JP2012058091A - Surface inspection device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To improve angle detection sensitivity and to widen a detectable angle range.SOLUTION: A zonal surface light source is configured so that chevron-shaped luminance distributions of mutually different wavelength regions overlap with each other, and the angle of an inspection surface is detected in two stages on the basis of a pixel value distribution corresponding to the chevron-shaped luminance distribution of a photographed image of each wavelength region. In the first stage, on the basis of a size relation among the pixel values of R, G and B and curve approximation information, an in-cycle reference light position x on a light source surface is determined. In the second stage, on the basis of the value of lightness and the curve approximation information, to which cycle i the in-cycle reference light position x belongs, that is, a cycle start point position Xi, is obtained. Then, the reference light position X=Xi+x is converted to a detection angle on an inspection line.

Description

  The present invention relates to a surface inspection apparatus that detects the angle of an inspection surface using regular reflection of light.

  As shown in FIGS. 8 and 9, the illumination device 20 and the imaging device 30 are attached to a robot arm (not shown) in order to detect minute irregularities on the inspection surface 10, for example, a painted surface of a car. While the illumination device 20 and the imaging device 30 are integrally moved relative to the inspection surface 10 in the direction of the arrow shown in the figure, the imaging device 30 captures regular reflection light on the inspection surface 10 of light emitted from the illumination device 20. To do. When the regular reflection light on the inclined surface of the concavo-convex portion 11 on the inspection surface 10 is imaged, the light emission point (reference light point) on the light source surface 21 of the illuminating device 20 corresponding to the pixel is this It changes according to the angle change of the inclined surface.

  Therefore, in Patent Document 1 below, the luminance distribution in the range from one end A to the other end B in the scanning direction of the light source surface 21 is given a gradient as shown in FIG. Corresponding to the change of the pixel value, the coating defect is detected based on the change of the pixel value.

  If the inclination angle of the concavo-convex portion 11 is relatively large, it is determined as a severe defect. On the other hand, even if the inclination angle is relatively small, it is necessary to determine that the defect is a severe defect if the area of the defect portion is large.

  However, when the tilt angle is small, the positional change amount of the reference light spot is small, so the SN ratio becomes small and a defect cannot be detected (detection sensitivity is low). When the brightness gradient is increased, the width between AB corresponding to the detectable angle range becomes narrow, and depending on the angle of the detection target, the detection sensitivity may be lowered or overlooked.

  In order to solve such a problem, in Patent Document 2 below, a slit is arranged on the light source surface 21 and the luminance distribution between AB is shown in FIG. The brightness gradient is increased. Since a defect cannot be detected in the dark part, two imaging devices 30 are arranged side by side so that the dark part viewed from one imaging apparatus 30 becomes the bright part viewed from the other imaging apparatus 30.

  However, this condition is met in some areas, resulting in blind spots. Also, the brightness distribution is flat in the bright area, and defects cannot be detected by the above principle.

  On the other hand, in Patent Document 3 below, an illumination device 20 having a luminance distribution of R (red), B (blue), and G (green) as shown in FIG. The apparatus 20 is arranged in parallel, the imaging apparatus 30 is arranged so that the optical axis thereof is perpendicular to the inspection surface 10, the B regular reflection light quantity at the upper part of the concavo-convex part 11, and the R at one end side of this upper part. The regular reflection light amount and the G regular reflection light amount on the other end side are read from the image, and the type of defect such as a concave-convex defect or a foreign matter is determined based on the combination of the magnitudes of these regular reflection light amounts.

  However, it is not possible to detect how much the inclination angle in the concavo-convex defect is.

  In Patent Document 4 below, a light whose wavelength gradually changes from red to blue is used in an illumination device in which the change in wavelength occurs along the direction in which the incident angle changes. From the photographed image of the reflected light, the color of the soldered portion is obtained by calculation, and based on this, the suitability of the solder surface state is determined.

  However, since the light whose wavelength gradually changes from red to blue is used, the position change amount of the reference light spot is small in the case of a small inclination angle. For this reason, the SN ratio is small, and the same problem as in the cited document 1 occurs.

Japanese Patent Laid-Open No. 5-209734 Japanese Patent Laid-Open No. 11-63959 JP 2010-112941 A JP 2009-128303 A

  In view of the above problems, an object of the present invention is to provide a surface inspection apparatus capable of improving angle detection sensitivity.

  Another object of the present invention is to provide a surface inspection apparatus capable of reducing the unevenness of the angle detection sensitivity while improving the angle detection sensitivity.

  Still another object of the present invention is to provide a surface inspection apparatus capable of improving the angle detection sensitivity and widening the detectable angle range.

  It is still another object of the present invention to provide a surface inspection apparatus that can improve the angle detection sensitivity, reduce the unevenness of the angle detection sensitivity, and widen the detectable angle range.

In the first aspect of the present invention,
Illumination means in which a plurality of strip light sources having different wavelength ranges are arranged in parallel,
Spectroscopic means for splitting the light emitted from the illumination means and regularly reflected by the light inspection surface into each of the plurality of wavelength ranges;
Imaging means comprising an image sensor on which each of the split light is imaged;
Image processing means for detecting an angle of the inspection surface based on data of an image picked up by the image pickup means;
A surface inspection apparatus comprising:
The illumination means has a luminance distribution with respect to a reference light position along a line perpendicular to the longitudinal direction of the strip light source on the light source surface of the plurality of strip light sources. The brightness decreases as the distance from the point decreases, and the brightness distribution of adjacent strip light sources is configured to overlap at least to the peak position of the brightness,
The image processing means, based on a combination of the wavelength range of the maximum pixel value and the wavelength range of the second largest pixel value, for pixel values corresponding to each of the plurality of wavelength ranges corresponding to the same reference light position A range including the reference light position is determined, a reference light position within the range is determined based on the maximum pixel value or based on the maximum pixel value and the second largest pixel value, and the reference light First processing means for obtaining the angle based on the relationship between the position and the angle is provided.

  In a second aspect of the surface inspection apparatus according to the present invention, in the first aspect, the first processing means of the image processing means is based on a ratio value between the second largest pixel value and the maximum pixel value. The reference beam position within the range is determined.

  In a third aspect of the surface inspection apparatus according to the present invention, in the second aspect, the first processing means of the image processing means does not depend on lightness corresponding to each of the plurality of wavelength ranges with respect to a reference light position. The distribution of pixel values normalized to a predetermined value from each normalized pixel value so that the normalized pixel value in a wavelength region different from the wavelength region of the peak pixel value at each peak pixel value is substantially zero. Subtraction is performed as an offset value, and after the subtraction, a reference light position within the range is determined based on the value of the ratio.

In the fourth aspect of the surface inspection apparatus according to the present invention, in the third aspect, the correction value multiplied by the ratio value so that the reference light position from the reference position is proportional to the ratio value does not depend on the lightness. Further comprising storage means for storing information for obtaining approximately as a function of the maximum pixel value normalized to or a function of the ratio,
The first processing unit of the image processing unit determines the reference light position by multiplying the ratio value by the correction value.

In a fifth aspect of the surface inspection apparatus according to the present invention, in the first aspect, the pixel value distribution with respect to the amount corresponding to the luminance distribution of each of the plurality of wavelength regions of the image captured by the imaging unit, First information for approximating the distribution of the maximum pixel value of each of the plurality of wavelength ranges corresponding to the same reference light position, normalized so as not to depend on lightness, is stored. A first approximate expression information storage means,
The first processing means of the image processing means calculates a reference light position corresponding to the normalized maximum pixel value within the range in the approximate expression of the normalized pixel value distribution of the maximum pixel value based on the first information. Ask.

In a sixth aspect of the surface inspection apparatus according to the present invention, in any one of the second to fifth aspects, the illuminating unit includes a plurality of strip-shaped light sources having different wavelength ranges arranged in parallel. The luminance distribution is different, and the ratio of the pixel value of the image to the in-group reference light position corresponding to each other for the plurality of sets is substantially constant regardless of the in-group reference light position,
Lightness distribution of the sets adjacent to each other with respect to the lightness, or approximately the lightness distribution based on the respective pixel values of the plurality of wavelength ranges corresponding to the same light reference position, A second approximate expression information storage means storing second information for approximately obtaining a boundary line for distinguishing between
The image processing means normalizes the pixel values corresponding to each of the plurality of wavelength ranges corresponding to the same reference light position so as not to depend on the set, and the first processing means performs the reference light position in the set. Seeking
The image processing means further determines which set of reference light positions in the set is based on the brightness distribution or the approximate expression of the boundary line based on the second information and the reference light position in the set. And second processing means for determining the reference light position in the entire range based on the reference light position in the set and the determined set.

  According to the configuration of the first aspect, the luminance distribution with respect to the reference light position along the line perpendicular to the longitudinal direction of the strip light source on the light source surface of the plurality of strip light sources is strip-shaped for each of the plurality of strip light sources. Each of the plurality of wavelength ranges corresponding to the same reference light position is configured such that the brightness decreases as it moves away from the midpoint of the light source surface, and the brightness distribution of the adjacent strip light sources overlaps at least to the peak position of the brightness. Is determined based on the combination of the wavelength range of the maximum pixel value and the wavelength range of the second largest pixel value, and the range including the reference light position is determined based on the maximum pixel value or the maximum pixel value. The reference light position within the range is determined based on the pixel value and the second largest pixel value, and the angle is obtained based on the relationship between the reference light position and the angle. The An effect that it is possible to above.

  According to the configuration of the second aspect, the reference light position in the range is determined based on the ratio value between the second largest pixel value and the maximum pixel value. With respect to the range, there is an effect that it is possible to reduce the unevenness of the angle detection sensitivity while improving the angle detection sensitivity.

  According to the configuration of the third aspect, with respect to the distribution of standardized pixel values corresponding to each of the plurality of wavelength regions with respect to the reference light position, the wavelength different from the wavelength region of the peak pixel value at each peak pixel value A predetermined value is subtracted from each normalized pixel value as an offset value so that the pixel value of the area becomes substantially zero, and after the subtraction, the reference light position within the range is determined based on the value of the ratio Therefore, it is possible to reduce the angle detection error by removing a signal having a low S / N ratio, and it is possible to simplify the calculation and improve the image processing speed.

  According to the configuration of the fourth aspect, since the reference light position is determined by multiplying the value of the ratio by the correction value, there is an effect that unevenness in angle detection sensitivity can be further reduced.

  According to the configuration of the fifth aspect, the first information for approximately obtaining the normalized pixel value distribution of the maximum pixel value among the pixel values of the plurality of wavelength ranges corresponding to the same reference light position. The reference light position corresponding to the standardized maximum pixel value within the range in the approximate expression of the standardized pixel value distribution of the maximum pixel value based on is obtained, so that the angle can be detected with a higher SN ratio, As a result, the angle detection sensitivity can be improved.

  According to the configuration of the sixth aspect, a plurality of band-shaped light sources having different wavelength ranges are arranged in parallel, the luminance distribution is different for each group, and the plurality of sets correspond to the in-group reference light positions corresponding to each other. Illumination means in which a ratio of pixel values of the image is substantially constant regardless of the reference light position in the set, and pixel values corresponding to each of the plurality of wavelength ranges corresponding to the same reference light position are assigned to the set The reference light position in the set is obtained by normalization so as not to depend on the first processing means, and based on the distribution of brightness or an approximate expression of the boundary line based on the second information and the reference light position in the set Since the reference light position in the set is determined and the reference light position in the entire range is obtained based on the reference light position in the set and the determined set, the angle detection sensitivity is improved. And the detectable angle range can be widened Achieve the results.

  Other objects, characteristic configurations and effects of the present invention will become apparent from the following description read in connection with the appended claims and the drawings.

It is a schematic block diagram of the surface inspection apparatus which concerns on Example 1 of this invention. It is a schematic perspective view of an illumination imaging apparatus mainly showing a schematic configuration of the imaging apparatus. (A) is a side view which shows schematic structure of an illuminating device, (B) is a front view of the LED array part in (A). FIG. 4 is a connection diagram of arbitrary one color LED row in FIG. 3. (A) is a diagram showing the luminance Y of each of R, G, and B with respect to the reference light position X along the scanning direction on the light source surface, and (B) shows the reference light position X by changing the angle of the inspection surface. 2 is a diagram showing changes in pixel values of specific points of line sensors corresponding to R, G, and B of the imaging apparatus when A is changed between AB in FIG. 1, and (C) is data of (B) It is a diagram which shows the brightness with respect to the reference light position X based on this. (A) is a diagram explaining a method for determining a reference light position x within a period based on a pixel value, (B) is a diagram obtained by normalizing the pixel value of (A), and (C) is a diagram. It is a diagram of the maximum pixel value among R, B, and G corresponding to the reference light position x in each cycle in (B), and (D) shows the brightness V in each cycle with respect to the reference light position x in the cycle. FIG. 2 is a schematic flowchart of processing for obtaining an inclination angle of an arbitrary point on an inspection line from R, G, and B pixel values, which is executed by the inspection point angle detection unit of FIG. 1. It is explanatory drawing which shows a mode that the reference light position X on the light source surface corresponding to a picked-up image changes, when an illuminating device and an imaging device are integrally moved relatively to the illustration arrow direction with respect to a test | inspection surface. . It is explanatory drawing which shows a mode that the reference light position X on the light source surface corresponding to a picked-up image changes, when an illuminating device and an imaging device are integrally moved relatively to the illustration arrow direction with respect to a test | inspection surface. . (A) is a side view of the uneven portion on the inspection surface, and (B) is an explanatory view of the angle distribution stored in the angle information storage portion of FIG. 1 when the uneven portion is scanned as shown in FIGS. is there. (A) is a side view showing a schematic configuration of a lighting device, (B) is a characteristic diagram showing transmittance between AB of an ND filter in (A), and (C) is a reference relating to Example 2 of the present invention. It is a diagram which shows the change of the pixel value of each specific point of the line sensor corresponding to R, B, and G when changing the light position X between AB of (A), (C) is a light shaping diffuser plate It is a diagram which shows the brightness with respect to the upper reference light position X. FIG. FIG. 4A is an explanatory diagram of offset subtraction according to a third embodiment of the present invention, and FIG. 5B is a diagram illustrating a pixel value distribution with respect to an intra-period reference light position x after subtracting the offset of (A) from each pixel. , (C) and (D) are approximate calculation explanatory diagrams. (A)-(C) are all explanatory drawing of the problem of background art, Comprising: It is a diagram which shows the luminance distribution with respect to the reference light position X of an illuminating device.

  FIG. 1 is a schematic configuration diagram showing an embodiment of a surface inspection apparatus according to the present invention.

  The inspection surface 10 is, for example, a painted surface of a car, and an illumination imaging device LID provided with an illumination device 20 and an imaging device 30 is attached to a robot arm (not shown) in order to detect the fine unevenness defect. . While the illuminating device 20 and the imaging device 30 are integrally moved relative to the inspection surface 10 in the direction of the arrow shown in FIG. Take an image.

  FIG. 3A is a side view illustrating a schematic configuration of the lighting device 20. FIG. 3B is a front view of the LED array portion constituting the illumination device 20.

  In the lighting device 20, a plurality of LED lamps 23 are arranged in a staggered pattern on a substrate 22 that is a rectangular heat conductive good conductor so as to narrow the space between rows. Each LED lamp 23 includes a small substrate 230 fixed to the substrate 22, an LED 231 fixed on the small substrate 230, and terminal portions respectively connected to the anode and the cathode of the LED 231. In the LED array, R (red), G (green), and B (blue) in FIG. 3, monochromatic R rows, G rows, and B rows having relatively narrow wavelength ranges are periodically arranged in this order. The LED array unit is configured. All R rows are provided with LEDs having the same characteristics except for manufacturing variations, and the same applies to the G row and the B row.

  Each of the three-color LED rows 240 to 246 includes a pair of R row, G row, and B row that are close to each other. A transparent substrate 25 is juxtaposed facing the LED array portion, and a sheet-shaped light shaping diffusion plate 26 is bonded onto the transparent substrate 25. Light shielding plates 270 and 271 are bonded to both ends in the scanning direction of the LED side surface of the transparent substrate 25 so as to shield each half of the R row at one end and the B row at the other end of the LED array. . Positions A and B in FIG. 3 correspond to positions A and B in FIGS. 1, 2, 8, and 9, respectively.

  As shown in FIG. 4, adjacent LED 231 is connected in series to each LED row, and a variable resistor 28 is further connected in series to this, and a predetermined constant voltage is applied to this series connection circuit, and the resistance of the variable resistor 28 is The light emission amount of the LED array is adjusted by the value. The brightness adjusting device 28G in FIG. 1 includes a variable resistor 28 for each LED row.

  The light shaping diffuser 26 is for diffusing and shaping light by its refraction function. For example, a surface, a relief, or a hologram pattern is used to randomly form a microlens on the surface. Is. The light shaping diffusion plate 26 shapes the light so that the diffusion angle in the direction of the LED row of the same color is sufficiently larger than that in the direction perpendicular thereto, and is parallel to the LED row on the light shaping diffusion plate 26. The luminance is made substantially uniform on a line in a different direction. Thereby, a strip-shaped light source is formed on the surface of the light shaping diffusion plate 26 corresponding to each LED row. Here, the diffusion angle is a full angle at which the luminance is half of the maximum value, and depends on the emission angle of the LED.

  The illumination device 20 has a luminance distribution with respect to the reference light position X along a line perpendicular to the longitudinal direction of the strip light source on the light source surface of the plurality of strip light sources, that is, a luminance distribution between ABs. For each, the luminance decreases as the distance from the middle point of the band-shaped light source surface increases, and the luminance distribution of the adjacent band-shaped light sources overlaps at least to the peak position of the luminance. ) As shown.

  Regarding the luminance peak value, the luminance adjusting device 28G is adjusted so as to change stepwise for each of the three-color LED rows. In FIG. 5A, R0, G0, and B0 are the luminance distribution curves of the band-like surface light sources corresponding to the R row, G row, and B row of the three-color LED row 240, respectively, and R1, G1, and B1 are each three colors. It is a luminance distribution of the strip-shaped surface light source corresponding to the R row, G row, and B row of the LED row 241.

  Next, a schematic configuration of the imaging device 30 will be described with reference to FIG.

  In the imaging device 30, the light reflected by the inspection line 12 on the inspection surface 10 is split into R, B, and G light by the dichroic prism 301 via the imaging lens 300, and the light is reflected on the line sensors 30 R, 30 G, and 30 B, respectively. Is imaged. The position (reference light position) of the incident light corresponding to the regular reflection at the inspection line 12 in the scanning direction AB of the light source surface 21 shown in FIG. 1 is determined by the angle at the inspection line 12 on the inspection surface 10. For example, the reference light spots RP0 to RP2 on the light source surface 21 correspond to the tangent planes S0 to S2 of the uneven portion at the position of the inspection line 12. The center point pixel values of the line sensors 30R, 30G, and 30B corresponding to, for example, the center line between AB on the light source surface 21 when the angle of the tangent plane is continuously changed in one direction are shown in FIG. It changes as shown in (B).

  DR0, DG0, DB0, DR1, DG1, and DB1 in FIG. 5B correspond to R0, G0, B0, R1, G1, and B1 in FIG. 5A, respectively. The horizontal axis in FIG. 5B is the reference light position X between AB in FIG. 1 as in FIG.

  Actually, the pixel value corresponding to the reference light position X differs depending on parameters such as the luminance depending on the emission angle from the light source surface 21, the reflection angle on the inspection line 12, and the reflectance depending on the wavelength. As a result, the brightness adjusting device 28G is adjusted so that the relationship as shown in FIG. FIG. 5A shows the above parameters ignored for simplicity of explanation.

  A small black dot in FIG. 5B indicates an arbitrary point in which the reference light positions x in one cycle are equal to each other. The brightness adjusting device 28G is adjusted so that these points are on a straight line indicated by a dotted line (the actual adjustment result is that these points should be on a substantially straight line within an allowable range. ). Utilizing such regular periodicity, the calculation of the reference beam position X is simplified as will be described later.

  The reason why the light shielding plates 270 and 271 are provided in FIG. 3A is that the detection error of the reference light position x is reduced and the periodicity is ensured by keeping the SN ratio equal to or higher than the predetermined value even in the end region. This is to simplify the processing. FIG. 6A shows one cycle in FIG.

  Next, the principle of determining the reference light position in this embodiment will be described. In this embodiment, the reference light position X is obtained in two stages. In the first stage, the reference light position x in one cycle shown in FIG. 6A is obtained, and in the second stage, the reference light position x in the period is determined by which period i of the pixel brightness distribution shown in FIG. (I is an integer value in the range of 0 to 4 shown in FIG. 5C) is determined, and Xi + x obtained by adding the intra-period reference light position x to the start position Xi of the period i is referred to as the reference light position. Find as X. That is, the reference light position X is determined hierarchically. Here, the pixel brightness V is set to V = DR + DG + DB for simplification. The brightness V may be obtained by weighting the R, G, and B pixel values in accordance with the relative sensitivities of the line sensors 30R, 30B, and 30G.

1). First Stage Of the set of R, B, and G pixel values corresponding to the reference light position x in one cycle, the SN ratio increases as the pixel value increases. In order to reduce the detection error, it is desired to obtain the in-period reference light position x using only this maximum pixel value. In FIG. 6A, for example, when the B pixel value is maximum and this value is D0, there are two reference light positions x in the period where the B pixel value becomes D0. If the second largest pixel value of a set of pixel values of R, B, and G is G, the in-period reference light position x corresponding to the pixel value on the uphill curve is obtained, and if R, R is on the down curve. An intra-period reference light position x corresponding to the pixel value is obtained.

  If this is generalized, as shown in FIG. 6A, the pixel value intersection point C0 between the peak points P0, P1, P2, and P3 of the R, G, and B pixel values and the adjacent peak points. , C1, and C2 are divided into six regions having a boundary point as one boundary. Then, a plurality of pixel values for approximating the curve of the maximum pixel value in the RGB1 set as shown in FIG. 6C with a curve, a broken line or a straight line are measured in advance, and are associated with the reference light position x in the period. The in-period reference light position x is obtained as follows.

  FIG. 6C shows the R, G, and B pixel values DR, DG, and R corresponding to the intra-period reference light position x in order to obtain the intra-period reference light position x uniformly regardless of the period. DB is standardized. Various standardizations are conceivable, but in this embodiment, these DR, DG, and DB are standardized so as not to depend on lightness, such as r = DR / V, g = DG / V, and b = DB / V, respectively.

  (1) If the maximum pixel value is R, it is determined that the region is A0 or A6. If the second largest pixel value is G, the region A0, that is, a point on the curve P0-C0, is regarded as a point on the cycle. The position x is obtained, and if the second largest pixel value is B, the in-period reference light position x is obtained assuming that the point is on the area A6, that is, the curve C2-P3.

  (2) If the maximum pixel value is G, it is determined that the region is A1 or A2, and if the second largest pixel value is R, the region A1, that is, a point on the curve C0-P1, is assumed to be a point on the cycle. The position x is obtained, and if the second largest pixel value is B, the intra-period reference light position x is obtained assuming that the point is on the area A2, that is, the curve P1-C1.

  The same applies to the following, and can be easily understood from the above description, so the description thereof is omitted.

  Note that the heights of the intersections C0, C1, and C2 of pixel values between adjacent peak points can increase the angle detection sensitivity by using a portion having a relatively large gradient and a relatively large SN ratio. It is preferably about half the peak point.

2). Second Stage FIG. 6 (D) shows FIG. 5 (C) in one period corresponding to FIG. 6 (C). By determining which curve the lightness V (x) at the intra-period reference light position x is closest to, the period of the intra-period reference light position x is determined. In other words, a curve connecting the midpoints of adjacent curves in FIG. 6D is a boundary line for period determination as shown in FIG. 6E, and between which boundary lines (V = 0 and V = The period of the reference light position x in the period is determined based on whether the brightness V (x) is included in the boundary line ∞.

  According to the reference light position determination principle as described above, the band-shaped surface light source is configured so that the mountain-shaped luminance distributions having different wavelength regions overlap each other, and the pixel value distribution corresponding to the mountain-shaped luminance distribution of the captured image in each wavelength region Since the angle of the inspection surface is detected based on the above, a gradient larger than the conventional gradient shown in FIG. 13A can be used, and this brings about the effect that the detection desire rate is improved. Furthermore, since the lightness V is varied for each period to determine which period the reference light position x within the period is, the effect of widening the inspectable angle range is achieved.

  Returning to FIG. 1, the input unit 41 is for inputting set values and instructions to the control unit 40, and the display unit 42 is used to interactively perform this input and display defect information and the like according to the instructions. It is. In the approximate expression information storage unit 43, as the curve approximation information for approximately determining the curve of FIG. 6C and FIG. 6D or 6E as a curve, a broken line, or a straight line. A plurality of measurement points (x, d) and (V, x) are stored. That is, the approximate expression information storage unit 43 corresponds to the same reference light position x with respect to the pixel value distribution with respect to x corresponding to the luminance distribution of each of a plurality of wavelength ranges of the image captured by the imaging device 30. Information for approximately obtaining the pixel value distribution of the maximum pixel value among the pixel values of the plurality of wavelength ranges, and a boundary line for distinguishing between the distribution of the brightness V with respect to x or the brightness distribution of adjacent groups with respect to the brightness Is stored for the purpose of approximating for each period.

  Usually, the normalized pixel value distribution corresponding to the reference light position x is substantially the same for the plurality of wavelength ranges, and the pixel distribution is symmetric with respect to the axis of the peak value, so that the approximate expression information is greatly reduced. be able to.

  Instead of the reference light position x, an amount corresponding to x, for example, an angle at the inspection line 12 with respect to the reference plane may be used. Of course, this information may be a coefficient of an approximate expression.

  The control unit 40 acquires position information of the illumination imaging device LID with respect to the inspection surface 10 from a robot arm (not shown). The control unit 40 supplies a control signal to the imaging device 30 to transfer image data from the line sensors 30R, 30G, and 30B to the image memory 44 by DMA (direct memory access). After this transfer, the ready signal for the inspection point angle detector 45 is activated. In response to this activation, the inspection point angle detection unit 45 starts the process shown in FIG. 7 for each pixel (one set of R, G, and B pixel values DR, DG, and DB). In the following, the step identification codes in the figure are in parentheses.

  (S10) The brightness V = DR + DG + DB is obtained.

  (S11) The pixel values DR, DG, and DB are normalized with the brightness V as described above.

  (S12) Based on the magnitude relationship among the pixel values DR, DG, and DB and the information corresponding to FIG. 6C in the approximate expression information storage unit 43, the intra-period reference light position x is determined as described above. decide.

  (S13) Based on the value of brightness V and the information corresponding to FIG. 6D in the approximate expression information storage unit 43, it is determined to which period i the intra-period reference light position x belongs. That is, the cycle start point position Xi is obtained.

  (S14) Based on the relationship between the reference light position X and the inclination angle (detection angle θ) with respect to the reference plane in the inspection line 12 determined by experiment or calculation, the reference light position X is converted into this detection angle θ.

  The inspection point angle detection unit 45 stores the detection angle θ at each point on the inspection line 12 corresponding to the pixel position in the imaging device 30 in the angle information storage unit 46 together with the position / posture information of the illumination imaging device LID. (A first-stage angle map is created in the angle information storage unit 46), and an angle detection completion signal is activated for the control unit 40.

  By repeating such processing while scanning the illumination imaging device LID relative to the inspection surface 10, an angle distribution on the inspection surface 10 is obtained in the angle information storage unit 46.

  In parallel with the above processing, the surface defect detection unit 47 determines the angle of the defect point with respect to a good surface (a wide range of surfaces where the detection angle changes relatively slowly) based on information in the angle information storage unit 46. (A second-stage angle map is created in the angle information storage unit 46).

  By performing such processing, for example, the illumination imaging device LID is scanned in the direction of the arrow as shown in FIGS. 8 and 9 with respect to the inspection surface 10 on which the uneven portion 11 as shown in FIG. 10A is formed. Angle information as shown in FIG. 10B can be acquired. FIG. 10B visually shows an outline of the angle map of the second stage, and the grid is a set of R, G, B pixels corresponding to the imaging device 30 on the inspection surface 10. It corresponds to the point. The hatched portions in FIG. 11B indicate that the detection angles are different from each other. A white portion indicates a good surface. An outline of the angle map for the inspection surface 10 on which the uneven portion 11A as shown in FIG. 10C is formed is visually shown in FIG. The blacked portion is a portion where the reference light spot does not exist on the light source surface 21 due to the large inclination angle, and the portion can be determined from the angle distribution on both sides thereof, and the absolute value is larger than the angle on each side. The angle is entered in this angle map.

  Next, the surface defect detection unit 47 determines whether each defect part is a severe or minor defect based on the distribution of the angle of defect points in the angle information storage unit 46 and the defect part size, and the result Is stored in the defect information storage unit 48 and is displayed on the display unit 42 to notify the worker with an alarm according to the degree of the defect.

  Conventionally, since the inspection of coating defects was performed by a person and an apparatus, and the apparatus was low in sensitivity, the level of automatic inspection did not reach that of humans, and the human inspection burden could not be reduced. Therefore, the angle detection sensitivity is high and the angle can be detected in a relatively wide range, so that it is possible to more reliably and accurately detect surface defects by the image processing, thereby reducing the human inspection burden. Therefore, it greatly contributes to reducing the manufacturing cost of automobiles and the like.

  FIG. 11 is an explanatory diagram of Embodiment 2 of the present invention.

  In Example 2, as shown in FIG. 11A, an ND (Neutral Density) filter 29 is bonded to the transparent substrate 25 in the illumination device 20A. As shown in FIG. 11B, the transmittance of the ND filter 29 linearly changes between ABs of the illumination device 20A.

  Each LED is adjusted so that the peak value of the pixel value in the imaging device 30 when the inclination angle of the inspection line 12 is continuously changed without using the ND filter 29 is constant as shown in FIG. Supply current. By using the ND filter 29 under these conditions, the brightness V of the pixel value changes substantially linearly with respect to the reference light position X as shown in FIG. Small black dots in FIG. 11D indicate points where the in-cycle reference light positions x are the same, and it is possible to determine which cycle belongs in the same manner as in the first embodiment. The intra-period reference light position x can also be obtained by the same method as in the first embodiment.

  The other points are the same as those in the first embodiment.

  In the first and second embodiments, since the curve becomes gentle near the peak level of the pixel value, the angle detection sensitivity decreases. If the sensitivity partially decreases, the detection error as a whole increases. Therefore, it is preferable to make the angle detection sensitivity as independent of the reference light position as possible.

  Therefore, in this embodiment, in order to solve this problem, instead of the processing in the first stage, the in-period reference light position x is obtained in consideration of two adjacent colors as follows.

  First, as shown in FIG. 12B, the level ofs of the straight line connecting the foot intersections is subtracted from each normalized pixel value r, g, b for one cycle as shown in FIG. A portion having a low S / N ratio is removed with a smooth pixel value distribution. As a result, x can be calculated using only two adjacent colors for each of the regions A0 to A5, so that the calculation is simplified and the image processing speed can be improved.

  Here, for the region A0, the ratio DG / DR has a one-to-one correspondence with x. Since DG / DR = (DG / V) / (DR / V), the ratio DG / DR is standardized. Therefore, x is approximately obtained by x = C * DG / DR (= C * g / r). Here, “*” is a multiplication symbol. The coefficient C is assumed to be a constant for simplification. In order to make the meaning of this expression easy to understand, as shown in FIG. 12C, each pixel value curve is considered by linear approximation like a dotted line.

  Here, the above linear approximation in the region A0 is performed by using the parameter p obtained by normalizing the DG in the range of 0 to 1 and the parameter q obtained by normalizing the range of x in the range of 0 to 1, q = p / (1 -P). The straight lines q = p and q = 1−p in FIG. 12D correspond to g (x) and r (x) on the right side of the above equation, respectively. In the range 0 ≦ p ≦ 0.5, it changes almost linearly as shown by the solid line in FIG. In FIG. 12C, in the region A0, except for both ends, the pixel value r is slightly larger than the corresponding approximate line and the pixel value g is slightly smaller than the corresponding approximate line. It becomes a substantially straight line as shown by the solid line in FIG.

  The advantage of this approximate expression is that the angle detection sensitivity is improved as compared with the case of the first embodiment because x changes substantially linearly with respect to the ratio g / r including the vicinity of the pixel value peak point in the region A0.

  The angle detection sensitivity is further improved by experimentally obtaining C as a function C (r) of the normalized pixel value r so that C * g / r changes linearly with respect to r. The value of C (r) is substantially equal to the width x0 of the region A0 and becomes x0 at both ends of the region A0.

  X is similarly obtained for other regions. That is, generally, x is obtained as follows.

When DR ≧ DG (region A0), x = C (r) * g / r
When DG> DR (area A1), x = C (g) * (2-r / g)
When DG> DB (area A2), x = C (g) (* (2 + b / g)
When DB> DG (area A3), x = C (b) * (4-g / b)
The same applies hereinafter. The function C (normalized pixel value) is substantially the same for each color.

  When the coefficient C is approximated by a constant, the curve approximation information storage unit 43 in FIG. 1 only needs to store curve approximation information for the second stage, and the configuration is simplified. When the function C (standardized pixel value) is used as in the above equation, information for approximately obtaining the function C is stored in the curve approximation information storage unit 43.

  Note that the angle detection sensitivity in the vicinity of the maximum pixel value may be further improved by using the coefficient C as a function of the ratio between the corresponding maximum pixel value and the second largest pixel value for each region. In general, the intra-period reference light position x is represented by a function x (ρ) of the ratio ρ between the corresponding maximum pixel value and the second largest pixel value for each of the above-described regions, and a curve of this function is approximately obtained. The information for this may be stored in the curve approximation information storage unit 43, and x may be obtained from the value of this information and ρ. For example, when DR ≧ DG (region A0), x (ρ) where x = C (ρ) * ρ and ρ = g / r is one of the expression forms of this generalized function.

  Next, when m = MIN (DR, DG, DB) and M = MIN (DR, DG, DB), the saturation degree S in the HSB color space is expressed as S = (M−m) / M. . Here, MIN and MAX are the minimum value and the maximum value among the pixel values in parentheses, respectively.

  The saturation S is 1 when the regular reflection is performed, and the value is low when the regular reflection is not performed. Therefore, the saturation S determines whether or not the regular reflection is performed. Since dust or the like diffuses light, the saturation S becomes small, and the saturation S is also used for determination of dust or the like.

  The other points are the same as those in the first embodiment.

  According to the third embodiment, the angle detection sensitivity is high, the sensitivity to the angle of the detection target is stable, and the angle can be detected in a relatively wide range. Therefore, the surface defect detection can be performed more reliably and accurately by the image processing. This makes it possible to further reduce the human inspection burden and contribute to the reduction of the manufacturing cost of automobiles and the like.

  In the above, preferred embodiments of the present invention have been described. However, the present invention includes various modifications, and other combinations of the components described in the above embodiments and functions of the components. Other configurations that use other configurations, and other configurations that would be conceived by those skilled in the art from these configurations or functions are also included in the present invention.

  For example, “the angle corresponding to the reference light position” in claim 1 is not limited to a normal unit of “angle”, and may be an expression corresponding to the angle, and may be the reference light position itself. Also good.

  Moreover, the structure of detecting the angle of the area | region near the peak of a pixel value with the processing method of Example 3, and detecting the angle of another area | region with the processing method of Example 1 may be sufficient.

  Moreover, the structure of FIG.6 (C) may be predetermined about each irregular period, without considering the said regular periodicity. In this case, the variable resistor 28 may be a fixed resistor. This irregular cycle may be, for example, one obtained by exchanging the curves of cycles 0 to 4 in FIG.

  From the above principle, the arrangement order of the R, B, and G LED rows may be arbitrary. Further, an infrared LED may be added to the R, G, and B LEDs. In this case, from the above principle, one cycle can be made into six columns (6 is the number of combinations in which two are extracted from four). The color of the LED may be two or more colors.

  Further, the configuration may be such that the reference light position on the monochromatic belt-like surface light source surface corresponding to this portion is also detected without providing the light shielding plates 270 and 271. Also in this case, an angle map as shown in FIG. 10D can be obtained.

  Moreover, the structure which stores the said curve approximation information in the memory | storage part 43 by using the distance between the illuminating device 20 or the imaging device 30, and the test | inspection surface 10 as a parameter, and detects an angle according to this distance may be sufficient.

  Further, an infrared LED having a relatively large divergence angle is also arranged on the substrate 22 without using the ND filter 29 in the second embodiment, and the luminance distribution on the light shaping diffusion plate 26 is between AB as shown in FIG. The gradient start point position Xi may be obtained in the same manner as the above-described processing described with reference to FIG. 11D by using an infrared detection pixel value instead of the lightness V described above. .

  Furthermore, the illumination device may be a device that uses a band-shaped surface light source in which the luminance distribution along the scanning direction on the light source surface is mountain-shaped and the adjacent wavelength band luminance distributions overlap, and a laser diode is used instead of the LED. It may be used, or a semi-cylindrical microlens array may be used as the light shaping diffusion plate 26.

  Further, the brightness V in the second stage is corrected according to the pixel position of the line sensor, for example, by multiplying V by a constant, and by determining which period it is, the sensitivity of each pixel of the line sensor is increased. It may be configured to correct variations and luminance variations along the longitudinal direction of the strip-shaped light source.

  Moreover, the structure which detects the defect other than the defect described in the present Example by image processing may be sufficient.

  In the above embodiment, the case where the imaging device 30 is a one-dimensional camera has been described for easy understanding. However, a two-dimensional camera may be used. In this case, a plurality of inspection lines 12 may be used, each line corresponding to the area sensor may be associated with each line, and the same processing as described above may be performed for each line.

LID illumination imaging device 10 inspection surface 11 uneven portion 12 inspection line 20, 20A illumination device 21 light source surface 231 LED
240 to 246 Three-color LED array 25 Transparent substrate 26 Light shaping diffusion plate 30 Imaging device 30R, 30G, 30B Line sensor 40 Control unit 43 Approximation expression information storage unit 44 Image memory 45 Inspection point angle detection unit 46 Angle information storage unit 47 Surface Defect detection unit 48 Defect information storage unit RP0 to RP2 Reference light spot S0 to S2 Tangent plane

Claims (6)

Illumination means in which a plurality of strip light sources having different wavelength ranges are arranged in parallel,
Spectroscopic means for splitting the light emitted from the illumination means and regularly reflected by the light inspection surface into each of the plurality of wavelength ranges;
Imaging means comprising an image sensor on which each of the split light is imaged;
Image processing means for detecting an angle of the inspection surface based on data of an image picked up by the image pickup means;
A surface inspection apparatus comprising:
The illumination means has a luminance distribution with respect to a reference light position along a line perpendicular to the longitudinal direction of the strip light source on the light source surface of the plurality of strip light sources. The brightness decreases as the distance from the point decreases, and the brightness distribution of adjacent strip light sources is configured to overlap at least to the peak position of the brightness,
The image processing means, based on a combination of the wavelength range of the maximum pixel value and the wavelength range of the second largest pixel value, for pixel values corresponding to each of the plurality of wavelength ranges corresponding to the same reference light position A range including the reference light position is determined, a reference light position within the range is determined based on the maximum pixel value or based on the maximum pixel value and the second largest pixel value, and the reference light First processing means for determining the angle corresponding to the position;
A surface inspection apparatus characterized by that. The first processing means of the image processing means determines the reference light position within the range based on a value of a ratio between the second largest pixel value and the maximum pixel value;
The surface inspection apparatus according to claim 1. The first processing means of the image processing means, for each peak pixel value, with respect to a distribution of pixel values corresponding to each of the plurality of wavelength regions and normalized so as not to depend on lightness with respect to a reference light position, A predetermined value is subtracted as an offset value from each normalized pixel value so that a normalized pixel value in a wavelength region different from the wavelength region of the peak pixel value becomes approximately 0, and after the subtraction, based on the value of the ratio Determining a reference beam position within the range,
The surface inspection apparatus according to claim 2. The correction value multiplied by the ratio value so that the reference light position from the reference position is proportional to the ratio value is approximated as a function of the maximum pixel value normalized so as not to depend on lightness or a function of the ratio. Further comprising storage means for storing information required for
The first processing means of the image processing means multiplies the value of the ratio by the correction value to determine the reference light position;
The surface inspection apparatus according to claim 3. Each pixel of the plurality of wavelength regions corresponding to the same reference light position with respect to the pixel value distribution corresponding to the amount corresponding to the luminance distribution of each of the plurality of wavelength regions of the image captured by the imaging unit A first approximate expression information storage means for storing first information for approximately obtaining a distribution of the maximum pixel value of the values normalized so as not to depend on lightness;
The first processing means of the image processing means calculates a reference light position corresponding to the normalized maximum pixel value within the range in the approximate expression of the normalized pixel value distribution of the maximum pixel value based on the first information. Ask,
The surface inspection apparatus according to claim 1. The illumination means includes a plurality of band-like light sources having different wavelength ranges arranged in parallel, the luminance distribution is different for each group, and the pixel values of the image with respect to in-group reference light positions corresponding to the plurality of groups Is substantially constant regardless of the reference beam position in the group,
Lightness distribution of the sets adjacent to each other with respect to the lightness, or approximately the lightness distribution based on the respective pixel values of the plurality of wavelength ranges corresponding to the same light reference position, A second approximate expression information storage means storing second information for approximately obtaining a boundary line for distinguishing between
The image processing means normalizes the pixel values corresponding to each of the plurality of wavelength ranges corresponding to the same reference light position so as not to depend on the set, and the first processing means performs the reference light position in the set. Seeking
The image processing means further determines which set of reference light positions in the set is based on the brightness distribution or the approximate expression of the boundary line based on the second information and the reference light position in the set. And a second processing means for obtaining a reference light position in the entire range based on the reference light position in the set and the determined set.
The surface inspection apparatus according to any one of claims 2 to 5, wherein

Images