The image processor 10 amplifies the input video signal, then performs predetermined image processing to convert it into bright and dark data, and outputs it to the host computer 8. Then, the host computer 8 detects the coordinates, shape and size of the position of the coating defect based on the change of the light and dark data fetched from the image processor 10 and stores it in the memory. Then, when repairing a coating defect, the stored contents in the memory are taken out, and a predetermined repair operation according to the type of the defect is executed.

Next, the surface defect detection processing which is a characteristic part of the present invention will be described. Specifically, this surface defect detection processing is performed as follows according to the flowchart of FIG.

First, the host computer 8 performs step S.
At 1, the minimum judgment value for defect detection is set. As the determination value for defect detection, for example, as indicated by the broken line in FIG. 7, the defect detection upper limit level UL indicating the upper limit of the light and dark data is displayed.
Then, the defect detection lower limit level DL indicating the lower limit of the data is set.

Next, the host computer 8 sets the detection section D of the light and dark data in steps S2 and S3, and then
Brightness data corresponding to the detection section D is used as the light-receiving image 1 in FIG.
Read along the X direction in 5. Then, the brightness data read in step S4 is delayed, and in step S5, the defect is detected by comparing the defect detection determination value with the brightness data. That is, the host computer 8 receives the signal level L of the bright and dark data per detection section taken from the CCD camera 6 via the image processor 10.
Is tentatively determined to be a defect when it is recognized as being higher than the defect detection upper limit level UL or smaller than the defect detection lower limit level DL, and subsequently, in step S6, the brightness data is 2
It is determined whether or not there is a delay for the section, and if NO is determined, the process returns to step S3 to read the light / dark data of the next detection section D, the light / dark data is delayed in step S4, and the minimum judgment is made in step S5. The value is used to determine whether there is a defect. That is, at the time of the first defect detection, it is regarded as a disturbance in order to prevent erroneous determination.

As the defect detection determination value immediately after the start, the minimum determination value set in step S1 is used.

On the other hand, when the host computer 8 determines in step S6 that the brightness data is delayed by two sections, it executes step S7 to calculate the brightness gradient. Specifically, the light / dark gradient is obtained as the amount of change in the signal level L with respect to the length of the detection section. That is, as shown in FIG. 7, when the gradient angle when the signal level L changes by δL with respect to one detection section D is defined as α, tanα indicating the light-dark gradient is expressed by the following relational expression: tanα = δL / It will be calculated based on D ...

Therefore, when the light / dark data is delayed by two sections, the light / dark gradient is calculated by the following equation: tan α = δL '/ 2D. It should be noted that δL ′ in the above equation represents the amount of change in the signal level L in the two sections.

Next, the host computer 8 makes a step S.
The optimum determination value for defect detection is determined in step 8. In this embodiment, the optimum determination value is determined by fuzzy inference.

That is, in the memory of the host computer 8, as shown in FIG. 8, the rising height FH of the defect indicated by the video signal, the falling height DH of the defect, the height difference PH of the defect, and the change width W of the defect. And a plurality of determination rules indicating the setting tendency of the defect detection upper limit level UL and the defect detection lower limit level DL that are preset according to the gradient angle α of the light-dark gradient.

Regarding the rising height FH of the defect, which is one of the antecedents constituting the above decision rule, three triangular membership functions are set as shown in FIG. 9 (a). That is, when the rising height FH is small, the function SM on the left side is normal, and the function MM in the center is normal,
When it is a little larger, the right function ML is selected.

Similarly, the trailing edge height DH of the defect,
Membership functions as shown in FIGS. 9B to 9E are set for the defect height difference PH, the defect change width W, and the gradient angle α of the light-dark gradient.

Here, the symbols representing the membership functions will be summarized and shown in Table 1 below.

[0036]

[Table 1] On the other hand, for the defect detection upper limit level UL which is the consequent part, the membership function is set as shown in FIG. 10, and the defect detection lower limit level DL is also the same as the defect detection upper limit level UL as shown in FIG. Membership functions are set.

As the decision rule for connecting the antecedent part and the consequent part, 16 kinds of decision rules are set as shown in Table 2 below.

[0038]

[Table 2] In Table 2 above, for example, when the defect rising height FH belongs to the function SM and the change width W of the defect belongs to the function GT, the decision rule with the rule number 1 is, for example, the defect detection upper limit level UL. It is shown that the function SA in FIG. 10 is selected as the membership function of.

That is, the rising height F indicated by the video signal
When H indicates FH ₁ as shown in FIG. 9A,
The function SM is adopted as the membership function, and 0.8 corresponding to the above FH ₁ is obtained as the matching value (grade).

Similarly, when the falling height DH indicates DH ₁ as shown in FIG. 9B, the function MM is adopted as the membership function and the adaptation value corresponds to the above DH ₁ . 0.2 is obtained, and in FIG.
When the height difference PH indicates PH ₁ as shown in (c), the function LA is adopted as the membership function, and 0.3 as the matching value is obtained, which corresponds to PH ₁ above. When the change width W indicates W ₁ as shown in ( ₁ ), the function GT is adopted as the membership function and 0.5 corresponding to the above W ₁ is obtained as its matching value. Further, as shown in FIG. 7E, when the gradient angle α of the light-dark gradient indicates α ₁ , the function MM is adopted as the membership function and
As the matching value, 0.3 corresponding to the above α ₁ is obtained.

In this case, since the function SM is adopted as the membership function of the rising height FH as described above, the function SA is selected according to the decision rule of rule number 1 from the relation of Table 2 and , The function SA is shown by the hatched area in FIG.
The portion truncated at 0.5, which is the smaller matching value of H and the variation width W, is adopted as the membership function.

Similarly, from the relationship between the falling height DH and the change width W, the function SM is selected according to the determination rule of rule number 5 in Table 2, and is shown by the shaded area in FIG. 12 (b). In this way, the part that is truncated at 0.2, which is the smaller matching value of both, is adopted as the membership function,
Further, from the relationship between the height difference PH and the change width W, the function LA is selected according to the determination rule of rule number 10 in Table 2, and the matching value of both is small as shown by the shaded area in FIG. The part truncated at 0.3 is adopted as the membership function, and the function MM is selected from the relationship between the gradient angle α and the change width W according to the determination rule of rule number 13 in Table 2, and As shown by the shaded area in FIG. 12D, the smaller portion of both matching values, which is truncated at 0.3, is adopted as the membership function.

Then, the host computer 8 operates as shown in FIG.
As shown by the shaded area in FIG. 3, the union of the membership functions is set as the final membership function of the defect detection upper limit level UL, and the value of the defect detection upper limit level UL corresponding to the barycentric position P ₁ of the membership function. Is determined as the optimum defect detection upper limit level UL ₁ .

The optimum defect detection lower limit level DL ₁ is also determined in the same manner.

Having determined the optimum defect detection upper limit level UL ₁ and the optimum defect detection lower limit level DL ₁ , the host computer 8 executes step S9 of the flow chart of FIG. 6, and the optimum defect detection upper limit level UL ₁ and these optimum defect detection upper limit levels UL ₁ and A defect detection calculation is performed by comparing the defect detection lower limit level DL ₁ with the delayed light / dark data, and it is determined in step S10 whether or not there is a defect, and when YES is determined, in steps S11 and S12. The type and size of the defect are measured and the data is stored in a predetermined area of the memory. That is, as shown in FIG. 14, for example, the host computer 8 re-evaluates the defect size, for example, the rising height FH, the falling height DH, and the change width W with reference to the optimum defect detection upper limit level UL _1. Therefore, it is determined whether or not the defect is present.

Then, the host computer 8 determines in step S13 whether or not the horizontal scanning has been performed up to the end. When the determination is YES, the host computer 8 updates the horizontal scanning by one line in step S14, and in step S15 the received light image. The above steps are executed until it is determined that one screen of 15 is finished.

Therefore, as shown in FIG. 7, the defect detection upper limit level UL and the defect detection lower limit level DL that appropriately reflect the video signal output from the CCD camera 6 can be obtained.

Next, an embodiment of the surface defect detection process using the neural network will be described.

That is, the host computer 8 sets the minimum judgment value for defect detection as in the case of the above fuzzy reasoning, and then sets the light and dark data detection section D as shown in FIG. The brightness data corresponding to the detection section D is read. Then, the read light / dark data is delayed, the defect is detected by comparing the defect detection determination value with the light / dark data, and the light / dark gradient is calculated when the light / dark data is delayed by two sections. Even in this case, the light / dark gradient is obtained as the amount of change δL of the signal level L per detection section D.

Next, the host computer 8 determines the optimum determination value for defect detection, that is, the defect detection upper limit level UL and the defect detection lower limit level DL, for example, with reference to FIG.
As shown in, the height difference H of the defect, the change width W of the defect,
Obtains a slope angle of the brightness gradient alpha, these values, as shown in FIG. 17, respectively input as the input signal ^{_{I 1 1, I 1 2,}} I 1 3 to the input layer of the neural network.

[0051] The neural network, which has an intermediate layer connected to the input layer in which the input signal ^{_{I 1 1, I 1 2,}} I 1 3 is input, and an intermediate layer connected to the output layer, From the output layer, output signals indicating the optimum defect detection upper limit level UL ₁ and the optimum defect detection lower limit level DL ₁ are respectively output.

Here, the characteristic parameters u ⁱ _j (i = 1 to 3; j = 1 to 3) of this neural network, θ
ⁱ _j (i = 1 to 3; j = 1 to 3), w ¹² _ij (i = 1 to 3;
j = 1 to 3) in advance the teaching data (input signal I ¹ _1, I
¹ _2, I ¹ ₃₎ as the optimum defect detection limit level UL ₁ as an output signal O ^{₁ 1,} O ¹ ₂ when the input and the optimum defect detecting lower limit level DL ₁ is output, for example backpropagation It is decided to learn using.

Therefore, the input signals I ¹ ₁ ,
As I ¹ ₂ and I ¹ ₃ , the height difference H of the defect and the change width W of the defect
When the gradient angle α of the light-dark gradient is input, the optimum defect detection upper limit level UL ₁ and the optimum defect detection lower limit level DL ₁ corresponding thereto are output from the output layer.

As a result, the host computer 8 detects the optimum defect detection upper limit level UL as shown in FIG.
_With reference to ₁ , the size of the defect, for example, the change width W is re-evaluated, and whether or not the defect is present is determined accordingly.

Therefore, also in this embodiment, FIG.
As shown in FIG. 5, the defect detection upper limit level UL and the defect detection lower limit level DL that appropriately reflect the video signal output from the CCD camera 6 are obtained.

Next, another embodiment of the light source to which the present invention is applied will be described with reference to FIG. 19. This light source 5'has one side face opened, as in the embodiment shown in FIG. A plurality of fluorescent lamps (not limited to fluorescent lamps) 12 '... 12' provided in the box 11 'and one side surface of the box 11' provided on the front surface of these fluorescent lamps 12 '. Blocking optical filter 13 'and diffusing screen 14'
It consists of and. And in this embodiment,
The optical filter 13 'has different light transmittance depending on the transmitting place so that the light emitted by each fluorescent lamp 12' can be converted into bright and dark light in which the brightness is repeatedly changed from dark to bright. Is configured. That is, in the present embodiment, the transmittance is repeatedly changed from small to large only in the x direction of the x and y directions orthogonal to each other shown in FIG.
20. As shown in FIG. 20, due to ′, bright and dark light is formed by repeating a change from dark to bright at a predetermined gradient along the x direction, and the bright and dark light is applied to the coating surface 1a of the vehicle body 1. It has become so. That is, as shown in FIG.
The relatively wide coating film surface 1a ′ covering the field of view F ′ of the CCD camera 6 ′ is irradiated with the bright / dark light 5a ′ whose brightness repeatedly changes from dark to bright in the x direction. Therefore, a light irradiation region in which the change from dark to bright is repeated is formed on the coating surface 1a '. Then, as shown in FIG. 19, in the received light image 15 ′ of the CCD camera 6 ′ that receives the reflected light from the light irradiation area, the bright and dark light 5 a ′.
The lightness repeatedly changes in the direction indicated by the arrow X corresponding to the direction x of change from dark to light. Note that FIG.
In FIG. 2, reference numeral X ₁ denotes a dark-to-light transition region in the bright-dark light 5a ′ in which the dark-to-light transition is repeated, and the line density is rough in each region X ₁ . The higher the lightness is, the denser the line is, the smaller the lightness is.

Therefore, for example, as shown in FIG.
If there is a convex defect z ′ in the light irradiation area of the coating film surface 1a ′ of the vehicle body 1 ′, a convex defect portion Z ′ having a brightness different from that of the surroundings is generated at the corresponding position in the light receiving image 15 ′. Become.

The present invention can also be applied to the case where the brightness of the light and dark light radiated on the coating surface 1a 'of the vehicle body 1'changes repeatedly in this way. In that case, the optimum defect detection determination value corresponding to each change region X ₁ ... X ₁ in the light-receiving image 15 ′ is automatically learned and set.

[0059]

As described above, according to the surface state detecting method of the present invention, since the defect detection determination value is changed by the light-dark gradient, for example, the surface inspection can be performed accurately even if the light-dark gradient changes. It will be well done.

According to the second aspect of the present invention, the defect detection determination value is changed depending on the brightness / darkness gradient and the size of the defect, so that it is not affected by the disturbance caused by the curvature of the surface to be inspected. The surface inspection will be performed with high accuracy.

Further, according to the third and fourth aspects of the present invention, since it is sufficient to store a simple decision rule indicating the tendency of setting the defect detection judgment value with respect to the brightness gradient, the memory capacity is saved, and at the same time, It is possible to set the determination value with higher accuracy than in the case where the defect detection determination value is approximated by using a regression equation represented by a straight line or the like.

Further, according to the fifth aspect, even if the gradient of the bright and dark light changes, an appropriate defect detection judgment value is set, and the detection accuracy is improved.

[Brief description of drawings]

FIG. 1 is a perspective view of a surface state inspection device.

FIG. 2 is an exploded perspective view of a light source.

FIG. 3 is a graph showing a pattern of light and dark gradients.

FIG. 4 is an enlarged view showing an irradiation state by a light source and a light receiving state by a CCD camera.

FIG. 5 is a partially enlarged view of a received light image.

FIG. 6 is a flowchart showing a surface state detection process.

FIG. 7 is a graph showing changes in the signal level of the light / dark data input to the host computer.

FIG. 8 is a graph showing a change in signal level of a defective portion before learning.

FIG. 9 is a graph showing a setting example of a membership function of the antecedent part used in fuzzy inference.

FIG. 10 is a graph showing a setting example of a membership function of a defect detection upper limit level.

FIG. 11 is a graph showing a setting example of a membership function of a defect detection lower limit level.

FIG. 12 is a schematic diagram showing a process of inferring a defect detection upper limit level.

FIG. 13 is a schematic diagram showing the final process of the inference calculation of the defect detection upper limit level.

FIG. 14 is a graph showing a change in signal level of a defective portion after learning.

FIG. 15 is a graph showing a change in signal level of light / dark data input to a host computer when a neural network is used.

FIG. 16 is a graph showing a change in signal level of a defective portion before learning.

FIG. 17 is a schematic diagram of a neural network.

FIG. 18 is a graph showing a change in signal level of a defective portion after learning and before learning.

FIG. 19 is an exploded perspective view of a light source according to another embodiment.

FIG. 20 is a graph showing a pattern of light and dark gradients in the example.

FIG. 21 is an enlarged view showing an irradiation state by a light source and a light receiving state by a CCD camera.

FIG. 22 is a partially enlarged view of a received light image.

FIG. 23 is an explanatory diagram of a problem of the conventional example.

[Explanation of reference numerals] 1 vehicle body 1a coating film surface 2 surface state inspection device 5 light source 6 CCD camera 8 host computer 10 image processor 15 light-receiving image z convex defect

Claims (5)

[Claims] 1. A light-dark image whose brightness gradually changes along a predetermined direction is applied to a surface to be inspected, a reflected light from the surface to be inspected is captured to create a light-receiving image, and this light-receiving image is changed to light-dark. A surface state inspection method for inspecting the surface state on the surface to be inspected by comparing the detected light and dark data with a predetermined defect detection determination value,
A surface state inspection method using bright / dark illumination, wherein the defect detection determination value is changed according to the light / dark data obtained by scanning the light-receiving image. 2. A surface to be inspected is irradiated with light and dark light whose brightness gradually changes along a predetermined direction, a reflected light from the surface to be inspected is captured to create a light reception image, and this light reception image is changed to light and dark. A surface state inspection method for inspecting the surface state on the surface to be inspected by comparing the detected light and dark data with a predetermined defect detection determination value,
Surface condition inspection by light / dark illumination, characterized in that the defect detection determination value is changed in accordance with the light / dark gradient calculated from the light / dark data obtained by scanning the light-receiving image and the size of the defect calculated from the data. Method. 3. A light-dark image whose brightness gradually changes along a predetermined direction is applied to a surface to be inspected, a reflected light from the surface to be inspected is captured to create a light-receiving image, and this light-receiving image is changed to light-dark. A surface state inspection method for inspecting the surface state on the surface to be inspected by comparing the detected light and dark data with a predetermined defect detection determination value,
Setting a decision rule in which the setting tendency of the defect detection determination value for the light and dark data is set in advance, and the defect detection determination value based on the light and dark data obtained by scanning the light-receiving image and the determination rule. A method for inspecting a surface condition by bright and dark illumination, which comprises calculating a degree of conformity and changing the defect detection determination value according to the degree of conformity. 4. A light-dark image whose brightness is gradually changed along a predetermined direction is applied to a surface to be inspected, a reflected light from the surface to be inspected is captured to create a light-receiving image, and the light-receiving image is changed to light-dark. A surface state inspection method for inspecting the surface state on the surface to be inspected by comparing the detected light and dark data with a predetermined defect detection determination value,
The decision rule in which the tendency of setting the determination value for defect detection is determined in advance corresponding to the light-dark gradient and the size of the defect is set in advance,
The degree of conformity of the defect detection determination value is calculated based on the lightness / darkness gradient calculated from the lightness / darkness data obtained by scanning the light-receiving image, the size of the defect calculated from the data, and the determination rule, and the degree of conformity is calculated as follows. A method for inspecting a surface condition by means of bright and dark illumination, characterized in that the judgment value for defect detection is changed accordingly. 5. A light-dark image whose lightness is repeatedly changed along a predetermined direction is applied to a surface to be inspected, a reflected light from the surface to be inspected is captured to create a light-receiving image, and the light-receiving image is changed to light-dark. Is a surface state inspection method for inspecting the surface state of the surface to be inspected by comparing the detected light and dark data with a predetermined judgment value for defect detection, which is obtained by scanning the received light image. A method for inspecting a surface condition by bright and dark illumination, characterized in that the judgment value for defect detection is changed according to the bright and dark data.