JP5844718B2 - Image processing method, image processing apparatus, surface inspection system, and program - Google Patents

Description

  The present invention relates to an image processing method, an image processing apparatus, a surface inspection system, and a program for performing desired image processing on a wide-angle image including an image portion corresponding to an inner wall side surface of a workpiece.

  Conventionally, the surface of a workpiece, which is an object to be inspected, is imaged by an imaging device, and the obtained captured image is analyzed to determine whether or not a defective part such as a scratch or a dent exists on the surface of the workpiece Surface inspection systems and methods have been developed. However, when detecting a defective part having a very small degree of deformation of the surface shape, the form is not sufficiently expressed (visualized) by the captured image, and the presence of the defective part may be missed. Therefore, various image processing techniques for further improving the detectability or visibility of a defective part have been proposed.

  In Patent Document 1, diffused light having a lightness gradation is projected onto the surface of a workpiece, and a defective portion is automatically extracted by sequentially performing differentiation processing and binarization processing on the obtained captured image. A device has been proposed.

International Publication No. 2008/123604 Pamphlet

  By the way, the above-described surface inspection method can be applied not only to the outer surface having a relatively large area including the vehicle body and the like, but also to the side surface of the inner wall of the fine nozzle hole such as an injector of an engine. At the time of this inspection, a thin tubular endoscope is accessed around the position to be inspected, and imaging using auxiliary light emitted from the distal end portion of the endoscope is performed. The reason for using the auxiliary light is that the outside light is usually blocked around the side surface of the inner wall, and the exposure amount for imaging is insufficient.

  In this case, due to the light diffusion phenomenon, the light amount of the auxiliary light tends to attenuate rapidly as the distance from the tip portion increases. That is, luminance unevenness in which the luminance (signal level) rapidly decreases along the depth direction of the space portion such as a hole occurs on the obtained captured image. As a result, there has been a problem that the detectability or visibility of a defective portion at a particularly recessed position is lowered.

  However, the apparatus described in Patent Document 1 does not consider the above-described inspection form at all, and does not solve any problems caused by luminance unevenness along the depth direction.

  The present invention has been made to solve the above-described problem, and an image processing method and an image that can improve the detectability or visibility of a defective part at a recessed position even when inspecting the inner wall side surface of a workpiece. It is an object to provide a processing apparatus, a surface inspection system, and a program.

  The image processing method according to the present invention is an image corresponding to an inner wall side surface of the work around the space, obtained by imaging with the endoscope facing in the depth direction of the space formed in the work. An input step of inputting a wide-angle image including a part; and a developed image of the inner wall side surface with the circumferential direction and the depth direction of the inner wall side surface as coordinate axes by performing geometric transformation on the input wide angle image Obtained by sequentially performing a first smoothing process along the depth direction and a second smoothing process along the circumferential direction on the developed image from the created creation step and the created developed image. The image processing apparatus is caused to execute a removal process for removing the trend.

  In this way, in the removal step of removing the trend obtained by sequentially performing the first smoothing process along the depth direction and the second smoothing process along the circumferential direction on the developed image from the developed image. Then, two types of smoothing processing are performed in the order in which the degree of occurrence of luminance unevenness is relatively large along the luminance transition direction on the developed image. In other words, the brightness distribution on the developed image can be made uniform globally by creating / removing the trend based on the tendency of uneven brightness. Thereby, even if it is a case where the inner wall side surface of a workpiece | work is test | inspected, the detectability or visibility of the defect site | part in the deep position can be improved.

  The first smoothing process preferably includes a process of determining an approximate curve of a pixel value profile along the depth direction. By using an approximate curve as the first smoothing process, it becomes difficult to be influenced by the singular point of the data.

  Furthermore, the approximate curve is preferably an exponential function. By applying an exponential function, it is possible to accurately match an actual pixel value profile.

  Furthermore, it is preferable that the approximate curve is the exponential function that minimizes the sum of residual absolute values with the pixel value profile. This makes it less susceptible to data singularities and improves fitting accuracy.

  Furthermore, it is preferable to further execute an adjustment step of adjusting a trend removal image obtained by removing the trend of the developed image according to the trend. Thereby, the detectability or visibility of the defect site | part in the image after adjustment can be improved further.

  Further, the removing step includes generating the trend image of the developed image by performing the second smoothing process after performing the first smoothing process on the developed image; It is preferable to include a step of subtracting the generated trend image from the developed image.

  Further, the removing step is generated by subtracting a first smoothed image obtained by performing the first smoothing process on the developed image from the developed image, and generating an intermediate image. It is preferable to include a step of subtracting a second smoothed image obtained by performing the second smoothing process on the intermediate image from the intermediate image.

  The image processing apparatus according to the present invention is an image corresponding to an inner wall side surface of the work around the space part, which is obtained by capturing an image with the endoscope facing the depth direction of the space part formed in the work. A wide-angle image input unit that inputs a wide-angle image including a part; and a geometric transformation on the wide-angle image input by the wide-angle image input unit, so that the circumferential direction and the depth direction of the inner wall side surface are coordinate axes. A developed image creating unit that creates a developed image of the side surface of the inner wall, a first smoothing process along the depth direction with respect to the developed image from the developed image created by the developed image creating unit, and the And a trend removing unit that removes the trend obtained by sequentially performing the second smoothing process along the circumferential direction.

  A surface inspection system according to the present invention includes the above-described image processing device and an imaging device that includes the endoscope and acquires the wide-angle image by imaging through the endoscope.

  The program according to the present invention obtains an image portion corresponding to an inner wall side surface of the workpiece around the space portion, which is obtained by imaging in a state where the endoscope is directed in the depth direction of the space portion formed in the workpiece. An input step of inputting a wide-angle image including the image, and a geometric transformation is performed on the input wide-angle image, thereby creating a developed image of the inner wall side surface with the circumferential direction and the depth direction of the inner wall side surface as coordinate axes A trend obtained by sequentially performing a first smoothing process along the depth direction and a second smoothing process along the circumferential direction on the developed image from the created process and the created developed image And a removal step of removing the image.

  According to the image processing method, the image processing apparatus, the surface inspection system, and the program according to the present invention, the first smoothing process along the depth direction and the second along the circumferential direction are performed on the developed image from the developed image. In the removal step of removing the trend obtained by sequentially performing the smoothing processing, the two types of smoothing processing are performed in the order of the occurrence of luminance unevenness along the transition direction of the luminance on the developed image. Has been given. In other words, the brightness distribution on the developed image can be made uniform globally by creating / removing the trend based on the tendency of uneven brightness. Thereby, even if it is a case where the inner wall side surface of a workpiece | work is test | inspected, the detectability or visibility of the defect site | part in the deep position can be improved.

It is a schematic perspective view which shows the structure of the surface inspection system incorporating the image processing apparatus which concerns on this embodiment. It is an electrical block diagram of the image processing apparatus shown in FIG. It is a flowchart with which operation | movement description of the test | inspection system of FIG. 1 is provided. FIG. 2 is an enlarged cross-sectional view of the endoscope and workpiece shown in FIG. 1. It is a figure which shows the wide angle image input from the wide angle image input part of FIG. It is a figure which shows the expansion | deployment image in the wide angle image of FIG. FIG. 3 is a detailed block diagram of a trend removing unit shown in FIG. 2. 8A and 8B are schematic explanatory diagrams regarding the operation of the first smoothing processing unit in FIG. 9A and 9B are schematic explanatory diagrams regarding the operation of the first smoothing processing unit in FIG. 7. 10A and 10B are schematic explanatory diagrams regarding the operation of the second smoothing processing unit in FIG. 11A and 11B are schematic explanatory diagrams regarding the operation of the second smoothing processing unit in FIG. FIG. 12A is a partially enlarged view of the original developed image. FIG. 12B is a partially enlarged view of the image after the trend processing according to the conventional example is performed. FIG. 12C is a partially enlarged view of the image after performing the differentiation process according to the conventional example. FIG. 12D is a partially enlarged view of the image after the trend removal process according to the present embodiment is performed. It is a detailed block diagram of the trend removal part which concerns on a modification.

  Hereinafter, an image processing method according to the present invention will be described with reference to the accompanying drawings by giving preferred embodiments in relation to an image processing apparatus, a surface inspection system, and a program for executing the image processing method.

[Configuration of Surface Inspection System 10]
FIG. 1 is a schematic perspective view showing a configuration of a surface inspection system 10 incorporating an image processing apparatus 14 according to this embodiment.

  The surface inspection system 10 includes an imaging device 12 that images the surface shape of the workpiece W, and an image processing device 14 that performs desired image processing on the imaging signal (wide-angle image 72 in FIG. 5) output from the imaging device 12. Is basically provided.

  The image processing device 14 is a computer that includes a main body unit 16, an input unit 18, and a display unit 20. The imaging device 12 is communicably connected to the main body 16 of the image processing device 14 via a cable 22. Note that the imaging device 12 and the image processing device 14 may be communicable wirelessly instead of wired communication by the cable 22.

  The imaging device 12 includes a camera body 24, a light guide unit 26, an endoscope 28, and an auxiliary light source 30. For example, the camera body 24 is held by a robot arm (not shown), and the position and posture of the imaging device 12 can be freely changed according to the driving of the robot arm.

  The camera body 24 includes an image sensor (image sensor array) that converts an input light image into an electrical signal. The imaging element is composed of, for example, a PD (PhotoDiode), a CCD (Charge Coupled Device), a CMOS (Complementary Metal Oxide Semiconductor), or the like. The imaging signal output from the camera body 24 may be a monochrome image or a color image, and the resolution may be high or low (pixel size).

  The light guide unit 26 optically connects between the auxiliary light source 30 and the endoscope 28 or between the endoscope 28 and the camera body 24.

  The endoscope 28 is an industrial fiberscope having a thin tube shape and an optical system incorporated therein. In this example, at least the distal end portion 36 of the endoscope 28 is inserted into a substantially cylindrical space 34 formed on the one surface 32 of the workpiece W.

  The auxiliary light source 30 is a light source that emits auxiliary light 66 (see FIG. 4) for illuminating the periphery of the distal end portion 36 when acquiring a light image 70 (see FIG. 4) from the endoscope 28. The auxiliary light 66 may be visible light, ultraviolet light, or infrared light, and is preferably light in a wavelength range suitable for imaging the workpiece W. In other words, the camera body 24 (sensitivity characteristics of the image sensor) and the auxiliary light source 30 may be variously combined depending on the detection target.

  In addition, although the space part 34 in this example is a cylindrical hole, its shape (cross section, depth, etc.) or size is arbitrary. The form of the space 34 is not limited to a hole, and may be either a hole or a recess.

  FIG. 2 is an electrical block diagram of the image processing apparatus 14 shown in FIG. The main unit 16 includes a control unit 40, a communication I / F 42, a display control unit 44, and a memory 46 (storage medium).

  The communication I / F 42 is an interface that transmits / receives an electrical signal to / from an external device (the imaging device 12 in the example of FIG. 1). The input unit 18 includes various input devices such as a mouse, a trackball, and a keyboard.

  The display control unit 44 is a control circuit that drives and controls the display unit 20 according to the control of the control unit 40. The display control unit 44 drives the display unit 20 by outputting a display control signal to the display unit 20 via an I / F (not shown). Thereby, the display unit 20 can display various images.

  The memory 46 stores programs, data, and the like necessary for the control unit 40 to control each component. The memory 46 may be a non-transitory and computer-readable storage medium including a non-volatile memory, a hard disk, and the like.

  The control unit 40 is configured by a processor such as a CPU (Central Processing Unit). The control unit 40 reads out and executes the program stored in the memory 46, whereby a wide-angle image input unit 48, a developed image creation unit 50, a trend removal unit 52, a defect determination unit 54, and a display data creation unit 56 (images). (Including the adjustment unit 58) can be realized.

[Operation of surface inspection system 10]
Next, the operation of the surface inspection system 10 shown in FIG. 1 will be described in detail with reference to the flowchart of FIG.

  In step S1, a workpiece W as an inspection object is prepared. In this embodiment, the surface shape of the inner wall side surface 38 of the workpiece W is inspected. For example, the workpiece W corresponds to an injector, and the space 34 corresponds to a nozzle hole of the injector.

  In step S2, the endoscope 28 and the workpiece W are relatively moved to set the endoscope 28 under a predetermined position / posture. For example, the camera body 24 is held by a robot arm (not shown), and the endoscope 28 is moved by driving the robot arm. Then, the endoscope 28 is held in a state where the distal end portion 36 side is directed in the depth direction of the space portion 34. In the example of FIG. 1, at least the distal end portion 36 is inserted into the space portion 34, but the endoscope 28 may be set outside the space portion 34 as long as the surface shape of the inner wall side surface 38 can be imaged.

  In step S <b> 3, the imaging device 12 is used to capture a wide-angle image 72 (see FIG. 5) that shows the surface shape of the inner wall side surface 38.

  FIG. 4 is an enlarged cross-sectional view of the endoscope 28 and the workpiece W shown in FIG.

  The endoscope 28 includes a first light guide 62 composed of one optical fiber cable or the like, an objective lens 63 (wide angle optical system) designed with a wide viewing angle, and a second light guide 64 composed of a bundle of optical fiber cables. Prepare.

  The auxiliary light 66 from the auxiliary light source 30 is guided through the light guide unit 26 and the first light guide 62, and is radiated from the distal end portion 36 side to the outside (space portion 34). As a result, due to the light diffusion phenomenon, the amount of the auxiliary light 66 is maximized around the distal end portion 36 and tends to attenuate rapidly as it moves away from the distal end portion 36 in the Z axis (depth direction). In addition, each arrow which attached | subjected the reference code | symbol 66 has shown typically that there is so much light quantity that the width | variety is wide.

  On the other hand, the light included in the field of view 68 of the endoscope 28 enters from the distal end portion 36 side and is collected by the objective lens 63. Here, since the objective lens 63 has a wide viewing angle of, for example, 100 degrees or more, a part of the inner wall side surface 38 is included in the field of view 68. As the objective lens 63, a fish-eye lens having a viewing angle of about 180 degrees may be applied.

  The condensed light image is guided through the second light guide 64 and the light guide unit 26 and is converted into an electric signal by the image sensor of the camera body 24. Thereafter, the electrical signal is output as an imaging signal from the camera body 24 and temporarily stored in the memory 46 via the cable 22 and the communication I / F 42 (see FIG. 2).

  For example, the wide-angle image 72 obtained here is a monochrome image, and this pixel value is defined by 8 bits (256 gradations). When the pixel value is 0, the pixel shows the lowest level of brightness (brightness), and when the pixel value is 255, the pixel shows the highest level of brightness (brightness).

  FIG. 5 is a diagram showing a wide-angle image 72 input from the wide-angle image input unit 48 of FIG. In this figure, a part with a high luminance level (part with a large pixel value) is expressed in white, and a part with a low luminance level (part with a small pixel value) is expressed in black. Hereinafter, the images shown in FIGS. 6, 9B, and 11B to 12D are similarly defined.

  As can be understood from the drawing, the wide-angle image 72 includes an image part (part drawn in white) corresponding to the inner wall side surface 38 of the workpiece W. In FIG. 4, when the optical axis 69 of the endoscope 28 coincides with the central axis of the space portion 34, the inner wall side surface 38 is drawn as an annular portion.

  By the way, the x axis-y axis is an orthogonal coordinate axis with the upper left corner of the wide-angle image 72 as the origin. The z axis-θ axis is a coordinate axis with the reference position on the inner wall side surface 38 as the origin. More specifically, the z axis is a coordinate axis along the depth direction of the space portion 34, and the θ axis is a coordinate axis along the circumferential direction of the inner wall side surface 38.

  In this way, the wide-angle image input unit 48 reads out and inputs the imaging signal indicating the wide-angle image 72 from the memory 46 (step S3).

  In step S4, the developed image creation unit 50 creates a developed image 74 of the inner wall side surface 38 from the wide-angle image 72 acquired in step S3. Specifically, the developed image creating unit 50 performs a predetermined geometric transformation to develop the wide-angle image 72 defined in the x-axis-y-axis coordinate system and the developed image defined in the z-axis-θ axis coordinate system. The image 74 is converted. In order to perform this conversion, each information (geometric information) related to the shape of the inner wall side surface 38 of the workpiece W and the position / posture of the imaging device 12 can be used to accurately map between the wide-angle image 72 and the developed image 74. The developed image creation unit 50 may acquire geometric information from a measurement result by a measurement unit (not shown), or may use a fixed value (default value) prepared in advance as the imaging condition is known. Good.

  FIG. 6 is a diagram showing a developed image 74 in the wide-angle image 72 of FIG. The developed image 74 is a two-dimensional image in which the depth direction (z axis) of the space 34 is the first coordinate axis and the circumferential direction (θ axis) of the inner wall side surface 38 is the second coordinate axis.

  As can be understood from the figure, luminance unevenness along the z-axis occurs on the developed image 74. More specifically, the luminance is higher as it is above the developed image 74 (closer to the tip portion 36), and the luminance is lower as it is below the developed image 74 (farther from the tip portion 36). This uneven brightness is caused by the attenuation of the auxiliary light 66 described above.

  In addition, streaky luminance unevenness slightly tilted with respect to the θ direction is also generated on the developed image 74. This can occur when the optical axis 69 (see FIG. 4) of the endoscope 28 is inclined with respect to the central axis of the space portion 34.

  In step S5, the trend removing unit 52 removes the trend from the developed image 74 created in step S4. This “trend” means a global change tendency on the image, and is specifically obtained by performing a smoothing process on the developed image 74. The “smoothing process” in this specification uses not only an image processing method for smoothing the developed image 74 by performing image processing, but also an optimization method for adapting the developed image 74 to a predetermined approximate curve. May be. As the former image processing technique, for example, various known techniques including moving average, convolution calculation including a Gaussian filter, median filter, statistical calculation including maximum / minimum value filter, and morphology calculation can be applied. As the latter optimization method, various known methods including function approximation calculation, fitting calculation, interpolation calculation, and multivariate analysis can be applied.

  FIG. 7 is a detailed block diagram of the trend removing unit 52 shown in FIG. The trend removing unit 52 includes a first smoothing processing unit 100, a second smoothing processing unit 102, and a subtractor 104. Hereinafter, the function of each component will be described in detail with reference to FIGS. 8A to 11B as appropriate.

  The first smoothing processing unit 100 in FIG. 7 performs a first smoothing process along the z-axis on the input developed image 74. 8A to 9B are schematic explanatory diagrams regarding the operation of the first smoothing processing unit 100 of FIG.

  As shown in FIG. 8A, the first smoothing processing unit 100 first extracts an arbitrary image sequence 78 existing in the image region 76. Here, FIG. 8B is an example of the pixel value profile along the direction of the broken line arrow in the image sequence 78. The horizontal axis of this graph is the z-axis position (pixel unit), and the vertical axis is the luminance level (pixel value). The base end side of the broken-line arrow corresponds to the “0” side position, and the tip end side corresponds to the “400” side position. In this graph, there is a tendency that the pixel value gradually decreases with respect to the z-axis, which is consistent with the light and shade tendency of the developed image 74 shown in FIG.

  Next, the first smoothing processing unit 100 determines an approximate curve of the pixel value profile in the image sequence 78 along the z axis. In this embodiment, an exponential function that can be accurately matched to an actual pixel value profile is used as the approximate curve. Further, as an evaluation value of the approximation error to be minimized, SAD (Sum of Absolute Difference), SSD (Sum of Squared Difference), or the like may be used. The value is not limited to these. In particular, in order to make it difficult to be affected by the singular point of data, it is preferable to apply SAD having a small residual order.

  As a result, as shown in FIG. 9A, an exponential function that minimizes the SAD with the pixel value profile is created as an approximate curve. In the same manner, the first smoothing processing unit 100 sequentially executes image sequence extraction and approximation function creation. Then, as shown in FIG. 9B, a smoothed image 80 as an aggregate of approximate functions is obtained. Note that the calculation range for creating the approximate function may be for each image sequence as shown in the example of FIG. 8A, or may extend to a plurality of image sequences.

  The second smoothing processing unit 102 in FIG. 7 performs a second smoothing process along the θ axis on the input smoothed image 80. 10A to 11B are schematic explanatory diagrams regarding the operation of the second smoothing processing unit 102 of FIG.

  As illustrated in FIG. 10A, the second smoothing processing unit 102 first extracts an arbitrary image sequence 82 present in the image area 76. Here, FIG. 10B is an example of the pixel value profile along the broken line arrow direction in the image sequence 82. The horizontal axis of this graph is the θ-axis position (pixel unit), and the vertical axis is the luminance level (pixel value). In this graph, the pixel values tend to change regularly (periodically) with respect to the θ axis. Note that the degree of variation of each value of the profile is greatly reduced as compared with the profile in the image sequence 82 of the developed image 74. This is due to the effect of performing the smoothing process along the z-axis where the degree of luminance unevenness is relatively large.

  Next, the second smoothing processing unit 102 determines a smoothing curve of the pixel value profile in the image sequence 82 along the θ axis. In the present embodiment, cubic spline fitting is used as a smoothing method. As a result, the smoothing curve shown in FIG. 11A is created. Similarly, the second smoothing processing unit 102 sequentially executes image sequence extraction and smoothing curve creation. Then, as shown in FIG. 11B, a trend image 84 as an aggregate of smoothing curves is generated. Note that the calculation range when creating the smoothing curve may be for each image sequence as in the example of FIG. 10A or may extend over a plurality of image sequences.

  The subtracter 104 in FIG. 7 generates and outputs a trend-removed image by subtracting a trend image 84 obtained by sequentially performing the first and second smoothing processes on the developed image 74 from the developed image 74. To do. This trend-removed image is an image from which a two-dimensional trend on the z-axis and the θ-axis has been removed, and the average pixel value is approximately zero. In this way, the trend removing unit 52 removes the trend from the developed image 74 (step S5).

  In step S <b> 6, the defect determination unit 54 performs various image processing on the trend-removed image created in step S <b> 5 and determines whether or not a defect exists on the inner wall side surface 38. Specifically, the defect determination unit 54 determines the presence and strength of a specific shape including an abrasion and a burrow using a known pattern detection process. As a determination criterion, not only the shape and position, but also an allowable value of area, depth, or luminance difference may be set.

In step S7, the image adjustment unit 58 performs brightness / contrast adjustment (hereinafter referred to as “B / C adjustment”) on the trend-removed image created in step S5. Specifically, the image adjustment unit 58 performs B / C adjustment by calculating according to the following equation (1) for each pixel of the trend-removed image.
O (z, θ) = B + C · R (z, θ) / T (z, θ) (1)

  Here, O (z, θ) is the pixel value of the visualized image after B / C adjustment. B is brightness and corresponds to an offset value in the visualized image. C is contrast and corresponds to the enhancement degree of the visualized image. R (z, θ) is a pixel value of the trend-removed image. T (z, θ) is a pixel value of the trend image 84. That is, by adjusting the trend-removed image according to the trend image 84 (trend), it is possible to further improve the detectability or visibility of the defective portion in the adjusted image.

  In step S8, the display control unit 44 causes the display unit 20 to display the visualized image adjusted in step S7. Prior to the display, the display data creation unit 56 creates data (hereinafter referred to as display data) used for image display. This display data includes not only the visualized image but also the determination result (for example, annotation information) by the defect determination unit 54. Then, the display data creation unit 56 supplies the created display data to the display control unit 44. Thereby, the inspector as a user can grasp | ascertain the surface shape of the inner wall side surface 38 of the workpiece | work W, or its automatic determination result by visually recognizing the display part 20. FIG.

  In step S9, it is determined whether or not the inspection of the surface shape of the workpiece W has been completed. If it is determined that the process has not been completed, the process returns to step S2, and steps S2 to S8 are sequentially repeated. On the other hand, when it is determined that the process has been completed, the surface inspection system 10 completes the surface inspection of the work W and then replaces the work W as necessary.

[Effect of the present invention]
According to the image processing method of the present invention, in the vicinity of the space portion 34 obtained by imaging in a state where the endoscope 28 is directed in the depth direction (z axis) of the space portion 34 formed on the workpiece W. An input step (step S3) for inputting a wide-angle image 72 including an image portion corresponding to the inner wall side surface 38 of the work W, and geometrical transformation on the wide-angle image 72 make it possible to obtain a circumferential direction (θ-axis) of the inner wall side surface 38. ) And a development step (step S4) for creating the developed image 74 of the inner wall side surface 38 with the depth direction (z axis) as a coordinate axis, and the developed image 74 along the depth direction (z axis). The image processing apparatus 14 is caused to perform a removal step (step S5) for removing the trend obtained by sequentially performing the first smoothing process and the second smoothing process along the circumferential direction (θ axis).

  In the above-described removal process, two types of smoothing processing are performed in the order of relatively large occurrence of luminance unevenness along the luminance transition direction on the developed image 74. In other words, the brightness distribution on the developed image 74 can be globally uniformed by creating and removing the trend based on the tendency of uneven brightness. Thereby, even if it is a case where the inner-wall side surface 38 of the workpiece | work W is test | inspected, the detectability or visibility of the defect site | part in the back position can be improved.

  Hereinafter, the effect of the present invention will be described with reference to an actual visible image.

  FIG. 12A is a partially enlarged view of the original developed image 74. Here, the portion drawn in white corresponds to a burrow-shaped defect portion.

  FIG. 12B is a partially enlarged view of the image after the trend removal process according to the conventional example is performed. Here, a trend image is created by applying a two-dimensional moving average mask to the developed image 74. Comparing FIG. 12A and FIG. 12B, in FIG. 12B, many white-out artifacts are generated around the defective part. As a result, the size of the defective part may be erroneously detected and recognized.

  FIG. 12C is a partially enlarged view of the image after performing the differentiation process according to the conventional example. Comparing FIG. 12A and FIG. 12C, in FIG. 12C, only the outline of the defective part having a sufficiently large degree of deformation of the surface shape is extracted. As a result, the presence of a burrow that is within the allowable range cannot be confirmed on the image.

  FIG. 12D is a partially enlarged view of the image after the trend removal process according to the present embodiment is performed. When comparing FIG. 12A to FIG. 12D, since the degree of occurrence of luminance unevenness in the image shown in FIG. 12D is the smallest, the visibility of the defective portion is improved.

[Modification]
FIG. 13 is a detailed block diagram of the trend removing unit 106 according to the modification. The trend removing unit 106 includes two subtracters 108 and 110 in addition to the first smoothing processing unit 100 and the second smoothing processing unit 102. Hereinafter, the function of each component will be described in detail.

  First, the first smoothing processing unit 100 performs a first smoothing process along the z-axis on the input developed image 74 to thereby obtain a smoothed image 80 (see FIG. 9B; first smoothing). Image). Since the specific method of the first smoothing process is the same as that of the present embodiment, the description thereof is omitted.

  Next, the subtractor 108 generates and outputs an intermediate image by subtracting the smoothed image 80 obtained by performing the first smoothing process on the developed image 74 from the developed image 74. This intermediate image is an image from which the one-dimensional trend on the z-axis has been removed.

  Next, the second smoothing processing unit 102 performs a second smoothing process along the θ axis on the input intermediate image to obtain a second smoothed image. Since the specific method of the second smoothing process is the same as that of the present embodiment, description thereof is omitted.

  Finally, the subtractor 110 generates and outputs a trend-removed image by subtracting the second smoothed image obtained by performing the second smoothing process on the intermediate image from the intermediate image. This trend-removed image is an image from which a two-dimensional trend on the z-axis and the θ-axis has been removed.

  Thus, even if the trend removing unit 106 is configured to remove the trend of the developed image 74 in two steps, the same effect as the trend removing unit 52 of FIG. 7 can be obtained.

  In addition, this invention is not limited to embodiment mentioned above, Of course, it can change freely in the range which does not deviate from the main point of this invention.

DESCRIPTION OF SYMBOLS 10 ... Surface inspection system 12 ... Imaging device 14 ... Image processing device 16 ... Main part 18 ... Input part 20 ... Display part 24 ... Camera main body 26 ... Light guide unit 28 ... Endoscope 34 ... Space part 36 ... Tip part 38 ... Inner wall side surface 40 ... control unit 46 ... memory 48 ... wide angle image input unit 50 ... development image creation unit 52, 106 ... trend removal unit 58 ... image adjustment unit 72 ... wide angle image 74 ... development image 76 ... development image 76 ... 82 Column 80 ... smoothed image 84 ... trend image 100 ... first smoothing processing unit 102 ... second smoothing processing unit 104, 108, 110 ... subtractor

Claims (10)

Input to input a wide-angle image including an image portion corresponding to the inner wall side surface of the work around the space, obtained by imaging with the endoscope directed in the depth direction of the space formed in the work Process,
A creation step of creating a developed image of the inner wall side surface with the circumferential direction of the inner wall side surface and the depth direction as coordinate axes by performing geometric transformation on the input wide angle image;
Removal that removes a trend obtained by sequentially performing a first smoothing process along the depth direction and a second smoothing process along the circumferential direction on the developed image from the created developed image An image processing method comprising causing an image processing apparatus to execute the steps. The image processing method according to claim 1,
The first smoothing process includes a process of determining an approximate curve of a pixel value profile along the depth direction. The image processing method according to claim 2.
The image processing method, wherein the approximate curve is an exponential function. The image processing method according to claim 3.
The image processing method according to claim 1, wherein the approximate curve is the exponential function that minimizes the total sum of residual absolute values with the pixel value profile. In the image processing method of any one of Claims 1-4,
An image processing method, further comprising an adjustment step of adjusting a trend-removed image obtained by removing the trend of the developed image according to the trend. In the image processing method according to any one of claims 1 to 5,
The removal step includes
Generating a trend image of the developed image by further applying the second smoothing process after the first smoothing process is performed on the developed image;
A step of subtracting the generated trend image from the developed image. In the image processing method according to any one of claims 1 to 5,
Instead of the removal step,
Generating an intermediate image by subtracting a first smoothed image obtained by performing the first smoothing process on the developed image from the developed image;
Subtracting a second smoothed image obtained by performing the second smoothing process on the intermediate image from the generated intermediate image, and causing the image processing apparatus to execute the image processing. Method.
A wide-angle image for inputting a wide-angle image including an image portion corresponding to the inner wall side surface of the work around the space, obtained by imaging with the endoscope directed in the depth direction of the space formed in the work An image input unit;
A developed image creation unit that creates a developed image of the inner wall side surface using the circumferential direction and the depth direction of the inner wall side surface as coordinate axes by performing geometric transformation on the wide angle image input by the wide angle image input unit. When,
Obtained by sequentially performing a first smoothing process along the depth direction and a second smoothing process along the circumferential direction on the developed image from the developed image created by the developed image creation unit. An image processing apparatus comprising: a trend removing unit that removes the trend. An image processing apparatus according to claim 8,
A surface inspection system comprising: an imaging device that includes the endoscope and acquires the wide-angle image by imaging through the endoscope. Input to input a wide-angle image including an image portion corresponding to the inner wall side surface of the work around the space, obtained by imaging with the endoscope directed in the depth direction of the space formed in the work Process,
A creation step of creating a developed image of the inner wall side surface with the circumferential direction of the inner wall side surface and the depth direction as coordinate axes by performing geometric transformation on the input wide angle image;
Removal that removes a trend obtained by sequentially performing a first smoothing process along the depth direction and a second smoothing process along the circumferential direction on the developed image from the created developed image A program that causes a computer to execute the process.

Images