JP4322230B2 - Surface defect inspection apparatus and surface defect inspection method - Google Patents

Description

  The present invention evaluates an illumination unit composed of a plurality of light emitting elements arranged continuously, an imaging camera for imaging a surface to be inspected illuminated by light emitted by the illumination unit, and an output signal of the imaging camera The present invention also relates to a surface defect inspection apparatus comprising defect evaluation means for detecting defects on the surface to be inspected, and a surface defect inspection method using the surface defect inspection apparatus.

  A typical example of this type of inspection apparatus is a technique used for inspecting the painted surface of an automobile body. In such a surface inspection, unevenness, scratches, etc. existing on the painted surface as the surface to be inspected are the inspection objects. As an inspection technique using pattern-shaped inspection light, so-called stripes, that is, illuminating illumination light creating vertical stripes of light and darkness on the painted surface, and imaging the painted surface in an irradiated state with an imaging camera, There is a technique for performing a surface inspection using the obtained captured image. (Technology disclosed in Patent Document 1 and Patent Document 2).

  For example, when the painted surface is moved in a predetermined direction (for example, the X direction), an image portion of a defect such as an uneven surface on the painted surface has coordinates in a direction orthogonal to the moving direction (for example, the Y direction). There is one that detects defects by using the image while changing the direction coordinate (X coordinate) without changing, and in the captured image of the defect area, the light stripe portion is dark and dark stripes. Since the defect is identified using the fact that it is brightly imaged at the part, the defect is captured as an intermediate gradation image of a bright part and a dark part of the stripe (see Patent Document 1).

In order to detect defects called "Yuzu skin" that are periodic irregularities on the surface, it tries to find a coating thickness group by fluctuation on the captured image of the boundary line of light and dark stripes that are inspection light There is a thing (refer patent document 2). In this inspection method, it is not necessary to move the surface to be inspected, but in general, there is a paint surface on which a disturbance that causes displacement in the boundary image of the stripe occurs over a relatively wide range of the paint surface. It becomes a detection target.
JP-A-8-145906 (FIGS. 5, 9, and 15) JP-A-9-126744 (FIG. 13)

  In the above-described conventional surface inspection methods such as Patent Documents 1 and 2, since the illumination unit that illuminates the surface to be inspected irradiates the painted surface with a stripe-shaped light / dark pattern, the wraparound of irradiation light used for surface inspection is striped. This occurs only in a direction orthogonal (crossing) to the light-dark pattern, and inspection efficiency is poor. However, in order to cause the irradiation light to wrap around from all sides to the defect present on the surface to be inspected, the illumination unit is continuously arranged with light emitting elements so as to leave a dark surface of a predetermined shape inside. A plurality of layout patterns are combined. In the image obtained through the imaging camera from the surface to be inspected illuminated by such an illuminating unit, there will be many light emitting images of the light emitting elements, and noise will increase at the same time. The problem arises that it becomes difficult to distinguish.

  In view of the above situation, the object of the present invention is to provide a surface inspection technique in which erroneous detection of defects is suppressed while using illumination that causes the irradiation light to wrap around from all directions to defects existing on the surface to be inspected. Is to provide.

In order to solve the above-mentioned problem, an illuminating unit composed of a plurality of light emitting elements arranged continuously, an imaging camera that images a surface to be inspected illuminated by light irradiated by the illuminating unit, and an output of the imaging camera evaluates the signal consisting of a defect evaluation means for detecting a defect in the inspected surface, the surface defect inspection apparatus according to the present invention, the layout pattern leaving a dark surface of a predetermined shape on the inside surrounded by the light emitting element in the light emitting element are repeatedly arranged, one or the defect evaluation means, smoothing by average value filter to the image data generated from the output signal, after performing edge enhancement processing by Sobel filter a preprocessing unit for performing repeated twice defect emphasizing process which processes, brightness of generating a contrast image of the inspection surface from the image data output from the preprocessing unit An image generating unit, and a defect determination unit for detecting defects from the outputted light-and-dark image from the dark image generation unit.

  In this configuration, for example, for the defect existing inside the irradiation point of the light emitting element group continuously arranged in a ring shape, that is, on the surface to be inspected opposite to the dark surface, the irradiation light from the entire circumferential direction of the defect. In the image data generated from the output signal of the imaging camera that captured the surface to be inspected, a pixel with a relatively high luminance value indicating a defect is mixed with a pixel with a relatively low luminance value indicating noise. Will do. By performing defect emphasis processing that performs smoothing processing after performing contour emphasis processing on such image data, the ratio between the luminance values of defects and noise, that is, the S / N ratio is improved, and the following two values are obtained: In the bright and dark image generated by the digitization process or the like, defects come to float with high reliability, and false detection in defect detection by the defect determination unit is reduced.

In particular, in a surface defect inspection such as a painted surface of an automobile body, a Sobel filter is used for the contour enhancement process and an average value filter is used for the smoothing process for image data acquired from an imaging camera, and It has been clarified by experiments of the inventor of the present application that the defect enhancement process is repeated twice to generate a bright and dark image that causes the defect area to rise well in the subsequent process. The Sobel filter has ((1,0, -1), (2,0, -2), (1,0, -1)) operators in the horizontal scanning, and ((1,1, 2,1), (0,0,0), (-1, -2, -1)) with an edge emphasis filter, and the average filter focuses on the average value of the target pixel and the surrounding eight pixels It is a smoothing filter that uses a moving average method that uses new pixel values of pixels.

In the present invention, a surface defect inspection method using the inspection algorithm of the surface defect inspection apparatus is also a subject of rights, and an illumination unit composed of a plurality of light emitting elements arranged in succession is used. The surface defect inspection method according to the present invention for detecting defects on the surface to be inspected by evaluating an output signal of an imaging camera that images the surface to be inspected illuminated by irradiation light is a luminescent image corresponding to the layout of the light emitting elements. And the step of applying the irradiation light from the illumination unit to the surface to be inspected so that a dark image is formed inside the light-emitting image , and the image data generated from the output signal of the imaging camera , and performing repeated twice defect emphasizing process for performing smoothing processing by the average value filter after performing edge enhancement processing by the Sobel filter, the defect strong Generating a contrast image of the inspection surface from the image data processed, and a step of detecting a defect from the generated contrast image. Of course, such a surface defect inspection method also Ru can be obtained all the operational effects mentioned in the surface defect inspection apparatus described above.
Other features and advantages of the present invention will become apparent from the following description of embodiments using the drawings.

  As an example of the surface defect inspection apparatus according to the present invention, FIG. 1 shows a schematic configuration diagram of an apparatus for inspecting a painted surface of an automobile body 1 after completion of a painting process conveyed to the left side of a sheet by a conveyor 2. ing. The surface defect inspection apparatus includes an illumination unit 3 that irradiates a painted surface of an automobile body 1 that is an inspection surface with illumination light as inspection light, and an imaging camera 4 that images the inspection surface illuminated by the illumination unit 3. A controller 5 that evaluates the presence of defects on the surface to be inspected using an output signal from the imaging camera 4 and outputs the evaluation defects, and a monitor as an output device connected to the output unit 10 of the controller 5 12 and the printer 13. The controller 5 includes an illumination / shooting control unit 9 that controls the illumination unit 3, and an image input unit 7 that takes in an output signal from the imaging camera 4 and develops it in the memory 8 as digital image data (hereinafter simply referred to as an input image). Defect evaluation means 6 for performing defect evaluation using the input image is provided. Further, the controller 5 is connected to the host computer 14 as a host controller of the surface defect inspection apparatus via the communication unit 11 so as to be able to transmit data. The host computer 14 stores information on the automobile body 1 to be inspected and operation information on the conveyor 2 that are downloaded to the controller 5 as necessary. Further, defect information on the painted surface generated by the controller 5 is also stored. Is also uploaded from the controller 5 to the host computer 14 and stored therein. Further, a projector 15 or a printer controlled by a terminal connected to the host computer 14 via a network is provided in the inspection / verification station, and is based on defect information sent from the controller 5 of the surface defect inspection apparatus via the host computer 14. The defect position is instructed to the inspector.

  The light emitting surface 3a of the illuminating unit 3 and the lens surface 4a of the imaging camera 4 are arranged so as to face the surface to be inspected of the automobile body 1 conveyed by the conveyor 2, or the light emitting surface 3a and Following control is performed with respect to the surface to be inspected so that the vertical line of the lens surface 4a and the vertical line of the surface to be inspected are as coincident or parallel as possible.

  The illuminating unit 3 has a large number of light emitting elements (in this embodiment, LED elements are used and hence referred to as LED elements hereinafter. Of course, the light emitting elements of the present invention are not limited to LED elements, but other light emitting elements). 30 may be a net-like (ring-shaped) layout pattern that leaves a hexagonal space, and the hexagonal layout pattern is repeated continuously (between adjacent LED elements 30). It has a configuration arranged). The space left by the LED elements 30 arranged in a hexagonal mesh shape is referred to herein as a dark surface 31 and is a black or dark plate surface.

  Although many dark surfaces 31 appear by the LED elements 30 arranged in a net shape, the imaging camera 4 illuminates so that the lens surface 4a of the imaging camera 4 is positioned on the dark surface 31 that is located at the most center among them. Part 3 is incorporated. The number of imaging cameras 4 is appropriately determined depending on the size of the light emitting surface 3a of the illumination unit 3.

  The controller 5 uses a CPU as a core member and constructs a functional unit for performing various operations of the surface defect inspection apparatus by hardware and / or software, as shown in FIG. As a functional unit particularly related to the present invention, a pre-processing unit 60A that adjusts an input image developed in the memory 8 to an image suitable for defect detection, and a light-dark image generation that generates a light-dark image from the pre-processed input image It can be divided into a part 60B and a defect determination part 60C that finds a defect on the surface to be inspected using this bright and dark image.

The pre-processing unit 60A includes a luminance adjustment unit 61a that performs luminance adjustment on the input image, an edge enhancement processing unit 61c that performs edge enhancement processing on the input image that has been subjected to luminance adjustment using a Sobel filter, and an edge enhancement. Here, a smoothing processing unit 61b that performs smoothing processing on the processed input image using an average value filter is included. The luminance adjustment unit 61a is not only for gamma adjustment, but the luminance level of the light emitting image of the LED element obtained from the normal surface to be inspected in which the luminance level of the light emitting image included in the input image is a reference for each paint color or paint surface. It is also possible to adjust the brightness of each pixel area so as to reach the level. The average value filter implemented in the smoothing processing unit 61b is a well-known filter that uses a moving average method in which an average value of a target pixel and its surrounding eight pixels is used as a new pixel value of the target pixel. The Sobel filter implemented in the processing unit 61c has an operator of ((1,0, -1), (2,0, -2), (1,0, -1)) in the horizontal scanning, and the vertical filter This is a filter with operators ((1,2,1), (0,0,0), (-1, -2, -1)) in the direction scan. In other words, the contour enhancement processing is not performed on the smoothed image as described above, but the smoothing processing is performed on the image subjected to the contour enhancement processing.

  FIGS. 4A and 4B are a density graph showing a density value of a typical input image before undergoing preprocessing and a density graph showing a density value of a typical input image after preprocessing, respectively. Has been. In this density graph, the horizontal axis represents the pixel position corresponding to the irradiation image of the portion indicated by the line AA in the layout pattern of the illumination unit 3 shown in FIG. 2, and the vertical axis represents the pixel gradation value (density). It is taken from. By performing the defect emphasis process that applies the average filter after applying the Sobel filter, the noise is suppressed and the defect image and the light emission image are sharply clarified, resulting in an image suitable for the subsequent binarization process. I understand that. In other words, in the present invention, it is important to perform defect emphasis processing that applies an average value filter after applying a Sobel filter to a luminance-adjusted input image. In this case, it is preferable to repeat the defect emphasis process of applying the average filter after applying the Sobel filter twice.

  In the light / dark image generation unit 60B, a binarization threshold value determination unit 62a that determines a binarization threshold value by a statistical method from the density histogram of the input image, and the binarization threshold value determined by the binarization threshold value determination unit 62a. A binarization processing unit 62b is included that uses the input image preprocessed by the preprocessing unit 60B as a binarized image to generate a light and dark image. An example of the light and dark image binarized by the binarization processing unit 62 is shown in FIG. In this binarized bright and dark image, areas with high luminance are displayed in white, but LED element groups, which are light-emitting images continuously arranged in a hexagonal layout pattern, are continuously connected in a hexagonal shape. Displayed as a white outline, the painted surface area facing the dark surface 31 is displayed as a dark region, and in some cases, an existing coating defect is displayed as a white independent region that floats in the dark region due to diffuse reflection from the surrounding light. Is done. For this reason, in the defect detection, it is only necessary to find an area where luminance is protruding (a white area in this embodiment) and is not continuous in a predetermined pattern, that is, an isolated point in the binarized image. Become. A well-known image processing algorithm for searching for a continuous pixel or an isolated area while having a luminance value (density value) of a predetermined level can be used.

  However, due to the variation of the reflection characteristics with respect to the irradiated light due to the shape of the surface to be inspected here, as shown in an enlarged view in FIG. 6, the light emission image of the LED element 30 that originally appears as a continuous line is interrupted. May occur, and the interrupted portion may be erroneously detected as a defect. The defect determining unit 60C is substantially configured by a program so as to appropriately avoid such erroneous detection. That is, the defect determination unit 60C includes a defect candidate extraction unit 63 that detects a discontinuous independent pixel region composed of a predetermined number of pixels as an isolated point to be a defect candidate, and LED elements that are continuously arranged. The defect candidate selection unit 64 that excludes defect candidates included in the region showing the 30 light emission images from the defect candidates, and the unnecessary image regions such as isolated point regions and backgrounds excluded from the defect candidates by the defect candidate selection unit 64 are integrated. Then, an image mask generation unit 65 that performs mask processing as a non-defect determination target region, and a labeling process that assigns different labels (numbers) to different defect candidate regions in order to identify a plurality of defect candidate regions located outside the image mask. Based on the area information from the area calculation unit 67, the label setting unit 66 to be performed, the area calculation unit 67 for calculating the area of each labeled defect candidate area, And a defect determination unit 68 for writing the defect map to determine the candidate true defect. The defect candidate selection unit 64 checks whether or not a defect candidate has been extracted a predetermined number of times from images sequentially sent from the imaging camera 4 in order to select the defect candidate extracted by the defect candidate extraction unit 63. A defect candidate time-series determination unit 64a that prevents a suddenly occurring bright region from being recognized as a defect candidate, and a light emission image in which defect candidates (isolated points) extracted as shown in FIG. 6 are continuous. A light emission image discontinuous portion searching unit 64b is provided that prevents the discontinuity of the light emission image from being recognized as a defect candidate by checking whether or not it is located on the extension line. This search for the non-continuous portion of the luminescent image can be performed using a shape feature extraction algorithm or the like that extracts a dark region located in the extended line region of the discontinuity while tracing the continuous luminescent image pixels, An isolated point existing in this discontinuous region is excluded from defect candidates.

The procedure for defect evaluation of the painted surface by the defect evaluation means 6 configured as described above will be described below with reference to the flowchart of FIG.
First, the frame images sequentially sent from the imaging camera 4 via the image input unit 7 are taken into the memory 8 (# 01). The captured input image is subjected to preprocessing by the preprocessing unit 60A. In this pre-processing, first, the luminance (density value) is adjusted by the luminance adjusting unit 61a (# 02). In this case, the feature amount of the input image is required. It is preferable that the feature amount is obtained by dividing the input image by a predetermined number of partitions and using the maximum value of the density average value calculated for each partition as the feature amount. Subsequently, in order to perform the defect emphasis process that applies the Sobel filter and then the average value filter twice, 0 is substituted for the count variable n, and 2 is substituted for the variable a that defines the number of times (# 03). First, contour enhancement processing using a Sobel filter is performed (# 04). Further, a smoothing process using an average value filter is performed (# 05). Thereafter, the variable n is incremented by 1 (# 06), and the values of the variable n and the variable a are compared (# 07). If the values of the variable n and the variable a are different (# 07 No branch), the process jumps to step # 04 to apply the Sobel filter and then repeats the process of applying the average value filter. If the values of the variable n and the variable a are the same (# 07 Yes branch), the defect enhancement processing for applying the average filter after applying the Sobel filter is performed twice, so the processing by the light and dark image generation processing unit 60B to go into. When the binarization threshold value is determined by the binarization threshold value determination unit 62a (# 08), as described above, the pre-processed input image is binarized with the binarization threshold value to obtain a binarized image. (# 09).

  From the binarized input image, the defect candidate extraction unit 63 extracts an isolated bright pixel region having the number of pixels within a predetermined number (predetermined from the image resolution or the like) as a defect candidate (# 10). . Among the extracted defect candidates, defect candidates belonging to isolated points that are instantaneously and locally generated by disturbance light or the like are excluded from the defect candidates by the defect candidate time-series determination unit 64a (# 11), and further extracted defect candidates. Among these, defect candidates belonging to isolated points located in the discontinuous region of the light emission image are excluded from the defect candidates by the light emission image discontinuous portion searching unit 64b (# 12).

  The peripheral region including the discontinuous region of the light emission image found by the light emission image discontinuous portion searching unit 64b is the shape information of the vehicle body 1 as the inspection object transmitted from the host computer 14 and the conveyance position information by the conveyor 2 The image mask generation unit 65 performs mask processing as an unnecessary pixel area together with the background area other than the painted surface as the inspection surface determined based on (# 13). In this embodiment, since the transport position information obtained from the host computer 14 may be different from the actual position, the positional deviation of the automobile body 1 is checked in real time using a laser sensor or the like, and the image is displayed. The position of the mask is corrected (# 14).

  After selecting defect candidates and removing the background image in this way, the remaining defect candidates (isolated points) are labeled (# 15), and the area of the isolated points to which each label is assigned is calculated ( # 16) Only isolated points satisfying a preset area condition (whether or not having an area equal to or greater than the threshold) are determined as true defects (# 17), and their coordinate positions and sizes are stored in the defect map. Write (# 18).

  The procedure of the defect evaluation of the painted surface by the defect evaluation means 6 is completed as described above. When the inspection of the painted surface is completed through this procedure, the controller 5 of the surface defect inspection apparatus via the host computer 14 at the painted surface inspection / collation station. Among the defect maps sent from, the defect matching is performed using the defect map assigned with the ID that matches the ID of the automobile body carried into the painted surface inspection matching station. At this time, it is advantageous to operate the projector 15 so as to point out the defect location based on the corresponding defect map in order to facilitate the collation work by the inspector. Of course, the defect information based on such a defect map may be output to paper by the printer 13 connected to the output unit of the surface defect inspection apparatus, and the output paper may be directly attached to the automobile body 1.

Configuration diagram schematically showing a surface defect inspection apparatus according to the present invention Schematic diagram showing the illumination unit and imaging camera Functional block diagram showing the configuration of the defect evaluation means mounted on the surface defect inspection apparatus A density characteristic graph of an input image and a density characteristic graph when the process of applying a Sobel filter and then applying an average value filter to the input image is performed twice. Explanatory drawing explaining the binarized input image Explanatory drawing explaining the isolated point which exists in the discontinuous part of a light emission image Flow chart showing the procedure for defect evaluation of the surface to be inspected by the defect evaluation means

3: Illumination unit 4: Imaging camera 5: Controller 6: Defect evaluation means 30: Light emitting element (LED element)
31: Dark surface 60A: Preprocessing unit 60B: Bright / dark image generation unit 60C: Defect determination unit 61a: Brightness adjustment unit 61b: Averaging processing unit (average value filter)
61c: Outline enhancement processing unit (Sobel filter)
62b: binarization processing unit 63: defect candidate (isolated point) extraction unit 64: defect candidate selection unit 65: image mask generation unit 66: label setting unit 67: area calculation unit 68: defect determination unit

Claims (2)

An illumination unit composed of a plurality of light emitting elements arranged in succession, an imaging camera for imaging a surface to be inspected illuminated by light irradiated by the illumination unit, an output signal of the imaging camera is evaluated, and the subject In the surface defect inspection apparatus consisting of defect evaluation means for detecting defects on the inspection surface,
Wherein and the inside surrounded by the light emitting element is the light emitting element repeatedly arranged in the layout pattern leaving a dark surface of a predetermined shape, One or the defect evaluation means, the image data generated from the output signal On the other hand , a pre-processing unit that repeatedly performs defect emphasis processing that performs smoothing processing using an average value filter after performing contour emphasis processing using a Sobel filter, and image data output from the pre-processing unit is used to perform the above processing. A surface defect inspection apparatus comprising: a light / dark image generation unit that generates a light / dark image of an inspection surface; and a defect determination unit that detects a defect from the light / dark image output from the light / dark image generation unit. While using an illuminating unit composed of a plurality of light emitting elements arranged continuously, an output signal of an imaging camera that images the inspected surface illuminated by the irradiation light from the illuminating unit is evaluated, and the inspected surface In the surface defect inspection method for detecting defects,
A step of applying the illumination light from the illumination unit so that the dark image is formed on the inside surrounded by the light emission image and the emission image corresponding to the layout of the light emitting element to be inspected surface, an output signal of said image pickup camera A step of repeatedly performing a defect emphasis process for performing an edge emphasis process using a Sobel filter and then a smoothing process using an average value filter on the image data generated from the image data, and the image data after the defect emphasis process A surface defect inspection method comprising: generating a light / dark image of the surface to be inspected, and detecting a defect from the generated light / dark image.

Images