KR102576853B1 - Image processing method for counting pill - Google Patents

Abstract

A pill counting image processing method is introduced.
To this end, the present invention includes the steps of (a) setting the parameters of a camera positioned toward a plurality of pills to a set reference value; (b) the camera captures an image toward a tray on which a plurality of pills are placed, and the captured initial image is stored in a separate central server; (c) the central server cutting the surrounding area of the captured image in a set manner and then storing image 1; (d) the central server storing image 2 by applying brightness and contrast set to image 1; (e) the central server converting image 2 into a grayscale image and storing image 3; (f) In order to remove noise from Image 3, the central server converts the images of Image 3 with a pixel value of 0 into white pixels and the images of Image 3 with a pixel value of 1 to 254 into black pixels. Converting and saving image 4; (g) the central server storing image 5 by applying a distance transform algorithm to separate the pill image composed of white pixels in contact with each other from image 4; (h) the central server storing image 6 using a preset threshold according to the type of pill so that in image 5, the pill image is expressed in white and the background is black; (i) the central server obtaining information including the area, center coordinates, and bounding box size of the white area in image 6 using the OpenCV program; and (j) the central server compares the size of each white area obtained in step (i) using the set pill size (area) and displays it in green if it is within the set error range, and if it exceeds the set error range, It includes a step of displaying in a color other than green.

Description

Pill counting image processing method {IMAGE PROCESSING METHOD FOR COUNTING PILL}

The present invention relates to a pill counting image processing method.

In the refined drug business field, drug detection and counting technology is an essential technology for automating the preparation and distribution of drugs, and for this purpose, devices (hardware) and algorithms (software) that detect and count multiple pills at high speed have been steadily developed.

Regarding the high-speed counting device for pill packaging, in the case of conventional patents such as KR101462673B1 and US5463839, for high-speed counting of pills, several tracks (guides) are installed in the movement path of the pills after injection, so that multiple pills take different paths. It is designed to move through and collect in the final packaging machine (medicine container).

This is a method for moving a large number of pills at high speed in a short period of time, and simply corresponds to a method that increases the efficiency of pill counting by device. The counting speed performance of the device depends on the number of tracks in the movement path, and the even movement of pills within the track There is a disadvantage in that the volume of the device also increases as it includes a vibrating part for .

In addition, in relation to the distribution and counting of pills, in the case of conventional patents US6237804, US6377648, US6370215, US6497339, etc., a stopper is installed in the device or a disc with a hole is rotated and the pills discharged through the hole are counted, etc. It suggests mechanical solutions, and also suggests ways to insert the cassette containing the pills and vibrate them so that the pills are ejected without getting caught.

However, like the above cases, this is also a simple device solution and focuses on accuracy rather than counting speed, and the counting speed also depends on the driving speed of the device.

Additionally, these methods have the disadvantage of being specialized for each device and cannot be expanded to other existing devices.

Next, in relation to tablet drug counting and medication management, in the case of conventional patents US8271128 and US8855811, when the cassette containing the pills is mounted on the device by using both a tablet cassette counter and an optical counter, the internal structure of the cassette rotates and the tablets are pilled. are released one by one.

At this time, the pills are counted using the bottom counter of the cassette, and the discharged pills fall along the funnel of the device, pass through the optical counter at the bottom of the device, and are counted again.

This not only allows pills to be counted more accurately through double counting, but also detects in advance the information on the pills to be counted and the number of pills to be counted through data obtained from the barcode attached to the cassette. This is a device that can perform the management of the overall dispensing work in places such as pharmacies that require dispensing and automation of tablets according to prescriptions, such as verifying whether this is correct and comparing whether the actual number of pills counted matches the number of preset pills. It was proposed as an automated dispensing management system.

KR 10-2013-0042089 (2013.04.17)

The present invention seeks to accurately count pills by embedding a separate program capable of counting pills in a pill counter. In particular, it relates to a pill counting image processing method that can more accurately count overlapping pills.

A pill counting image processing method is introduced.

To this end, the present invention includes the steps of (a) setting the parameters of a camera positioned toward a plurality of pills to a set reference value; (b) the camera captures an image toward a tray on which a plurality of pills are placed, and the captured initial image is stored in a separate central server; (c) the central server cutting the surrounding area of the captured image in a set manner and then storing image 1; (d) the central server storing image 2 by applying brightness and contrast set to image 1;

(e) the central server converting image 2 into a grayscale image and storing image 3; (f) In order to remove noise from Image 3, the central server converts the images of Image 3 with a pixel value of 0 into white pixels and the images of Image 3 with a pixel value of 1 to 254 into black pixels. Converting and saving image 4; (g) the central server storing image 5 by applying a distance transform algorithm to separate the pill image composed of white pixels in contact with each other from image 4; (h) the central server storing image 6 using a preset threshold according to the type of pill so that in image 5, the pill image is expressed in white and the background is black; (i) the central server obtaining information including the area, center coordinates, and bounding box size of the white area in image 6 using the OpenCV program; and (j) the central server compares the size of each white area obtained in step (i) using the set pill size (area) and displays it in green if it is within the set error range, and if it exceeds the set error range, It includes a step of displaying in a color other than green.

According to the present invention comprised as described above, the effect of accurately distinguishing overlapping pills and counting them is realized.

Figures 1 to 9 are photographs imaged at each step to explain the pill counting image processing method of the present invention.

Hereinafter, a preferred embodiment of the pill counting image processing method of the present invention will be described with reference to the attached drawings.

The present invention is carried out in the following steps.

(a) Setting the parameters of a camera positioned toward a plurality of pills to a set reference value; (b) the camera captures an image toward a tray on which a plurality of pills are placed, and the captured initial image is stored in a separate central server; (c) the central server cutting the surrounding area of the captured image in a set manner and then storing image 1; (d) the central server storing image 2 by applying brightness and contrast set to image 1; (e) the central server converting image 2 into a grayscale image and storing image 3; (f) In order to remove noise from Image 3, the central server converts the images of Image 3 with a pixel value of 0 into white pixels and the images of Image 3 with a pixel value of 1 to 254 into black pixels. Converting and saving image 4; (g) the central server storing image 5 by applying a distance transform algorithm to separate the pill image composed of white pixels in contact with each other from image 4; (h) the central server storing image 6 using a preset threshold according to the type of pill so that in image 5, the pill image is expressed in white and the background is black; (i) the central server obtaining information including the area, center coordinates, and bounding box size of the white area in image 6 using the OpenCV program; and (j) the central server compares the size of each white area obtained in step (i) using the set pill size (area) and displays it in green if it is within the set error range, and if it exceeds the set error range, It includes a step of displaying in a color other than green.

The central server is not shown in the drawing, but may be a separate central server built into a general pill counter.

As shown in Figure 1, (a) setting the parameters of a camera positioned toward a plurality of pills to a set reference value and (b) the camera captures an image toward a tray on which a plurality of pills are placed, and the captured initial image is The steps are performed and stored on a separate central server.

Camera parameters are set (image horizontal and vertical size, brightness, contrast, gamma, saturation, hue, etc.).

The parameters used in the pill counter to which the present invention is applied are as follows.

1) Width of camera-acquired image (horizontal): 640mm, 2) Height of camera-acquired image (vertically): 480mm, 3) Camera brightness: 0, 4) Camera contrast: 100, 5) Camera gamma: 30, 6) Camera Color: -90, 7) Camera saturation: 0.

As shown in Figure 1, the area where the pill is placed is expressed in black, and the background image is expressed in gray.

Next, (c) the central server cuts the surrounding area of the captured image in a set manner and then stores Image 1.

This is a step to cut out unnecessary areas for accurate pill counting. To this end, coordinates are expressed for the width and height at the upper left and lower right corners of the initial image, respectively, so that all pill images expressed in black in the initial image are included. The surrounding area is cut into a square image divided by and the image 1 is stored.

As shown in FIG. 2, specifically, the width (horizontal) direction coordinate of the upper left corner of the image to be cut (x1), the width (horizontal) direction coordinate of the lower right corner of the image to be cut (x2), and the height (vertical) of the upper left corner of the image to be cut. ) If the direction coordinates (y1) and the height (vertical) direction coordinates (y2) of the lower right corner of the image to be cut are set in the central server, the square image is divided, and image 1 is stored in the central server according to the square image.

As shown, unnecessary areas are cut out, and Image 1 is created to include all black pill images and then stored on the central server.

Next, (d) the central server stores image 2 by applying the brightness and contrast set for image 1. Specifically, as shown in FIG. 3 ( In step d), the set brightness is 10 and the set contrast is 60.

Next, (e) the central server converts image 2 into a grayscale image and stores image 3, as shown in FIG. 4. The area of the image consists of a plurality of pixels, Each pixel is given a scale value between 0 and 254.

Next, (f) in order to remove the noise of Image 3, the central server uses white pixels for the pixel value of 0 among the images of Image 3, and white pixels for the images with a pixel value of 1 to 254 among the images of Image 3. The step of converting to black pixels and saving image 4 is performed.

That is, as shown in Figure 5, the pill image is displayed in black in Figure 4, so the threshold value is 0 (black) and the inversion type program logic (images expressed in black are converted to white, and images expressed in white are converted to white). When applying the logic that converts the image to black (the parameter used in the threshold function already stored in the central server), black pixels with a pixel value of 0 are white pixels, and gray and white pixels with values between 1 and 254 are black. Converted to color pixels.

Next, (g) the central server stores image 5 by applying a distance transform algorithm to separate the pill image composed of white pixels that are in contact with each other from image 4.

Specifically, the distance transform algorithm in step (g) uses the Euclidean distance calculation formula to calculate the closest distance value of the shortest distance from the current pixel to the pixel with a value of 0 (black). A matrix with a is obtained, and the distance value is mapped to a value between 0 and 254, so that it is expressed as a white area, and the overlapping pill image, which is the area where the images are in contact with each other, is expressed in black.

To explain in more detail, the distance transform uses the Euclidean distance formula to obtain a matrix with the closest distance value from the current pixel to the pixel with a value of 0 (black), and the distance values range from 0 to 254. It is mapped to the value between. (To express as a grayscale image)

The Euclidean distance calculation formula is explained in [Table 1] below.

According to the distance transform algorithm, the center of the white circle (pill image) is farther from the black pixel (background image) than the edge, so the edge pixels of the circle (pill image) have a value close to white. The distance to the black pixel (background image) is short, so the value is close to 0, so it is expressed as close to black.

Through this process, the area where the two pills come into contact is rendered close to black.

The image stored in the central server through the above process is shown in Figure 6, and it can be seen that not only the pill image but also the overlapping pill images are all expressed in close to black.

Next, (h) the central server stores image 6 using a preset threshold according to the type of pill so that in image 5, the pill image is expressed in white and the background is black.

The pill image is expressed in white and the background in black according to a preset threshold, and is converted by a pre-stored reduction ratio according to the type of pill to separate the overlapping pill images.

Specifically, the threshold applied at this stage uses a preset value depending on the type (size, shape) of the pill and is expressed as a reduction ratio. The smaller the value, the original size (Image 4) is expressed, but the attached The optimal reduction ratio was set in advance for pill separation. In the specific example, in the case of 2 mg of diazepam, when the reduction rate (threshold) is 50, a standard size of 117 (pixels) is applied.

In the above step, the pill image expressed in white is converted into a grayscale image by going through distance transform.

In addition, the central pixel of the pill image has a large value because it is the furthest from the background image, and the value becomes smaller as it moves to the edge of the pill. When a threshold is applied to this grayscale image (the area below the threshold is filled with black), (attributable to background) can result in the pill shrinking.

Ultimately, in the present invention, the relevant threshold is specified as a reduction rate, and for the convenience of users, the central server, which is a component of the present invention, performs tests to obtain the reduction rate and standard size for various pills on the market in advance to create a database. It has been saved.

The image stored in the central server through the above steps is shown in Figure 7.

As shown, the background image is black and the pill image is white, but it can be seen that the white area is smaller than the previous image. Furthermore, the area where the pills overlap is also clearly visible. You can see that each pill is separated and expressed as an image.

This step is actually the preparation stage for pill counting.

Next, (i) the central server acquires information including the area, center coordinates, and bounding box size of the white area in Image 6 using the OpenCV program.

Specifically, OpenCV's ‘connectedComponentsWithStats’ function is applied to Image 6 to obtain information about the white area (area (number of white pixels), center coordinates, bounding box size, etc.).

For actual pill counting, labeling is performed using an 8-step binarized image (Image6).

As shown in Table 2 below, when the upper left pixel of the initial image is (x, y) = (0, 0), moving from left to right in order from top to bottom, The correlation with the neighboring pixels (8 directions, pixels surrounding the corresponding pixel (left and right, top and bottom, diagonal)) is found and grouped (see Table 3). (Based on gray pixels in the binary image in Table 2 below)

Information such as number of groups, area, center coordinates, and bound size is obtained from the grouping results.

The number of groups refers to the number of pills, and the area expresses the size of the pills and is used for comparison with the standard value.

Next, (j) the central server compares the size of each white area obtained in step (i) using the set pill size (area), displays it in green if it is within the set error range, and matches the set error range. If it exceeds it, displaying it in a color other than green.

Specifically, if it is below the lower limit within the set error range, it is displayed in yellow and the user is warned. If it is above the upper limit within the error range, the number of pills is displayed as a quotient of the area divided by the reference area and then displayed in red to warn the user. It is characterized by a warning.

The specific embodiment will be described with reference to FIGS. 8 and 9 as follows.

FIG. 8 is a diagram showing a case where the white part of level 8 in Image 1 is painted green (within the standard size range), and FIG. 9 is a diagram illustrating a warning to the user.

The color of the pill warns the user. For example, if the standard size is 100, if the size is less than 5 (5%), it is treated as noise and not included in the count. If the size is between 5 and 50 (50% or less), it is counted, but it is expressed in yellow and a warning is issued (Figure 2). pill at the bottom right of 9).

Additionally, if the size is 51 to 150 (51% to 150%), it is considered a normal size and displayed in green, and if the size is 151 (over 150%), the quotient divided by the standard size is taken as the number of pills and displayed in red to warn. do. (When the size of the pill in the upper left corner of Figure 9 is 350, the quotient divided by the standard size is 3, so it is counted as 3)

Claims (6)

In the pill counting image processing method,
(a) Setting the parameters of a camera positioned toward a plurality of pills to a set reference value;
(b) taking an image by the camera toward the tray on which a plurality of pills are placed, and storing the captured initial image in a separate central server;
(c) the central server cutting the surrounding area of the captured image in a set manner and then storing image 1;
(d) storing image 2 by applying brightness and contrast set to image 1 by the central server;
(e) the central server converting image 2 into a grayscale image and storing image 3;
(f) In order to remove the noise of the image 3, the central server converts black pixels having a pixel value of 0 in the image 3 to white pixels, and pixels having a pixel value of 1 or more and 254 or less in the image 3. Converting to black pixels and saving image 4;
(g) storing image 5 by applying a distance transform algorithm to separate pill images composed of white pixels in contact with each other in the central server in image 4;
(h) The central server applies a preset threshold according to the type of pill, converts pixels with pixel values below the threshold to black, and converts pixels with pixel values exceeding the threshold to white. Saving image 6;
(i) obtaining, by the central server, information including the area, center coordinates, and bounding box size of the white area in the image 6 using the OpenCV program; and
(j) The central server compares the size of each white area obtained in step (i) using the set pill size (area), and displays it in green if it is within the set error range, and green if it exceeds the set error range. Including a step of displaying in a color other than that of
In Image 3 above, the pill is expressed in black,
In image 4 above, the pill is expressed in white,
Through the distance conversion algorithm, the image 4 is converted into a pixel value of 0 or more and 254 or less in correspondence with the distance value between each pixel and the closest black pixel, and is stored as the image 5,
In Image 5 saved by applying the distance conversion algorithm, the center of the pill is expressed closer to white than the edge of the pill, and the edge of the pill is expressed closer to black than the center of the pill, so in Image 5, the pill is It is expressed in grayscale, and the parts where the pills come into contact with each other are expressed closer to black.
The pill counting image processing method of claim 1 , wherein the size of the white pill in image 6 is smaller than that of the white pill in image 4. In claim 1,
Each of the parameters of the image taken by the camera in the step (a) is image width (width) of 640 mm, image height (length) of 480 mm, camera brightness of 0, camera contrast of 100, camera gamma of 30, camera It is characterized by a color of -90 and a camera saturation of 0.
The pill counting image processing method according to claim 1 , wherein the brightness set in step (d) is 10 and the contrast set is 60. In claim 2,
The method set in the step (c) is a rectangle delimited by coordinates representing width and height at the upper left corner and lower right corner of the initial image so that all of the pill images represented in black are included in the initial image. A pill counting image processing method characterized in that a peripheral region is cut out into an image and the image 1 is stored. delete delete In claim 1,
In step (j) above,
If it is less than the lower limit within the set error range, it is displayed in yellow and warns the user, and if it exceeds the upper limit within the error range, the number of pills is displayed as the quotient of the area divided by the reference area, and then displayed in red to warn the user. Characterized by a pill counting image processing method.