JP2005043260A - Visual inspection device of grain of rice - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a visual inspection device of a grain of rice capable of specifying a trace of rice bran adhering to the grain of rice by setting a threshold highly accurately. <P>SOLUTION: This device has an illumination means 2 for illuminating the grain of rice, an imaging means 4 for imaging the grain of rice 1 illuminated by the illumination means 2 by an imaging camera 3 to acquire an analog image signal, and an image conversion means 5 for performing binarization processing of the acquired analog image signal and converting it into a digital signal showing density. In the device, image processing of a digital image signal acquired by the image conversion means 5 is performed, and a density distribution histogram on the image density is produced, and the density distribution acquired from the density distribution histogram is accumulated to thereby produce an accumulative density distribution histogram, and the threshold T is determined from the accumulative density distribution histogram, and binarization determination relative to the grain of rice and the adhering rice bran is performed. <P>COPYRIGHT: (C)2005,JPO&NCIPI

Description

The present invention relates to a rice grain appearance inspection apparatus.

  Conventionally, a grain sorter has been proposed that sorts grains such as rice according to the reflectance and transmittance of light irradiated on the surface thereof. As this grain sorter, light is irradiated to the granular material flowing continuously, the appearance of the granular material is imaged, the obtained image signal is processed to obtain the concentration distribution, and a predetermined threshold is used as a reference There is one that creates a binarized image and determines whether or not it is a defective particle based on the binarized image (see, for example, Patent Document 1).

  However, when a granular object is imaged with an imaging device under illumination light and image processing is performed, the illuminance changes due to voltage fluctuations of the light source, etc. A drift phenomenon may occur. Therefore, it is necessary to set the threshold value for binarizing the image signal to the optimum value for these changes, but it is difficult to automatically set the threshold value to the optimum value.

  In addition, when identifying wrinkles attached to rice grains, the proportion of the image area of the attached wrinkles to the image area of the captured rice grains is only about 1 to 2%, so a fine area such as a pinhole In this case, the conventional grain sorter cannot set a high-accuracy threshold, and it is difficult to identify wrinkle adhesion.

JP2001-145855

  In view of the above problems, an object of the present invention is to provide a rice grain appearance inspection apparatus capable of setting a threshold value with high accuracy and identifying a small amount of rice cake attached to rice grains.

In order to solve the above-mentioned problems, the invention of claim 1 is directed to an illuminating means for illuminating rice grains, an imaging means for taking an image of the rice grains illuminated by the illuminating means and obtaining an analog image signal, and an obtained analog image Image conversion means for binarizing the signal to convert it into a digital signal representing light and shade, and density distribution for performing image processing of the digital image signal obtained by the image conversion means and creating a density distribution histogram relating to the density of the image Histogram creation means, cumulative density distribution histogram creation means for accumulating the density distribution obtained by the density distribution histogram creation means to create a cumulative density distribution histogram, and cumulative density distribution obtained by the cumulative density distribution histogram creation means Binarization discriminating means for determining a threshold value T from the histogram to binarize rice grains and adhered wrinkles; The binarization discriminating means calculates the cumulative number of pixels in an appropriate ratio with respect to the total number of pixels in at least two locations on the cumulative density distribution histogram, and determines each cumulative pixel number. A density value calculating means for calculating a corresponding density value, a density value corresponding to the cumulative number of pixels calculated by the density value calculating means, and a linear expression that passes between the two plotted points is created. A function creating means, a provisional true value creating means for substituting the total number of pixels into the linear formula obtained by the linear function creating means, and setting the density value at that time as a provisional true value; and the provisional true value Deviation calculating means for comparing the provisional true value obtained by the preparation means with a reference value set in advance by measurement of a sample, for obtaining the deviation, and the deviation obtained by the deviation calculating means and the provisional true value And the threshold T The technical means of providing a threshold value determining means for determining was provided.

  As a result, if it is known that the pixel ratio of rice bran in rice grains does not exceed a predetermined value (for example, 25%) in advance, the cumulative pixel number is designated as 50% and 75% on the cumulative density distribution histogram. Each density value is calculated and a linear function is created. When the threshold value T is calculated with high accuracy from this linear function and this threshold value T is set on the cumulative density distribution histogram, it is possible to identify a small amount of cocoons attached to the rice grains. If it is known that the image ratio of minute wrinkles will not exceed 10% in advance, specify the cumulative number of pixels as 50% and 90%, calculate the respective density values, and create a linear function. The threshold value T may be calculated.

Further, in the subsequent process of the density distribution histogram creating means, the threshold value T is determined by the discriminant analysis method from the obtained density distribution histogram to binarize and discriminate the rice grain and the background, and only the background is removed. A background removing means may be provided.

As described above, according to the present invention, the illuminating unit that illuminates the rice grains, the imaging unit that captures the rice grains illuminated by the illuminating unit with the imaging camera, and obtains an analog image signal, and the obtained analog image signal Image conversion means for converting to a digital signal representing density by performing a digitization process, and density distribution histogram creation means for performing image processing of the digital image signal obtained by the image conversion means and creating a density distribution histogram relating to the density of the image And cumulative density distribution histogram creation means for accumulating the density distribution obtained by the density distribution histogram creation means and creating a cumulative density distribution histogram, and cumulative density distribution histogram obtained by the cumulative density distribution histogram creation means, Rice grain comprising: binarization discriminating means for binarizing and discriminating rice grain and adhered wrinkles by determining threshold T The binarization discriminating means obtains a cumulative number of pixels in an appropriate ratio with respect to the total number of pixels in at least two locations on the cumulative density distribution histogram and density values corresponding to the respective cumulative pixel numbers. A density value calculating unit that calculates the density value, a linear function generating unit that plots a density value corresponding to the cumulative number of pixels calculated by the density value calculating unit, and generates a linear equation that passes between the two plotted points; Substituting the total number of pixels into the linear equation obtained by the linear function creating means, and obtaining a temporary true value creating means for setting the density value at that time as a temporary true value, and the temporary true value creating means The provisional true value is compared with a reference value set in advance by measurement of a sample and the deviation is calculated, and the deviation obtained by the deviation calculation means is added to the provisional true value. Threshold to determine threshold T Therefore, it is possible to provide a rice grain appearance inspection apparatus that can set the threshold value T with high accuracy and identify a small amount of rice cake adhering to the rice grain.

Further, in the subsequent process of the density distribution histogram creating means, the threshold value T is determined by the discriminant analysis method from the obtained density distribution histogram to binarize and discriminate the rice grain and the background, and only the background is removed. Since the background removing means is provided, the background in the captured image can be identified and removed from the density distribution histogram.

The best mode for carrying out the present invention will be described below with reference to the drawings. FIG. 1 is a block diagram showing an example of a rice grain appearance inspection apparatus used in a rice grain appearance inspection apparatus.

Reference numeral 1 is a rice grain to be inspected, reference numeral 2 is a light source serving as an illuminating means for illuminating the rice grain obliquely from above, and reference numeral 3 is an imaging camera for imaging the rice grain illuminated by the illuminating means (for example, A solid-state camera (CCD camera), which can obtain analog image signals such as RGB. Reference numeral 4 denotes a camera control device for controlling the imaging camera 3. Reference numeral 5 denotes an image processing apparatus for converting an analog image signal obtained by imaging with the imaging camera 3 into a digital signal, storing it in an image memory, and processing the stored data. Reference numeral 6 denotes an image monitoring device for monitoring an image obtained by imaging with the imaging camera 3.

The image processing apparatus 5 has a computer system and has the following processing means. That is, the rice grain is imaged by the imaging camera 3, the obtained analog image signal is binarized and converted into a digital signal representing light and shade, and the image of the digital image signal obtained by the image conversion unit A density distribution histogram creating means for performing processing to create a density distribution histogram relating to the density of the image, and a cumulative density distribution histogram creating means for creating a cumulative density distribution histogram by accumulating the density distribution obtained by the density distribution histogram creating means And binarization discriminating means for binarizing discriminating rice grains and attached wrinkles by determining a threshold value T from the cumulative density distribution histogram obtained by the cumulative density distribution histogram creating means.

FIG. 2 is a flowchart showing the flow of processing in an embodiment of the rice grain appearance inspection apparatus of the present invention. The appearance inspection apparatus will be described with reference to this.

In step 1, an analog signal of an image obtained by imaging rice grains with a CCD imaging camera 3 under oblique illumination is binarized to convert it into a digital signal representing light and dark, and this digital signal is stored in an image memory. To do. The rice grain image obtained in step 1 includes a background image, an example of which is shown in FIG.

In step 2, the density of each pixel point in the region in the image obtained in step 1 (see FIG. 3) is analyzed. In the case of the present embodiment, the density is expressed in 256 levels from 0 (black) to 255 (white) by an 8-bit signal.

In step 3, a density distribution histogram is created by taking the density of each pixel point obtained in step 2 on the horizontal axis and the number of pixels on the vertical axis. An example of this density distribution histogram is shown in FIG.

また、クラス間分散σ_Ｂ ^２は式（２）のようになる。

そして分散比ηは式（３）のようになる。

このηが最大となるＴをしきい値Ｔとして決定すると、２値化判別することができる。図４の例では、しきい値Ｔは濃度値５０となり、濃度値５０未満の領域が背景画像として決定される。 Steps 4 and 5 show background removal means for binarizing and discriminating rice grains and the background and removing only the background. In step 4, in order to remove the background image from the density distribution histogram obtained in step 3, a threshold value T by the discriminant analysis method is determined. In this discriminant analysis method, there are two groups of a background image and a rice grain image, and the threshold value T is determined so that the variance ratio of intra-class variance and inter-class variance relating to these two groups is maximized. Now, the number of pixels n ₁ in the background image _{group 1} concentration values mean _mu, variance and sigma ₁ ^2, rice grain number of pixels n ₂ in the image _{group, 2} concentration values mean _mu, variance sigma ₂ ² and then, to represent the average density of the total number of pixels in the mu _T, within-class variance sigma _w ² is as equation (1).

Further, the interclass variance σ _B ² is as shown in the equation (2).

The dispersion ratio η is as shown in the equation (3).

If T at which η is maximum is determined as the threshold value T, binarization can be determined. In the example of FIG. 4, the threshold value T has a density value 50, and an area less than the density value 50 is determined as a background image.

  In step 5, the background image having a density value less than 50 is removed based on the threshold value T obtained in step 4, and a density distribution histogram of the rice grain image is created. An example of this is shown in FIG.

  Steps 6 to 13 show binarization determination means for determining the threshold value T from the cumulative density distribution histogram and binarizing the rice grains and the attached wrinkles. In step 6, the density distribution obtained in step 5 is accumulated, the density value is taken on the horizontal axis, and the cumulative number of pixels is taken on the vertical axis to create a cumulative density distribution histogram. An example of this is shown in FIG.

In step 7, the cumulative number of pixels of 50% of the total number of pixels and the cumulative number of pixels of 75% of the total number of pixels are obtained from the cumulative density distribution histogram obtained in step 6, and the density value and cumulative number of the cumulative number of pixels are 50%. A density value of 75% of pixels is calculated. A density value of 50% cumulative pixels is the central value of the cumulative pixels, a density value of 75% cumulative pixels is three-quarters of the cumulative pixels, and specifying these cumulative pixels has a small pixel ratio. It is effective to identify nasty traps. That is, it is determined from the precondition that the pixel ratio of minute wrinkles does not become larger than 25%. Therefore, if it is known that the image ratio of minute wrinkles does not exceed 10% in advance, a density value with a cumulative pixel number of 50% and a density value with a cumulative pixel number of 90% may be calculated.

In step 8, a straight line formula is created that passes between two points (cumulative pixel number 50%, its density value) and (cumulative pixel number 75%, its density value). This is represented by a linear equation as y = mx + b as shown in FIG.

In step 9, the total number of pixels (accumulated pixel number 100%) is substituted into y in the straight line expression obtained in step 8, and the density value x at that time is obtained, which is set as a temporary true value.

In step 10, a value set in advance by measurement of a sample or the like is set as a reference value, and a difference from the provisional true value obtained in step 9 is set as δ.

In step 11, a difference δ between the reference value set in step 10 and the provisional true value is calculated. That is, as shown in FIG. 7, the difference δ between the reference value in the space and the provisional true value, the slope m obtained in step 8, and the density value C75 with the cumulative number of pixels of 75% are mutually related. I understand. Since the plane equation can be expressed by the equation aX + bY + cZ + d = 0, it is expressed as δ = αm + βC75 + γ, and δ can be obtained by solving this plane equation. Note that the values of α, β, and γ may be those calculated by multiple regression analysis.

  In step 12, the threshold value T is determined based on the true value obtained by adding δ obtained in step 11 and the provisional true value.

  In step 13, based on the threshold value T obtained in step 12, the rice grains and the adhered wrinkles are binarized on the cumulative density distribution histogram. For example, when the density value 121.5 is calculated as the threshold value T, this is plotted on the x-axis of the cumulative density distribution histogram shown in FIG. 6 and when a line parallel to the y-axis is drawn, the cumulative pixel count A boundary is formed at a point of 98.2%. That is, it can be seen that 98.2% of the total number of pixels of rice grains is the number of pixels of rice grains, and 1.8% is the number of pixels of attached wrinkles.

10 samples of non-washed rice made by A company were extracted and imaged with low illuminance light, and 10 samples of non-washed rice made by A company were extracted and imaged with high illuminance light. 10 samples of polished rice were extracted and imaged at normal illuminance, the background image was removed, the density distribution was accumulated for the obtained density distribution histogram, the density value was plotted on the horizontal axis, and the number of accumulated pixels Is a vertical axis, and a cumulative density distribution histogram is created (see FIGS. 8, 9 and 10).

FIG. 11 shows the correlation between the difference δ between the reference value and the provisional true value and the slope m of the straight line passing between the two points in each cumulative density distribution histogram thus obtained. Further, in each cumulative density distribution histogram, the correlation between the slope m of the straight line passing between two points and the density value C75 of the cumulative number of pixels 75% is shown in FIG. As shown in FIGS. 11 and 12, it can be seen that none of the parameters are correlated in two dimensions. However, as shown in FIG. 7, in a three-dimensional space, δ, m, and C75 are correlated with each other. You can see that there is sex.

Then, the threshold value T is calculated by executing steps 10 to 11 described above. When this threshold value T is used as a calibration value (cal value) and the measured values (true values) of samples measured in advance are plotted, the relationships shown in FIGS. 13 to 15 are obtained.

FIG. 13 shows the relationship between the cal value and the true value of a non-washed rice image photographed with light of low illuminance, but the correlation coefficient is 0.907, and the cal value calculated in this example is extremely different from the true value. It can be seen that the correlation is strong. FIG. 14 shows the relationship between the cal value and the true value of the unwashed rice imaged with high-illuminance light, but the correlation coefficient is 0.995, and the cal value calculated in this example is extremely true. It can be seen that the correlation is strong. FIG. 15 shows the relationship between the cal value and true value of an image of polished white rice taken at normal illuminance. The correlation coefficient is 0.977, and the cal value calculated in this example is highly correlated with the true value. Is strong.

It is a block diagram which shows an example of the external appearance inspection apparatus of a rice grain. It is a flowchart which shows the flow of the process in one Embodiment of the external appearance inspection apparatus of a rice grain. It is an image of the rice grain containing the background image obtained by imaging with an imaging camera. It is a density distribution histogram of the image containing a background. It is a density distribution histogram of the rice grain image which does not include a background. It is a cumulative density distribution histogram composed of density values and cumulative number of pixels. It is a figure which shows the relationship between deviation (delta) in space, inclination m, and density | concentration value C75. 10 samples of non-washed rice made by company A are extracted, images are taken with light of low illuminance, background images are removed, the density distribution is accumulated with respect to the obtained density distribution histogram, and the obtained cumulative density distribution histogram is obtained. is there. 10 samples of non-washed rice made by company A are extracted, images are taken with light of high illuminance, background images are removed, the density distribution is accumulated with respect to the obtained density distribution histogram, and the obtained cumulative density distribution histogram is obtained. is there. This is a cumulative density distribution histogram obtained by extracting 10 samples of polished rice, photographing an image with light of ordinary illuminance, removing a background image, and accumulating the density distribution for the obtained density distribution histogram. It is a figure which shows correlation with deviation (delta) and the inclination m of the straight line which passes between 2 points | pieces. It is a figure which shows the correlation with the inclination m of the straight line which passes between 2 points | pieces, and the density value C75 of the accumulation pixel number 75%. It is a figure which shows the relationship between the calibration value (cal value) at the time of high illuminance image photography of the non-washing rice made from A company, and the measured value (true value) of the sample measured beforehand. It is a figure which shows the relationship between the calibration value (cal value) at the time of low illuminance image photography of the non-washing rice made from A company, and the measured value (true value) of the sample measured beforehand. It is a figure which shows the relationship between the calibration value (cal value) at the time of normal illuminance image photography of polished rice, and the measured value (true value) of the sample measured beforehand.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Rice grain 2 Light source 3 Imaging camera 4 Camera control apparatus 5 Image processing apparatus 6 Image monitor apparatus

Claims (2)

Lighting means for illuminating the rice grains;
Imaging means for capturing rice grains illuminated by the illumination means with an imaging camera to obtain an analog image signal;
Image conversion means for binarizing the obtained analog image signal and converting it into a digital signal representing light and shade,
A density distribution histogram creating means for performing image processing of the digital image signal obtained by the image converting means and creating a density distribution histogram relating to the density of the image;
Cumulative density distribution histogram creating means for accumulating the density distribution obtained by the density distribution histogram creating means and creating a cumulative density distribution histogram;
A rice grain appearance inspection comprising: binarization discriminating means for binarizing discriminating rice grains and attached wrinkles by determining a threshold value T from the cumulative density distribution histogram obtained by the cumulative density distribution histogram creating means A device,
The binarization determining means obtains a cumulative pixel number in an appropriate ratio with respect to the total number of pixels in at least two locations on the cumulative density distribution histogram, and calculates a density value corresponding to each cumulative pixel number; ,
Plot density values corresponding to the cumulative number of pixels calculated by the density value calculation means, and a linear function creation means for creating an equation of a straight line passing between the plotted two points;
Substituting the total number of pixels into the linear equation obtained by the linear function creating means, and a provisional true value creating means for setting the density value at that time as a provisional true value;
Deviation calculation means for comparing the provisional true value obtained by the provisional true value creation means with a reference value set in advance by measurement of a sample and for obtaining the deviation;
Threshold value determining means for determining the threshold value T by adding the deviation obtained by the deviation calculating means and the provisional true value;
An apparatus for inspecting the appearance of rice grains. In the subsequent step of the density distribution histogram creating means, the threshold value T is determined by a discriminant analysis method from the obtained density distribution histogram to binarize and discriminate the rice grain and background, and background removal for removing only the background. The rice grain appearance inspection apparatus according to claim 1, further comprising means.

Images