KR20230139166A - Inspection Method for Wood Product - Google Patents

Abstract

The defect inspection method for wood products involves taking color photos of the wood product using a photography means, obtaining a brightness histogram of the color distribution of the color image captured by the image processing means using an image processing means, and determining the brightness of the color distribution of a normal wood product as determined in advance. A color distribution of the color image that deviates by a predetermined value or more from a reference value based on the average and standard deviation of the histogram is set as an abnormal color distribution, and an area having the abnormal color distribution is determined on the surface of the wood product photographed by the photographing means, If the area with the abnormal color distribution is wider than a predetermined value, it is detected as a defect in the wood product, and in the case where the inspection target is a wood product with a defect area relatively small with respect to the inspection target area, the color distribution of the normal wood predetermined above is detected. As such, the image distribution acquired for each inspection object is changed and used each time.

Description

{Inspection Method for Wood Product}

The present invention relates to a method for inspecting defects in wood products.

In general, in order to manufacture wood products, a log is cut with a knife to continuously obtain veneers with a thickness of several millimeters. These veneers are aligned to a predetermined size, dried, and then multiple veneers are glued together with an adhesive to integrate them. In these manufacturing processes, depending on the degree of location, number, and area of defects such as defects due to discoloration of the wood surface, distortion, places where knots in the veneer become holes, cracks, etc. due to discoloration of the wood surface, which affect the quality of the veneer, It is necessary to select those that make up the surface layer of plywood, that is, those that have few cosmetic defects, and those that make up the inner layer of plywood, that is, those that do not cause problems even if there are many defects. This selection can be divided into, for example, 5 to 7 stages.

Conventionally, selection of what constitutes the surface layer of plywood and what constitutes the inner layer of plywood is performed by visually judging the veneer being conveyed on a conveyor by an operator.

In addition, as a conventional method of inspecting defective parts of wood, wood is photographed with a color CCD camera, the image signal is binarized by comparing it with the reference color of rosin and discoloration using a color image extraction device, and the binarized image matches the measurement target area. There was a technology to detect defective areas such as adhesive resin such as rosin, corrosion, or discoloration by labeling and comparing with the judgment value.

The above-described prior art had problems such as visual judgment, where judgments made by humans were inconsistent and inaccurate, and productivity was poor because the speed of the conveyor could not be increased.

The present invention photographs a wooden product using a photographing means, determines the color distribution of the color image photographed by the photographing means using an image processing means, compares the obtained color distribution with a predetermined color distribution of normal wood, and determines the obtained color. A distribution whose distribution deviates from the color distribution of the normal wood by a predetermined value or more is designated as an abnormal color distribution, and a spot where the abnormal color distribution is greater than a predetermined value in an area on the wood surface photographed by the photographing means is detected as a wood defect. For this reason, we provide a defect inspection method for wood products that can accurately detect defects caused by discoloration of the wood surface, which affect the quality of wood, using color distribution.

In one embodiment, a method of inspecting defects in a wood product involves taking color photos of the wood product using a photography means, obtaining a brightness histogram of the color distribution of the color image captured by the image processing means using an image processing means, and obtaining a predetermined normal wood product. The color distribution of the color image that deviates by a predetermined value or more from a reference value based on the average of the brightness histogram of the color distribution and the standard deviation of the distribution is set as an abnormal color distribution, and the abnormal color distribution is defined as the abnormal color distribution on the surface of the wood product photographed by the photographing means. Find the area with the abnormal color distribution, and if the area with the abnormal color distribution is wider than a predetermined value, it is detected as a defect in the wood product. In the case where the inspection target is a wood product with a defect area that is relatively small with respect to the inspection target area, the predetermined value is determined above. As the color distribution of normal wood, the image distribution acquired for each inspection object is changed and used each time.

In the method of inspecting defective wood products, the brightness histogram of the color distribution of the predetermined normal wood product follows a normal distribution as a whole, and the overall normal distribution is estimated from the cumulative frequencies of some areas.

The method for inspecting wood products for defects according to the present invention uses image processing means to determine the color distribution of a color image captured by a photographing means, compares the obtained color distribution with a predetermined color distribution of normal wood, and determines the obtained color distribution. Since a color distribution that deviates from the color distribution of the normal wood by a predetermined value or more is considered an abnormal color distribution, and a color distribution in which the abnormal color distribution is greater than a predetermined value in an area on the wood surface photographed by the photographing means is detected as a wood defect, Defects caused by discoloration of the wood surface that affect the quality of wood can be accurately detected using color distribution.

In the case where the inspection target is wood with a defect area that is relatively small relative to the inspection target area, the color distribution of normal wood is set in advance, and the image distribution acquired for each inspection target is changed each time, so the predetermined color distribution is used. The color distribution of normal wood can be easily obtained.

Since the brightness histogram of the color distribution of the captured color image is obtained and the brightness abnormality is detected, the brightness abnormality such as pressing can be easily detected.

Since the brightness histogram of the predetermined color distribution of normal wood is assumed to follow a normal distribution as a whole and estimates the overall normal distribution from the cumulative frequency of some areas, the color distribution of normal wood can be determined from the wood being inspected even if the color distribution of normal wood is not determined in advance. The color distribution of can be estimated.

1 is an explanatory diagram of a single plate sorting device of the present invention.
Figure 2 is an explanatory diagram of the image processing device of the present invention.
Figure 3 is an explanatory diagram of converting the color of each point of image g into the HS plane according to the present invention.
Figure 4 is an explanatory diagram of the relaxation method of the present invention.
Figure 5 is an explanatory diagram of color distribution in the HSV color space of the present invention.
Figure 6 is an explanatory diagram of pixel distribution in the specific brightness v plane of the present invention.
Figure 7 is an explanatory diagram of the cumulative frequency F(x) of the normal distribution of the present invention.
Fig. 8 is an explanatory diagram of pixel distribution (normal distribution shape) in the central axis direction of the present invention.
Figure 9 is an explanatory diagram for estimating the average value Vm when the distribution shape of the present invention is abnormal.
Figure 10 is an explanatory diagram for the case where the area of the brightness abnormality part of the present invention is large.

The present invention described below can be modified in various ways and can have various embodiments. Specific embodiments will be illustrated in the drawings and described in detail in the detailed description.

However, this is not intended to limit the present invention to specific embodiments, and should be understood to include all transformations, equivalents, and substitutes included in the spirit and technical scope of the present invention. In describing the present invention, if it is determined that a detailed description of related known technologies may obscure the gist of the present invention, the detailed description will be omitted.

The terms used in this application are only used to describe specific embodiments and are not intended to limit the invention. Singular expressions include plural expressions unless the context clearly dictates otherwise. In this application, terms such as “comprise” or “have” are intended to designate the presence of features, numbers, steps, operations, components, parts, or combinations thereof described in the specification, but are not intended to indicate the presence of one or more other features. It should be understood that this does not exclude in advance the possibility of the existence or addition of elements, numbers, steps, operations, components, parts, or combinations thereof.

Additionally, terms such as first and second may be used to separately describe various components, but the components should not be limited by the terms. The above terms are used only for the purpose of distinguishing one component from another.

In addition, when at least two different embodiments are described in the present application, all or part of the components of each embodiment can be merged and used interchangeably without departing from the technical spirit of the present invention. .

Figure 1 is an explanatory diagram of a single plate sorting device. In Figure 1, 1 is an image processing device (image processing means), 2 is a sorter control device, 3 is an operating panel, 4 is a belt conveyor, 5 is a lighting for transmitted light, 6 is a lighting for reflected light, 7 is a distribution device for each grade, and 8 is a lighting device for transmitted light. Line sensor camera (photography means), 9 shows a veneer (wood).

(1) Description of veneer sorting device

Figure 1 is an explanatory diagram of a single plate sorting device. In Figure 1, the overall configuration of the single sheet sorting device is shown. The single plate sorting device includes an image processing device (1), a sorter control device (2), an operating panel (3), a belt conveyor (4), lighting for transmitted light (5), lighting for reflected light (6), distribution device for each grade (7), and a line sensor camera (8) are installed.

The image processing device 1 is an image processing means that processes image data from the line sensor camera 8 and outputs processing results such as single plate quality grade to the sorter control device 2. The sorter control device 2 outputs a sorter conveyor control signal, such as operation and stop of the conveyor, and a control signal for the grade distribution device 7 based on the output of the image processing device 1. The operating panel 3 is an operating panel that performs operations such as changing the setting values of the image processing device 1 and controlling the sorter control device 2. The belt conveyor (4) is a conveyance means that conveys the single plate (9). The transmitted light illumination 5 is an illumination means (light source) such as an LED for detecting holes, cracks, etc. in the end plate 9, and uses lighting of a different color (for example, green lighting) than the reflected light illumination 6. use. This is to distinguish the reflected light from the reflected light 6 (differentiated by color and intensity) and detect holes (knot holes), cracks, etc. in the veneer. The lighting 6 for reflected light is an illumination means (light source) such as LED for detecting the reflected light of the end plate 9, and usually uses white lighting. The line sensor camera 8 is a photographing means that captures images of the lines of the end plate 9.

The operation of this plate sorting device is to photograph the plate 9 sent from the belt conveyor 4 with the line sensor camera 8 and output image data to the image processing device 1. The image processing device 1 processes this image data and outputs processing results, such as single plate quality grade, to the sorter control device 2. The sorter control device 2 outputs a control signal to the grade distribution device 7 to sort the single plate 9 by grade. This selection includes the number of insect holes, number of holes and missing knots, number of living knots, number of dead knots, number of splays, number of cracks, number of bark pockets, blueness number, and warpage value. This is done depending on the degree of etc. and their size (area).

(2) Description of image processing device

Figure 2 is an explanatory diagram of an image processing device. In Figure 2, the image processing device includes three line sensor cameras (8a, 8b, 8c), bases for camera image acquisition (11a, 11b, 11c), laser markers (12a, 12b), and laser drivers (13a, 13b). , the main computer 14 is installed.

The line sensor cameras 8a, 8b, and 8c are three cameras that divide the single plate 9 into three parts in a direction perpendicular to the transport direction and take pictures in color. Each time the camera image acquisition bases 11a, 11b, and 11c receive one line of images from each line sensor camera, they digitize and transmit the image data to the main computer 14.

The laser markers 12a and 12b irradiate thin light rays from the line sensor cameras 8a, 8b and 8c in the transport direction of the veneer as a mark for compositing (combining) each image. This irradiated light ray can be irradiated as a thin ray of a different color from the color of the veneer (wood) (for example, a red laser monochromatic ray) so that it can be easily removed in later processing.

The laser drivers 13a and 13b are connected to an AC power source to drive the laser markers 12a and 12b. The main computer 14 is equipped with processing means, storage means, output means, etc., and performs image processing (image compositing, knot detection, defect detection processing, etc.) of the single plate 9. Here, the camera image acquisition bases 11a, 11b, 11c and the main computer 14 serve as image processing means.

The operation of the image processing device is to illuminate the conveyed single plate 9 with rays from the light source 5 for transmitted light and the light source 6 for reflected light, and to use bases 11a, 11b, and 11c for camera image acquisition, and a line sensor camera. Each time one line of images is received from (8a, 8b, 8c), the data is transmitted to the main computer 14. The main computer 14 corrects the received images, detects the warpage value, and sequentially combines the images. Finally, when the camera image acquisition bases 11a, 11b, and 11c have finished accepting images, the main computer 14 has almost finished compositing color images and converting black and white images. Then, the main computer 14 combines the images from the camera image acquisition bases 11a, 11b, and 11c, in which the single plate image is divided into three.

Here, the end plate 9 is divided into three by irradiating laser marks from the laser markers 12a and 12b, and the line sensor cameras 8a, 8b and 8c combine the line images up to each laser mark to simply create an image. allows you to combine.

Additionally, in order to improve the image processing speed, the knot exploration process may be performed using a black-and-white grayscale image with a large number of pixels, and the color image, such as the exploration of dead knots, may be performed using a reduced image (reducing the number of pixels).

Hereinafter, the operation of the image processing device will be explained by dividing it into processing during shooting and processing after shooting.

<Description of processing being filmed>

Image data captured by the line sensor cameras 8a, 8b, and 8c is transmitted to the main computer 14 for each line and synthesized as one entire image.

· Processing of camera image acquisition base (11a, 11b, 11c)

A one-line color image is received from the line sensor cameras 8a, 8b, and 8c, the position of the laser mark (joining position) is detected, and the one-line color image is transmitted to the main computer 14 along with the information.

· Processing of the main computer (14)

While correcting the arrived 1-line color image, the warpage value is detected and synthesized based on the position information (laser mark). Here, if there is bending in the inspection object, the detection trace of the position of the laser mark is distorted in a non-linear form, so the bending value can be detected from the amount of distortion.

At this stage, the shooting is completed in the camera image acquisition bases 11a, 11b, and 11c, and the final one-line color image is received, the synthesis of all color images is completed in the main computer 14. Additionally, in order to effectively utilize the shooting time, processing that can be performed for each line, such as black-and-white conversion or reduction processing, can be performed simultaneously and in parallel.

<Explanation of processing during image analysis after shooting>

· Processing of camera image acquisition base (11a, 11b, 11c)

Wait until the arrival of the next plate (single plate) is detected.

· Processing of the main computer (14)

Based on preset information such as the size and type of the target plate, knot detection processing and defect detection processing using transmitted light are performed according to the area to be calculated or set values, and finally classification including the warpage value. Process it. The results are displayed on a display device (not shown), and the results are output to the sorter control device 2.

In addition, in the above explanation, it has been explained that the camera image acquisition bases 11a, 11b, 11c and the main computer 14 in the image processing device use computers (PCs), but the number of computers used is It can be changed depending on the amount of data or the processing speed of the computer. Also, it can be processed with one computer.

In addition, we explained using three line sensor cameras, but depending on the size and type of the target plate and the processing power of the computer, one, two, or four or more cameras can be used.

(3) Description of detection of defective parts due to discoloration of veneer surface

Defects due to discoloration of the wood surface that affect the quality of wood can be detected by the following means and methods.

Defects due to discoloration are areas of discoloration caused by mold entering the wood from the outside, or in veneers used for plywood, burning by a dryer, traces of bark, or wood stains formed inside the wood. What appears on the surface, etc.

This is to enable detection of these as defective parts.

(Explanation of detection means for defective parts)

1) The surface of the wood is photographed with a color line sensor camera, and the image is received into a computer using an input device (input means).

2) The captured image is stored in the computer memory (storage means) as a color image in which each pixel is composed of RGB (red, green, blue).

3) The RGB image is converted into an HSV (hue, saturation, brightness) image by a computer image processing program (image processing device).

4) Defects are detected from the HSV image by the following method.

(Explanation of defect detection method)

The surface color of normal (healthy) wood of the same species is distributed in almost a specific range of saturation and color, regardless of its lightness or shade. However, defects such as mold cause differences in saturation and hue from the healthy color distribution due to differences in materials. In addition, defective parts such as depressions are distributed in significantly black (low-brightness) areas compared to the distribution of healthy colors.

Accordingly, the method is characterized by examining the saturation, color deviation, and brightness deviation of the color distribution of the surface of the wood to be inspected and the color distribution of the surface of the healthy wood, and detecting the portion with the large deviation value as a defective part.

(4) Description of the method for obtaining the color distribution of the standard sound wood surface

1) For the tree species being inspected, the surface of the healthy wood is photographed with a color line sensor camera.

2) For the same tree species above, take multiple photos (about 20 or more) of the surface in different states to obtain sufficient statistical precision.

3) Each pixel color of all the above images is arranged in a 3D color space in the computer memory to create a 3D color distribution.

As a three-dimensional color space, RGB (red, green, blue space) is collected, HSV (hue, saturation, brightness color space) or Lab color space (“L” is brightness, “a” is from green to red, “ b" refers to a color element from blue to yellow), etc. can be used.

4) Obtain the 2-dimensional distribution for each isometric view of the 3-dimensional color distribution, and obtain the point representing the maximum frequency.

5) By changing the brightness step by step, a curve that approximately connects the maximum frequency points of 4) above can be obtained. This curve is taken as the standard central axis of the three-dimensional color distribution.

For example, when the brightness ranges from 0.0 to 10 in the HSV color distribution, the two-dimensional distribution of color and saturation of pixels with equal brightness values for each brightness division 002 is obtained, and the maximum frequency points are connected. Obtain a curve and use this as the reference central axis of the 3D color distribution. At the same time, the standard deviation σc(v) of the two-dimensional distribution of color and saturation is also obtained.

6) If it is known in advance that the defective area is relatively small with respect to the inspection target area, this reference distribution can be replaced with an image distribution acquired for each inspection target. In other words, it is good to know the mean and standard deviation values of the distribution of healthy parts.

(5) Description of defect inspection method

1) Photograph the wooden surface of the inspection target with a color line sensor camera.

2) Each pixel of the image is placed in a 3D color space and a 3D color distribution is created.

3) Calculate the color deviation value from the standard central axis of the 3D color distribution as follows.

For example, the pixel at the x,y position of the target image is g[x,y], and the color in the HSV color distribution space is

Color value:h(g[x,y])

Saturation value: s(g[x,y])

Brightness value:v(g[x,y])

, and the coordinates of the reference center axis at the specific brightness v of the reference center axis of the three-dimensional color distribution obtained previously are

Hue value: H(v), saturation value: S(v)

If the horizontal axis of the light plane is X and the vertical axis is Y, it becomes the same as Figure 3. Fig. 3 is an explanatory diagram of converting the color of each point in image g onto the HS plane. In Fig. 3, the pixel g[x,y] of the target image tree is converted to orthogonal coordinates X2,Y2 on the HS plane. Also, the distribution of colors (see net shape) is not a circle but is distributed in various shapes, but the standard deviation is approximately a circle.

Here, the orthogonal coordinates X1 and Y1 of the reference central axis coordinates H(v) and S(v) are as follows.

X1=S(v)·cos(2π·H(v)/360)

Y1=S(v)·sin(2π·H(v)/360)

The orthogonal coordinates X2, Y2 of h(v), s(v) of pixel g[x,y] are as follows.

X2=s(v)·cos(2π·h(v)/360)

Y2=s(v)·sin(2π·h(v)/360)

The square spatial distance d from the reference central axis is obtained as follows.

Here, the color deviation value Zc[x,y] is as follows.

Here, σc(v) is the standard deviation σc(v) of the two-dimensional distribution of hue and saturation at the brightness v of the reference central axis. βc is a coefficient that determines how many times σc(v) a color deviates from the reference central axis is considered abnormal, and takes a value of about 1.0 to 2.0, for example.

In addition, the spatial distance deviation can be obtained in the same way even if other color distributions such as Lab color space are used.

4) Next, in order to find the area of the actual defect, only pixels in the color distribution space where the color pixels that deviate from the standard locally form a cluster in the original wood pixel g[x,y] must be selected. There is a need. Here, discontinuous isolated points are removed while considering the colors of surrounding pixels, and areas with large deviations are emphasized, and a method called a general relaxation method as an image processing technique can be used.

As an example, the defective area due to color abnormality can be determined by a relaxation method that uses the color deviation value Zc[x,y] from the reference central axis as the initial label (see explanation in FIG. 4).

(Explanation of detection of defective parts due to abnormal brightness)

5) Obtain a histogram in the direction of the standard central axis (brightness axis) of the 3D color distribution.

6) The brightness histogram of the sound part is assumed to be a normal distribution (Gaussian distribution) with a mean value Vm and a standard deviation σv, and the brightness deviation value Zv[x,y] is obtained as follows.

Zv[x,y]=｜Vm-g[x,y]v｜/(σv×βv)

βv is a coefficient that determines how many times σv the brightness deviates from the brightness average value Vm is considered ideal, and takes a value of about 1.0 to 4.0, for example.

The overall deviation value Zt[x,y] of color and brightness is as follows.

Zt[x,y]=Zc[x,y]+Zv[x,y]

If the brightness histogram does not show a normal distribution due to camera characteristics, etc. (for example, bright parts with brightness close to 1.0 are saturated), the average value Vm and standard deviation σv may not be obtained accurately. In this case, the histogram follows the standard normal distribution, and the normalized cumulative probability distribution function F(x) can be expressed as follows.

(Here, x is the brightness, μ is the average value of brightness, and σ is the standard deviation.)

The brightness histogram is integrated starting from the side with the lowest brightness, and the integrated value is divided by the total number of pixels (accumulated frequency) to obtain the brightness corresponding to the following p1, p2, p3, and p4, respectively. V1, V2, V3, Make it VM.

p1=F(μ-2.0σ)=0.0228

p2=F(μ-1.0σ)=0.1587

p3=F(μ-0.5σ)=0.3085

p4=F(μ)=0.5

Additionally, the effective areas Vmin and Vmax that can be taken by V1, V2, V3, and Vm are empirically obtained from reference wood, etc., and can be set as Vmin = 0.25 and Vmax = 0.9, for example.

a) Among V1, V2, V3, and Vm, if you find the brightness position corresponding to V2:F(Vm-σ)=0.1587 and Vm:F(Vm)=0.5 within the effective area, the estimated average value x=Vm and standard deviation σv is obtained (see Figure 8)

b) However, in cases where the brightness distribution is smaller than x=μ and the distribution shape is disturbed, Vm is outside the effective area. At that time, find V1:F(Vm-2.0σv) and V2:F(Vm-1.0σv) that exist within the effective area,

σv and Vm can be estimated (see Figure 9).

c) Or, if the brightness abnormality becomes relatively large,

V1:F(Vm-2.0σv) and V2:F(Vm-1.0σv) are outside the effective area. So, from the brightness values V3:F(Vm-0.5σ)=0.3085 and Vm:F(Vm)=0.5 that exist within the effective area, σv=(Vm-V3)×2 can be used to estimate σv and Vm ( (see Figure 10)

Using this method, the average value and standard deviation of the sound part can be obtained without using reference wood and regardless of the distribution shape (also, if reference wood is used, no processing is performed to obtain it). (without using the average value and standard deviation)

7) Next, in order to obtain the area of the actual defective area, it is necessary to select only pixels in the color distribution space where pixels of colors away from the standard locally form a cluster with the original wood pixel g[x,y]. . There, it is possible to use a method called a general relaxation method as an image processing technique, which removes discontinuous isolated points while considering the colors of surrounding pixels and emphasizes parts with large deviations.

As an example, a defective area due to color and brightness abnormalities can be determined by a relaxation method (see explanation in FIG. 4) using the overall deviation value of color and brightness Zt[x,y] as an initial label.

Until now, in automatic quality inspection of wood, defects have been detected only by the lightness and darkness of the surface color or by specifying a specific color, but this cannot respond to changes in bright areas or colors, and all products have become non-defective products.

Among the surface colors of wood, many parts that cause discomfort to the human eye and affect quality are not the wood's original natural colors, and this appears as a difference in color distribution in the three-dimensional color space. A unified method of separating and detecting these differences makes it possible to detect defective parts with high precision.

In addition, since the color of fungi that affect the quality of wood varies depending on the type of wood and the production area, it was difficult to detect all of them with high precision using a single method. However, according to the present invention, even if the type of wood being inspected is different, detection is possible without changing the reference central axis coordinate value in most cases. Even if the detection accuracy deteriorates depending on the tree species, the detection accuracy can be restored simply by changing the reference central axis coordinate of the 3D color distribution, which is the initial value.

In addition, although it has been difficult to detect tree pulp and bark through visual inspection using image processing, detection of these substances is now possible.

Since it is now possible to detect blackened parts of the bark, etc. with high precision, it becomes easier to determine whether a knot is alive or dead by determining whether or not the blackened bark remains.

(6) Description of relaxation method

Figure 4 is an explanatory diagram of the relaxation method. Hereinafter, explanation will be given according to processes S1 to S3 in FIG. 4.

In this process, the defect probability Pi(x,y) is set for each pixel g(x,y) of the target image. Here, Pi(x,y) is the defect probability for pixel g(x,y) after the ith repetition.

S1: The image processing device assigns an initial probability P0(x,y(O to 10)) to each pixel g(x,y) of the target image and moves to process S2. Here, P0(x,y) is calculated from the color deviation value Z(x,y) as follows.

P0(x,y)=Z(x,y):0<Z<1.0

1.0:Z>=1.0

S2: Image processing device for all pixels

If (0.0<Pi(x,y)<1.0)

Find the probability average value <P> of the neighboring pixels of Pi(x,y),

Pi+1=Pi+α(<P>-Pi)

(α is the influence coefficient of surrounding pixels, approximately 1 to 4)

Update the probability and move to process S3.

When (Pi(x,y)≤0.0 or Pi(x,y)≥1.0)

Pi+1=Pi

, the probability is not updated, and the process moves to process S3.

S3: The image processing device checks the convergence conditions.

If, for Pi(x,y),

The repetition number I is greater than the specified number (I > specified number),

If the ratio of the number of pixels of Pi = 0.0 and pi = 1.0 to all pixels is greater than the specified rate (> specified rate), the process is terminated.

Otherwise, repeat process S2.

Here, the repetition designation number is approximately 10 times, and the designation rate of the number of pixels of Pi = 0.0 and pi = 1.0 for all pixels is approximately 99%.

(7) Description of color distribution in HSV color space

Figure 5 is an explanatory diagram of color distribution in the HSV color space. In Figure 5, the upward direction is brightness (V: here, V = 0.0 to 1.0), the radial direction of the same brightness plane is saturation (S: here, S = 0.0 to 1.0), and the circumferential direction is color (H: here, H = 0). ° to 360°). The color distribution of sound wood has a large upper and lower color distribution area, and the central axis (reference central axis) of this color distribution is indicated by an upward arrow.

In addition, the color distribution of the color abnormality area, which is a discoloration area caused by mold or the like, is shown as a color abnormality area on the right. In addition, the brightness abnormality, such as scuffing caused by a dryer, is shown as a small area of color distribution at the bottom.

(8) Description of pixel distribution in a specific brightness plane

Figure 6 is an explanatory diagram of pixel distribution in a specific brightness v plane. In Fig. 6, the distribution of pixels of color abnormalities in the specific brightness v plane is shown. Here, the distribution of pixels in the color abnormality is distributed in a specific area (refer to the net-shaped part) with a standard deviation σc(v).

(9) Description of pixel distribution in the central axis direction

Figure 7 is an explanatory diagram of the cumulative frequency F(x) of a normal distribution. In Figure 7, a general normal distribution (distribution probability) is indicated by a dotted line, and the cumulative frequency (cumulative probability distribution function) F(x) is standardized by dividing the integrated value of the sound part of the wood by the total number of pixels N. Here, p1 = 0.0228 (μ-2σ), p2 = 0.1587 (μ-1.0σ), p3 = 0.3085 (μ-0.5σ), p4 = 0.5 (average value = μ).

Fig. 8 is an explanatory diagram of pixel distribution (normal distribution shape) in the central axis direction. In Figure 8, a histogram in the direction of the reference central axis (brightness axis) of the three-dimensional color distribution is shown. In this brightness histogram, the brightness corresponding to p1, p2, p3, and p4 is set to V1, V2, V3, and Vm, respectively, and the effective area that can be taken by the sound part of the wood surface is set to Vmin and Vmax.

In the case of this figure, if the brightness position corresponding to V2:F(Vm-σ)=0.1587 and Vm:F(Vm)=0.5 within the effective area can be found, the estimated average value x=Vm and standard deviation σv can be obtained. do.

Figure 9 is an explanatory diagram for estimating the average value Vm when the distribution shape is abnormal. In Figure 9, a histogram in the direction of the reference central axis (brightness axis) of the three-dimensional color distribution is shown. In this brightness histogram, if the brightness distribution of the sound part of the wood surface is disturbed in a part smaller than x=μ, Vm is outside the effective area. In this case, V1:F(Vm-2.0σv) and V2:F(Vm-1.0σv) that exist within the effective area can be obtained, and σv and Vm can be estimated from σv=V2-V1 and Vm=V2+σv.

Figure 10 is an explanatory diagram for the case where the area of the brightness abnormality part is large. In Figure 10, a histogram in the direction of the reference central axis (brightness axis) of the three-dimensional color distribution is shown. In this brightness histogram, when the area of the brightness anomaly on the wood surface becomes relatively large, V1:F(Vm-2.0σv) and V2:F(Vm-1.0σv) fall outside the effective area. So, from the brightness values V3:F(Vm-0.5σ)=0.3085 and Vm:F(Vm)=0.5 that exist within the effective area, σv=(Vm-V3)×2, and σv and Vm can be estimated. .

Likewise, if the brightness histogram does not show a normal distribution due to camera characteristics, etc. (for example, when the sensor sensitivity characteristics are not linear in the area where the brightness is close to 1.0), the average value Vm and standard deviation σv cannot be obtained accurately. There are cases. Even in this case, since the histogram follows the standard normal distribution, the entire distribution can be estimated from the histogram (two points) at the bottom of the distribution, and the average value Vm and standard deviation σv can be obtained.

As a result, in the image processing apparatus, the average value Vm and the standard deviation σv of the healthy part can be obtained (estimated) using two points among V1 to Vm, which are points within the effective area (Vmin to Vmax). The priority of the two points used is a point different from the average value Vm (one point from V1 to V3), and when the average value Vm is not within the effective area, two points from V1 to V3 are used.

(1O) Description of program installation

The image processing device (image processing means) (1), the sorter control device (sorter control means) (2), the camera image acquisition base (11a, 11b, 11c), the main computer (14), etc. can be configured with a program. , It is mainly executed by the control unit (CPU) and stored in main memory. This program is processed on a computer. This computer mainly consists of hardware such as a control unit, main memory, file device, output devices such as display device, and input device.

The program of the present invention is installed on this computer. This installation stores these programs on portable recording (storage) media such as floppy disks and magneto-optical disks, and allows access to the recording media included in the computer through a drive device or network such as LAN. It is installed on the file device created on the computer. As a result, it is possible to easily provide a wood inspection device that can accurately detect defects caused by discoloration of the wood surface that affect the quality of wood using color distribution.

As described in detail above, the method for inspecting wood products for defects according to the present invention uses image processing means to obtain the color distribution of a color image captured by a photographing means, and determines the color distribution of normal wood in advance by determining the obtained color distribution in advance. Compared with, the obtained color distribution deviates from the color distribution of the normal wood by more than a predetermined value as an abnormal color distribution, and the abnormal color distribution is larger than a predetermined value in the area on the wood surface photographed by the photographing means. Since it is detected as a defect in wood, defects due to discoloration of the wood surface that affect the quality of wood can be accurately detected using color distribution.

In the case where the inspection target is wood with a defect area that is relatively small relative to the inspection target area, the color distribution of normal wood is set in advance, and the image distribution acquired for each inspection target is changed each time, so the predetermined color distribution is used. The color distribution of normal wood can be easily obtained.

Since the brightness histogram of the color distribution of the captured color image is obtained and the brightness abnormality is detected, the brightness abnormality such as pressing can be easily detected.

Since the brightness histogram of the predetermined color distribution of normal wood is assumed to follow a normal distribution as a whole and estimates the overall normal distribution from the cumulative frequency of some areas, the color distribution of normal wood can be determined from the wood being inspected even if the color distribution of normal wood is not determined in advance. The color distribution of can be estimated.

Meanwhile, the embodiments disclosed in these drawings are merely provided as specific examples to aid understanding, and are not intended to limit the scope of the present invention. It is obvious to those skilled in the art that in addition to the embodiments disclosed herein, other modifications based on the technical idea of the present invention can be implemented.

1...Image processing device (image processing means) 2...Selector control device
3...Operation panel 4...Belt conveyor
5...Lighting for transmitted light (lighting means) 6...Lighting for reflected light (lighting means)
7...Distribution device by grade 8...Line sensor camera (photography means)
9...wood

Claims (2)

Color photography of wooden products is carried out as a means of photography.
Using image processing means, a brightness histogram of the color distribution of the color image captured by the photographing means is obtained,
The color distribution of the color image that deviates by a predetermined value or more from a reference value based on the average of the brightness histogram of the color distribution of the normal wood product determined in advance and the standard deviation of the distribution is set as an abnormal color distribution,
Obtaining an area having the abnormal color distribution on the surface of the wood product photographed by the photographing means, detecting the area having the abnormal color distribution as a defect of the wood product if the area having the abnormal color distribution is wider than a predetermined value,
In the case where the inspection target is a wood product whose defect area is relatively small with respect to the inspection target area, a defect inspection method for wood products in which the image distribution acquired for each inspection target is changed each time as the color distribution of the predetermined normal wood. According to paragraph 1,
The brightness histogram of the color distribution of the predetermined normal wood product follows a normal distribution as a whole, and a defect inspection method for wood products in which the overall normal distribution is estimated from the cumulative frequency of some areas.