JP3185916B2 - Image binarization method and surface defect inspection device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method for binarizing an image and an apparatus for inspecting a surface defect of an object to be inspected, for example, a surface defect such as unevenness of a painted surface of an automobile body.

[0002]

2. Description of the Related Art A conventional surface defect inspection apparatus is disclosed in, for example, Japanese Patent Application Laid-Open No. 2-73139. This is to detect a defect on the inspected surface by imaging the inspected surface on which the light and dark pattern is projected by the light and dark pattern projecting means, and by differentiating and binarizing the imaged image signal.

[0003]

However, the conventional surface defect inspection apparatus as described above has the following problems. An image signal obtained by imaging the inspected surface on which the light and dark patterns are projected is differentiated and binarized by using an average value of the change signal as a threshold value, so that the defect contrast is low or the noise level is high. In some cases, for example, the threshold value is higher than the defect level, so that the defect cannot be detected.

FIGS. 1 and 2 show an example of a conventional example. FIG. 1A shows an original image, which shows that there are two circular black objects with respect to a white bright background, and below that is a luminance (density) cross section of a dotted line portion of the image. When edge detection such as differentiation is performed on such an original image, an image and a luminance cross section as shown in FIG. 1B are obtained. Therefore, when the image of FIG. 1B is binarized using the average luminance value as a threshold value, an edge component and a noise component can be separated, and a good binary image can be obtained. However, in the case of an object having a low contrast with respect to the background as shown in FIG. 2, since the level of edge detection is reduced, it is difficult to perform good binarization with the above-described conventional threshold value.

SUMMARY OF THE INVENTION The present invention has been made to solve the problems of the prior art as described above, and has an appropriate threshold value for reliably separating a signal component and a noise to be detected in an edge detection processing image by differentiation or the like. It is an object of the present invention to provide a binarization method using the method described above, and a surface defect inspection apparatus capable of accurately detecting a defect on a surface to be inspected by applying the binarization method to defect detection.

[0006]

SUMMARY OF THE INVENTION In order to solve the above-mentioned problems, the present invention is to extract only an area having a high luminance level from an image obtained by performing edge detection processing such as differentiation on an original image. When performing binarization with a predetermined threshold value, the edge detection processing image is divided into predetermined regions, and a predetermined calculation is performed for each region based on a physical quantity related to noise generated in the edge detection processing image. A value is calculated, and a threshold value satisfying a predetermined condition among threshold values calculated for each region is used to binarize the entire image.

[0007]

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

(First Embodiment) FIG. 3 and FIG.
It is a figure showing a 1st embodiment of the present invention. FIG. 3 is a processing flow of the present invention. FIG. 4 is an explanatory diagram using a luminance section of the present invention. As a result of performing processing such as differentiation to detect edges in the image as shown in FIG.
It is assumed that a luminance section as shown in (b) is obtained. Normal images have noise due to electrical noise, the surface condition of the subject, and the like. Therefore, in order to extract only edges by binarization, as shown in FIG.
It is sufficient that h is a value larger than the noise level and smaller than the edge level. A method for calculating such a threshold value Th will be described below with reference to FIG.

First, as shown in FIG. 5A, an upper limit value and a lower limit value of a range of a luminance level at which noise occurs are set. In the present embodiment, the maximum value of the noise is set as the upper limit (Nma
x), 0 is the lower limit. Therefore, the noise level is set between the upper limit value Nmax and the lower limit value 0, and the average value Navg in this range is obtained (FIG. 5B). This noise average value Navg is multiplied by a predetermined coefficient k as shown in, for example, Expression (1) to calculate a threshold value Th larger than the noise level and smaller than the edge level as shown in FIG.

Th = k × Navg (1) For example, if the noise average value Navg is doubled, it almost becomes the upper limit of noise. If 1.2 times the threshold value is used as the threshold value, k = 2 × 1 .2 = 2.4. If the noise level does not change, the threshold value Th is equal to the noise maximum value N
The value may be a value obtained by adding a predetermined offset value to max, but when the noise level fluctuates, it is better to calculate by multiplying a coefficient so that the threshold follows the fluctuation. Here, the above coefficient and calculation formula, and the upper limit of noise /
The lower limit may be determined experimentally in advance. By binarizing with the threshold value thus obtained, only the edge can be detected as shown in FIG.

(Second Embodiment) Next, a second embodiment of the present invention will be described.
The embodiment will be described with reference to FIG.

It is assumed that an edge detection process is performed on an original image as shown in FIG. 2A to obtain an image as shown in FIG. 2B. Perform a maximum filter on this image,
When the noise valley is converted to the surrounding bright luminance value, FIG.
The brightness cross section is as shown in FIG. What is a maximum filter?
The maximum value of the luminance of the pixels within the predetermined range is set as a new luminance value of the center pixel. A high-luminance (bright) portion spreads in a local area, and a low-luminance (dark).
Acts to shrink the part.

Next, a minimum value filter process is performed on the image (FIG. 6A) obtained by the maximum value filter process, and a high-luminance (bright) portion in the image is contracted. The brightness cross section is as follows. This is obtained by connecting the peaks of the noise in the original image, and this is used as the noise peak value. Therefore, if a threshold value is calculated by adding a predetermined offset value to the noise peak value or multiplying the noise peak value by a predetermined coefficient, it is possible to separate from the noise and detect only the edge as shown in FIG. In the present invention, the binarization threshold value is calculated from the physical quantity related to noise, and the calculation method, calculation formula, and the like are not limited to the present embodiment.

(Third Embodiment) Next, a third embodiment of the present invention will be described.
An embodiment will be described.

This means, for example, that the image is A, as shown in FIG.
The area is divided into four areas B, C, and D, and a threshold value is calculated for each area according to the procedure described in the first and second embodiments and binarized. The present invention is effective when the noise level is partially different and not constant in an image.
The number of divisions, the shape, and the size of the area may be determined according to the state of noise generation.

(Fourth Embodiment) Next, a fourth embodiment of the present invention will be described.
An embodiment will be described.

According to the present invention, the edge detection image is divided into arbitrary regions, and a predetermined condition among a plurality of threshold values calculated according to the procedure described in the first and second embodiments is determined for each region. The entire edge detection image is binarized by a threshold value that satisfies the condition. For example, as shown in FIG. 8, when an object and a background that does not require processing appear in an image at the same time, to perform edge detection / binarization processing on the object part, a threshold value is set only for the object area without including the background part. Needs to be calculated.

For example, as shown in FIG. 8, when there is a characteristic that the brightness is high and the noise level is high in the subject area of the image but the noise level is low in the background because the brightness is low and dark, the noise level is the highest in the four areas. Binarization may be performed using a threshold value, and the predetermined condition is the maximum value of the threshold values of a plurality of regions obtained. therefore,
If the processing unnecessary area (background) in the image is reflected on either the top, bottom, left or right of the image as shown in FIG. 8, the areas A, B, C, and D in the image are set as shown in FIG. If the threshold value is calculated and the entire image is binarized with the maximum value of the four threshold values, the edge of the subject can be detected satisfactorily. For example, when the subject is reflected on the left side of the screen as shown in FIG. 8A, the threshold value calculated in the area B in FIG. 9 is calculated, and in the scene shown in FIG. 8B, the threshold value is calculated in the area C in FIG. The threshold value becomes the maximum value, and binarization is performed. The number of divisions, the shape, and the size of the area may be determined according to the processing-unnecessary area (background) reflected in the image, the state of occurrence of noise, and the like.

(Fifth Embodiment) Next, a fifth embodiment of the present invention will be described with reference to FIG.

In FIG. 10, reference numeral 1 denotes an illumination means, which is arranged so as to project a light and dark pattern on the surface 5 to be inspected. Reference numeral 2 denotes an image pickup unit, such as a CCD camera, which picks up an image of the surface 5 to be inspected on which a light and dark pattern is projected in a predetermined field of view by attaching a lens having a predetermined focal length. The video signal output from the camera 2 is input to the image processing means 3, and the processing result is connected to the tracking processing means 4. Further, in this embodiment, it has been described that the received light image is processed at an arbitrary time while moving either the inspected surface or the imaging unit, but in this embodiment, the inspected surface 5 is not shown in FIG. The following description is based on the assumption that the object moves in the direction of the arrow. FIG. 11 is a schematic diagram illustrating an example of the illumination unit 1. Light source 10 in lighting box 101
2 is attached, and on the irradiation side of the lighting box 101, a lighting plate 103 having a matte black predetermined light and dark pattern formed on a diffusion plate is arranged. Therefore, the light from the light source 102 passes through the illumination plate 103, and a predetermined light / dark pattern is projected on the surface 5 to be inspected. Here, if there is an uneven defect 6 on the surface 5 to be inspected, light is irregularly reflected due to the unevenness of the defect.
The received light image of the camera 2 appears as an area where the defect has a brightness (brightness) different from that of the surroundings as shown in FIG. Therefore, the image processing means 3 performs edge detection processing of the camera light reception image by differentiation or the like, and performs a predetermined calculation based on the procedure described in the first and second embodiments, that is, the physical quantity relating to noise generated in the edge detection processing image. Is performed to calculate the threshold value and binarize it, or the procedure described in the third embodiment, that is, the edge detection processing image is divided into predetermined regions, and the noise generated in the edge detection processing image for each region The threshold value is calculated and binarized by performing a predetermined operation on the basis of the physical quantity related thereto, or the procedure described in the fourth embodiment, that is, the edge detection processing image is divided into predetermined regions, and A threshold value is calculated by performing a predetermined calculation based on a physical quantity related to noise generated in the edge detection processing image, and a predetermined condition is satisfied among the threshold values calculated for each region. Binarizing the entire image using the threshold,
A binary image as shown in FIG. 12B is obtained.

Further, when the surface 5 to be inspected moves as shown in FIG.
Tracing processing means 4
In, an object that matches the moving amount and moving direction of the inspected surface 5 is detected from a plurality of images at arbitrary times. As a result, the defect 6 on the inspection surface 5 moves in accordance with the movement of the inspection surface 5 and is determined to be a defect. In addition, other objects do not coincide with the movement of the inspection surface 5 and are not determined to be defects. Therefore, only the genuine defect 6 can be detected from the image.

(Sixth Embodiment) Next, a sixth embodiment of the present invention will be described.
The embodiment will be described with reference to FIG.

According to the present invention, the threshold value is determined based on the relationship between the surface condition of the surface to be inspected, for example, the surface condition such as painting and surface roughness, and the physical quantity relating to noise such as noise average value.

In this embodiment, a method of determining a threshold value based on the relationship between a paint color and a physical quantity relating to noise will be described, taking as an example a case where the surface to be inspected is painted. For example, it is assumed that the higher the brightness of the paint color itself, that is, the brighter the white or the like, the higher the luminance level, that is, the voltage level of the video signal. Therefore, the noise level increases in proportion to this voltage level. In such a case, the threshold value for separating the edge component and the noise component only needs to change according to the lightness of the paint color. Therefore, the relationship between the physical quantity N relating to noise and the threshold value Th is shown in FIG.
It looks like 3. Such a relational expression is experimentally obtained in advance, and the threshold value is calculated.

The method of calculating the threshold value and the calculation formula are not limited to the present embodiment.

[0026]

As described above, according to the present invention, an edge-detected image processed by differentiation or the like is divided into predetermined regions, and a threshold value is determined for each region based on a physical quantity relating to noise in the edge-detected image. , And binarizing the entire image using a threshold value that satisfies a predetermined condition among the threshold values calculated for each region, thereby reliably separating a signal component to be detected and noise from each other. The binarization method can be used, and by applying the above-described binarization method to a surface defect inspection apparatus, an effect that a defect on a surface to be inspected can be accurately detected can be obtained.

[Brief description of the drawings]

FIG. 1 is an image and a luminance cross-sectional view illustrating a conventional example.

FIG. 2 is an image and a luminance cross-sectional view illustrating a conventional example.

FIG. 3 is a diagram showing an example of a processing flow according to the first embodiment of the present invention.

FIG. 4 is a luminance cross-sectional view for explaining the first embodiment of the present invention.

FIG. 5 is a luminance cross-sectional view for explaining the first embodiment of the present invention.

FIG. 6 is a luminance sectional view for explaining a second embodiment of the present invention.

FIG. 7 is a diagram illustrating an example of image division according to a third embodiment of the present invention.

FIG. 8 is a diagram illustrating an example of an original image including a processing unnecessary area according to the fourth embodiment of the present invention.

FIG. 9 is a diagram illustrating an example of division of an image according to the fourth embodiment of the present invention.

FIG. 10 is a schematic configuration diagram showing a fifth embodiment of the present invention.

FIG. 11 is a schematic configuration diagram of a lighting unit according to a fifth embodiment of the present invention.

FIG. 12 is a diagram illustrating an example of a received light image and a processed image according to a fifth embodiment of the present invention.

FIG. 13 is a schematic diagram showing a relationship between a threshold, a physical quantity relating to noise, and a paint color according to the sixth embodiment of the present invention.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Illumination means 2 Imaging means 3 Image processing means 4 Tracking processing means 5 Inspection surface 6 Defect 101 Lighting box 102 Light source 103 Lighting plate

Continuing on the front page (72) Inventor Hiroshi Suzuki 2 Nissan Motor Co., Ltd., 2 Takaracho, Kanagawa-ku, Yokohama City, Kanagawa Prefecture (72) Inventor Masafumi Tsuji 2 Nihonsan Takaracho, Kanagawa-ku, Yokohama City, Kanagawa Prefecture (56 References JP-A-6-160302 (JP, A) JP-A-7-14,092 (JP, A) JP-A-8-86634 (JP, A) JP-A-8-94333 (JP, A) (58) Field surveyed (Int.Cl. ⁷ , DB name) G01B 11/00-11/30 102 G01N 21/84-21/958 G06T 1/00

Claims (3)

(57) [Claims] 1. An image obtained by performing edge detection processing such as differentiation on an original image and performing binarization with a predetermined threshold value in order to extract only a region having a high luminance level.
The edge detection processing image is divided into predetermined regions, a predetermined calculation is performed based on a physical quantity relating to noise generated in the edge detection processing image for each region, and a threshold value is calculated, and the threshold value calculated for each region is calculated. Wherein the entire image is binarized by using a threshold value satisfying a predetermined condition. 2. A surface defect inspection apparatus which irradiates a surface to be inspected with light, creates a light receiving image based on the reflected light from the surface to be inspected, and detects a defect on the surface to be inspected based on the light receiving image. An illumination means for forming a predetermined light and dark pattern on the surface of the object to be inspected; an imaging means for converting a received image obtained by imaging the surface to be inspected into image data of an electric signal; and edge detection processing for the image data The image obtained by performing the above is divided into predetermined regions, a predetermined operation is performed for each region based on a physical quantity relating to noise generated in the edge detection processing image, and a threshold value is calculated. Image processing means for binarizing the entire image using a threshold value which satisfies a predetermined condition among threshold values, and image processing means at any time while moving either the inspected surface or the imaging means Place If the defect candidate object present in each of the plurality of images processed in time series obtained therefrom matches the moving amount and moving direction of the inspection surface or the imaging means under predetermined conditions, the defect A surface defect inspection apparatus comprising: a tracking processing unit that determines a candidate object as a defect. 3. The surface defect inspection according to claim 2, wherein the binarization threshold in the image processing means is determined based on a relationship between an object amount relating to noise and a surface state of the inspection surface. apparatus.