JP2007017163A - Evaluation method for granular products - Google Patents

Abstract

The present invention provides a method for evaluating the uniformity and the number of granular products in which granular products having a substantially constant size are arranged at approximately equal intervals as an overall tendency.
A gray-scale image obtained by imaging a granular product is binarized with a predetermined threshold value, a labeling process is performed on the extracted grain region, an area of each grain region is obtained, and a first area threshold value set in advance is obtained. Grain regions having less than the area are excluded from the evaluation target as noise, and the grain regions having an area that is greater than or equal to the first area threshold and less than the second area threshold are extracted as normal grain regions. Obtain the total area, extract the grain regions with an area equal to or larger than the second area threshold as abnormal grain regions, find the number and the sum of the areas of the abnormal grain regions, and calculate the number of normal grains, normal grain total area, abnormal grains When the number of normal grains or the total area of normal grains is less than the standard value, or the abnormal grain total area exceeds the standard value, the quality of the granular product is Judged to be non-uniform.
[Selection] Figure 1

Description

  The present invention relates to a method for evaluating the uniformity of the size of granular products in which granular products having a substantially constant size are arranged at approximately equal intervals.

  In the case of measuring individual sizes of a granular product having a substantially constant size, conventionally, the background and the granular product are distinguished by binarizing an image in which a large number of granular products are captured with a predetermined threshold value. Then, the obtained binary image is subjected to labeling processing, and the feature amount (area, side length of circumscribed rectangle, etc.) of each granular product is obtained by image analysis, and the tolerance value for each feature amount is set in advance. Often used is a method for determining the quality of a granular product based on whether or not it is in the container. Labeling is a general method for treating bright regions or dark regions to be connected in a binary image as one block and assigning labels (serial numbers) to individual connected regions.

  In the case of counting the number of granular products having a substantially constant size, conventionally, it is obtained by discriminating between the background and the granular product by binarizing an image in which a large number of granular products are imaged with an appropriate threshold value. In many cases, a labeling process is performed on the binary image and the number of granular products is counted based on whether or not the number of labels is within a preset allowable range.

  It also labels the binary image, counts the number of white and black pixels in each labeled area, determines whether the pattern is a foreign object based on the number of white pixels, and determines white and / or black pixels. There is known a method of determining whether or not a grain is present by dividing by the reference pixel number and improving the precision of determining the number of granular products (see Patent Document 1).

However, when producing a granular product with a large granulator or the like, it is necessary to manage the quality based on the overall tendency of the grain size and the approximate number of grains. In such a case, rather than accurately measuring the size and number of individual grains, it is necessary to grasp the approximate number of normal grains, the approximate number of abnormally sized grains, and the approximate total number of grains. I must.
JP-A-5-165959

  An object of the present invention is to provide a method for evaluating the uniformity of the size of a granular product in which granular products having a substantially constant size are arranged at approximately equal intervals as an overall tendency.

  In order to solve such a problem, the present inventor has intensively studied a method for evaluating the uniformity of the size of the granular product. As a result, the grayscale image obtained by imaging the granular product is binarized and binarized. The extracted grain area is labeled, the area of each labeled grain area is calculated, the area threshold is set to remove the noise in the grain area, the normal grain area is extracted and the number of normal grains is counted, the area of the normal grain area The total number of abnormal grains is extracted and the number of abnormal grains is counted. The total sum of the areas of abnormal grains is obtained, and the number of normal grains, the total area of normal grains, the number of abnormal grains, or the total area of abnormal grains is preset. When it was less than the reference value, it was found that the size of the granular product was determined to be non-uniform so that the uniformity of the size of the granular product could be evaluated as an overall trend, leading to the present invention.

  That is, the present invention is a method for evaluating the uniformity of the size of a granular product in which granular products having a substantially constant size are arranged at approximately equal intervals. (1) A grayscale image is obtained by imaging the granular product. An imaging process to be generated; (2) a binarization process for binarizing a grayscale image with a predetermined threshold; (3) a labeling process for performing a labeling process on a grain region extracted by the binarization process; ) An area calculating step for obtaining an area of each labeled grain region; (5) a noise removing step for excluding a grain region having an area less than a preset first area threshold from noise as an evaluation target; ) Extracting a grain region having an area that is greater than or equal to the first area threshold and less than the second area threshold as a normal grain region, and counting the number of normal grains, and (7) the total area of the normal grain regions Normal grain total area calculation step to obtain (8 Extracting a grain region having an area equal to or larger than the second area threshold as an abnormal grain region and counting the number thereof, and (9) calculating an abnormal grain total area calculating a sum of the areas of the abnormal grain regions. And (10) when compared with the reference values set in advance for the number of normal grains, the total area of normal grains, the number of abnormal grains and the total area of abnormal grains, and if the number of normal grains is less than the standard value, A uniformity evaluation step for determining that the size of the granular product is non-uniform when the number of abnormal grains exceeds the reference value or the abnormal grain total area exceeds the reference value, It is the evaluation method of the uniformity of the granular product characterized by providing.

  According to the method of the present invention, the number of normal grains having an area value in a predetermined range and the total area thereof, and the number of abnormal grains having an area value exceeding the predetermined value and the total area thereof as feature quantities are used as normal grains. When the pre-set reference value is not satisfied, and when the pre-set reference value is exceeded for abnormal particles, it is determined that the size of the granular product is non-uniform. The uniformity can be determined based on the above.

  Further, the present invention determines that the size of the granular product is non-uniform when the total area of normal grains having an area value in a predetermined range is less than a preset reference value. This can also be detected when the number of grains decreases abnormally.

  Furthermore, the present invention determines that the size of the granular product is non-uniform when the total area of abnormal grains having an area value exceeding a predetermined value exceeds a preset reference value. Even when one grain is generated or when many abnormal grains slightly exceeding a predetermined value are generated, it can be determined as non-uniform.

  FIG. 1 shows an embodiment of the method for evaluating the uniformity of a granular product according to the present invention. On the upper surface of the granulating belt 1 of the belt type granulator, the granular product 2 is arranged at approximately equal intervals to form grains, and is conveyed. The granular product 2 is imaged by a camera 3 installed above the granulation belt 1, and image data is transmitted to the image analysis device 4.

FIG. 2 shows an image analysis procedure performed in the image analysis apparatus 4. The procedure starts from step s0, and in step s1 (imaging), for example, an image for one frame is input using an image capture board or the like. The input image is normally stored in the memory of the image analysis device 4 as grayscale image data of 8 bits (256 gradations) per pixel.
In step s2 (binarization), binarization processing is performed on the input grayscale image using a predetermined threshold value. For example, a pixel having a density value less than the threshold is dark (density value = 0), and a pixel having a density value greater than or equal to the threshold is bright (density value = 1). In this embodiment, the dark area is a grain area. As the threshold used here, a fixed value may be input in advance, or may be dynamically determined using various threshold determination algorithms.

In step s3 (labeling), a labeling process is performed on the grain region extracted by the binarization process. Labeling is to treat light areas or dark areas to be connected in a binary image as one block and to give labels (serial numbers) to individual connection areas. In the present embodiment, since the dark region is a grain region, labeling is performed on the dark region.
In step s4 (area calculation), the area is obtained from the number of pixels in each labeled grain region.
In step s5 (noise removal), a region in which the grain region area does not satisfy the preset first area threshold is determined as noise and is excluded from subsequent processing targets.

In step s6 (normal grain count), areas where the grain area is not less than the first area threshold and less than the second area threshold are extracted as normal grain areas, and the number is counted.
In step s7 (calculate normal grain total area), the total area of the normal grain region extracted in step s6 is calculated.
In step s8 (abnormal grain count), labels whose area is equal to or larger than the second area threshold are extracted as abnormal grain areas, and the number is counted.
In step s9 (abnormal grain total area calculation), the total area of the abnormal grain region extracted in step s8 is calculated.

In step s10 (uniformity evaluation), each of the number of normal grains, normal grain total area, abnormal grain number, and abnormal grain total area is compared with a preset reference value. If the normal grain number is less than the reference value, normal When the total grain area is less than the reference value, when the number of abnormal grains exceeds the reference value, or when the total abnormal grain area exceeds the reference value, it is determined that the size of the granular product is non-uniform.
After executing the above processing continuously, the process shifts to the end state of step s11, maintains the standby state, and steps when the next measurement time is reached based on a preset measurement interval (unit: millisecond). Move to s0.

  In step s2 of FIG. 2, when the original image includes shading unevenness due to illumination unevenness or the like, it may be effective to perform appropriate shading correction processing prior to binarization processing. Shading correction is a technique that removes shading unevenness by creating an image with only shading of the background by, for example, averaging the number of images, and subtracting it from the original image input at each measurement. This is a typical image processing technique.

  In step s3 of FIG. 2, there may be a problem in image processing such as a void is generated inside the grain region, or because the grains are close to each other, they are pseudo-connected on the image. is there. In such a case, the cavity can be filled or the pseudo-connected grain regions can be separated by applying an image expansion / contraction process. The expansion / contraction is, for example, a process of forcibly brightening 8 pixels around a bright pixel (expansion of a bright pixel), and is a general image processing technique.

  FIG. 3 shows an example of an original image captured by the camera 3 and input to the image analysis device 4 and an image obtained by binarizing the original image. In the example of FIG. 3, for each pixel, an average value of a local region of 15 pixels × 15 pixels with the pixel as the center is obtained, and binarization is performed using the value as a threshold for the pixel. In this method, a local region of 15 pixels × 15 pixels cannot be obtained in the region of 7 pixels around the image. For this reason, the area of 7 pixels around the image is excluded from the evaluation target as an invalid area after the binarization process. In the binarized image of FIG. 3, the invalid area is displayed in black. The invalid area is not displayed in subsequent figures. The binarization method shown here is an example, and an appropriate binarization method can be selected according to the background and contrast of the target image.

  Next, a labeling process is performed on the grain region extracted by the binarization process. In this embodiment, a black area is a labeling target, and serial numbers are assigned to all grain areas. Furthermore, the area of each labeled grain region is determined.

  Table 1 shows the first area threshold value, second area threshold value, normal grain number reference value, normal grain total area reference value, abnormal grain number reference value, and abnormal grain total area used after the noise removal step. An example of a reference value is shown. The reference value is for an evaluation range of 150 pixels × 150 pixels.

In the present embodiment, in the noise removal step, a grain region having an area less than the first area threshold is excluded from the evaluation target as noise.
In the normal grain number counting step, a grain area having an area that is equal to or larger than the first area threshold and less than the second area threshold is extracted as a normal grain area, and the number is counted. In the normal grain total area calculating step, the total area of the normal grain regions is obtained.
In the abnormal particle number counting step, a particle region having an area equal to or larger than the second area threshold is extracted as an abnormal particle region, and the number is counted. In the abnormal grain total area calculating step, the total area of the abnormal grain regions is obtained.

  In the uniformity evaluation process, the number of normal grains, the total area of normal grains, the number of abnormal grains, and the total area of abnormal grains are compared with the above reference values. If the number of normal grains is less than the standard value, the total area of normal grains is When the reference value is not satisfied, the size of the granular product is determined to be non-uniform when the number of abnormal particles exceeds the reference value or the total area of abnormal particles exceeds the reference value. The determination result may be displayed on the monitor screen of the image analysis device 4, or a digital signal or the like may be output by an external device as necessary.

4 to 8 show a binarized image, the number of normal grains, the total area of normal grains, the number of abnormal grains, and the total area of abnormal grains in other examples.
The example in FIG. 4 is an example in which all items are determined to be normal, and the examples in FIGS. 5 to 8 are examples in which any item is not satisfied and thus determined to be non-uniform. In FIG. 5, the number of normal grains is 65, which is less than the normal grain number reference value 70 shown in Table 1. In FIG. 6, the number of normal grains is 72 and satisfies the reference value of 70. However, since the area of each individual grain is small as a whole, the value of the total area of normal grains is 3493 pixels. The area reference value of 5000 pixels is not satisfied. In FIG. 7, some abnormal grains in which several grains are connected are seen, the number of abnormal grains is 7, which exceeds the reference value of the number of abnormal grains. FIG. 8 shows one region where a large number of grains are connected. For this reason, the number of abnormal grains is one and does not exceed the reference value of 5, but the abnormal grain total area is 8717 pixels, which exceeds the reference value of the abnormal grain total area of 3500 pixels. At the same time, the number of normal grains is correspondingly reduced.

It is a figure which shows one Embodiment of the uniformity evaluation method of the granular product of this invention. It is a figure which shows the analysis procedure in the image analysis apparatus. It is a figure which shows the example of the original image and the image which binarized it. It is a figure which shows the example of the image which shows the grain state determined to be a normal state, and its determination result. It is a figure which shows the image determined as a non-uniform | heterogenous state because the number of normal grains does not satisfy the reference value, and the determination result. It is a figure which shows the image and its determination result which are determined to be in a non-uniform state because the normal grain total area does not satisfy the reference value. It is a figure which shows the image determined as a non-uniform state because the number of abnormal grains exceeds a reference value, and the determination result. It is a figure which shows the image determined as a non-uniform | heterogenous state because the abnormal grain total area exceeds a reference value, and its determination result.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Granulation belt 2 Granular product 3 Camera 4 Image analyzer

Claims (1)

A method for evaluating the uniformity of the size of granular products in which granular products having a substantially constant size are arranged at approximately equal intervals,
(1) An imaging process for imaging a granular product to generate a grayscale image;
(2) a binarization step for binarizing the grayscale image with a predetermined threshold;
(3) a labeling process for performing a labeling process on the grain region extracted by the binarization process;
(4) an area calculation step for obtaining the area of each labeled grain region;
(5) a noise removing step of excluding, as noise, a grain region having an area less than a preset first area threshold value from the evaluation target;
(6) Normal grain number counting step of extracting a grain region having an area which is equal to or larger than the first area threshold and less than the second area threshold as a normal grain region,
(7) a normal grain total area calculating step for obtaining a total area of normal grain regions;
(8) An abnormal particle number counting step of extracting a particle region having an area equal to or larger than the second area threshold as an abnormal particle region and counting the number thereof,
(9) An abnormal grain total area calculating step for obtaining a total area of abnormal grain regions;
(10) Compared to the preset reference values for the number of normal grains, normal grain total area, abnormal grain count, and abnormal grain total area, and the normal grain total area is the reference value when the normal grain count is less than the standard value A uniformity evaluation step for determining that the size of the granular product is non-uniform when the number of abnormal grains exceeds the reference value or the total area of abnormal grains exceeds the reference value. A method for evaluating the uniformity of granular products.

Images