A method for detecting the quantity of circular medicines based on the power theorem.

Abstract

This invention discloses a method for detecting the number of circular medicines based on the power theorem. The method involves acquiring an image of a medicine package, performing image preprocessing to obtain the edge points of the circular medicines, selecting any point P in the medicine package image, drawing a ray from P to the edge of the medicine, recording the set of intersection points of the ray and the edge points, calculating the product of line segments between P and each intersection point, and, according to the power theorem, calculating the set of edge points with the same product to construct a power histogram. Peak detection is performed on the power histogram, with a peak value representing the cumulative number of points on the same circle. Cross-validation is performed on each detected peak point, and the circle parameters are calculated and output, representing the detected circular regions. The number of circles is then counted to determine the number of medicines. This method has low data processing volume, small memory footprint, high detection accuracy, and fast speed, effectively reducing errors in the number of medicines during the packaging process.

Description

Technical Field

[0001] This invention relates to secondary verification of drug packaging, and more particularly to a method for detecting the quantity of circular drugs based on the power theorem. Background Technology

[0002] With the increasing automation of drug management, fully automated drug packaging systems are increasingly replacing manual packaging and dispensing. Even after drugs are packaged by the packaging machine according to the doctor's prescription, problems such as excess tablets, missing tablets, and broken tablets may still occur. If these issues are not detected and addressed promptly, they can lead to accidental ingestion by patients, worsening their condition or even posing a life-threatening risk. Therefore, secondary verification of the packaged drugs to ensure that the types and quantities of drugs in the package match the electronic prescription is a necessary measure to ensure drug packaging safety. Summary of the Invention

[0003] To address the shortcomings of existing technologies, the present invention aims to provide a method for detecting the quantity of circular medicines based on the power theorem.

[0004] To achieve the objectives of this invention, the technical solution adopted is as follows:

[0005] A method for detecting the quantity of circular medicines based on the power theorem includes the following steps:

[0006] (1) Acquire images of pharmaceutical packaging, perform image preprocessing, including image enhancement, image filtering, image segmentation, and edge extraction, and obtain the edge points of circular pharmaceutical packaging.

[0007] (2) Select any point P in the image of the medicine package and draw a ray from it to the edge of the medicine. Record the set of intersection points of the ray and the edge points. Calculate the product of the line segments of point P and each intersection point. According to the power theorem, calculate the set of edge points with the same product and construct a power histogram.

[0008] (3) Perform peak detection on the power histogram, where a certain peak value represents the cumulative value of points on the same circle;

[0009] (4) Perform cross-validation on each detected peak point, calculate and output the circle parameter, which is the detected circular area, and count the number of circles, which is the number of drugs.

[0010] Furthermore, step (1) specifically includes:

[0011] (1.1) Perform image grayscale conversion to achieve image enhancement;

[0012] (1.2) Median filtering is used to filter the grayscale image to eliminate isolated noise points;

[0013] (1.3) Perform image segmentation to divide the filtered image into foreground pills and background;

[0014] (1.4) Edge extraction: Obtain the edge points of the circular drug.

[0015] Further, in step (1.2), before performing median filtering, salt and pepper noise is added to the grayscale image to fill in the holes in the image.

[0016] Further, in step (1.3), the Otsu thresholding method is used to segment the image. First, the optimal threshold is obtained, and then the image is binarized to segment the image into foreground pills and background.

[0017] Further, in step (1.4), the Sobel operator is used for edge detection. This operator is convolved with the original image to obtain the horizontal and vertical gradient values. When the gradient value is greater than a certain threshold, the point is considered to be an edge point.

[0018] Furthermore, step (2) specifically includes:

[0019] (2.1) Define E as the set of edge pixels of the image, P as a point outside the edge pixels in the image, and R as M rays emanating from point P into the circle, R = {r j |j=1,2,......,M}, for one of the rays r j ∈R, define a subset E of E j ={e k |k=1,2,......,K}, the points in this subset are ray r j The point of intersection with the edge of the circle;

[0020] (2.2) Calculate E j Find the distances between all points and point P, calculate the product of the distances between all pairs of points, and store the product in set p. j :

[0021]

[0022] (2.3) Define the power histogram of the image as PH, count the set of intersections of the M rays emanating from point P with the edge points, calculate the product set, and obtain:

[0023]

[0024] Where, r k Let n be the product order value, n be the x-axis of the power histogram, and n be the product order value. k The product is r k The number of edge pixels, p(r) k ) represents the product order value r k The probability of occurrence is represented by the ordinate of the power histogram, for each product order value r. kEach corresponds to a series of edge pixels, forming a pixel set.

[0025] Furthermore, two steps are involved before constructing the power histogram:

[0026] The first step is to perform preprocessing to obtain a set of refined edge pixels and increase the number of reference points;

[0027] The second step is to set the density of rays emitted from each reference point.

[0028] Furthermore, in step (3), a zero-crossing peak detection operator is used to detect the peak value of the histogram, and a peak detection operator k is defined. σ The peak detection result is obtained by convolving the detection operator with the power histogram. The peak detection operator k σ It is the difference operator Γ and the Gaussian filter operator g σ The product of.

[0029] Further, in step (4), cross-validation refers to emitting rays from N different points in the image to the detection circle, obtaining N power histograms, and storing the peak values ​​in all power histograms into a set S = {S i |i = 1, 2, ..., N};

[0030] Each S i It contains a series of peaks. For one of the peaks p, ψ(p) is defined as the set of edge points corresponding to the peak point p, including real edge points and false edge points. χ(p) is defined as a subset of ψ(p), and the points in χ(p) that satisfy the cross-validation condition are real edge points on the same circle corresponding to the peak p.

[0031] Furthermore, χ(p) needs to satisfy the following cross-validation conditions:

[0032] (1)χ(p) is a common subset of the point sets corresponding to multiple peaks representing the same circle in N power histograms;

[0033] χ(p)=ψ(p)∩ψ(p1)∩ψ(p2)......ψ(p n p1, p2, ..., p n These are different peaks, and they come from other power histograms that are different from the P peak;

[0034] (2) χ(p) must have a certain scale, and the number of elements in the point set must be greater than a certain threshold, |χ(p)|≥λ n ;

[0035] (3) The points on the detection circle are required to have a certain density. Based on the pixels in χ(p), the least squares method is used to fit the circle C to obtain the circle parameters (x). c ,yc Let N be the number of points on circle C. c The density distribution of points on the circle must be greater than a certain set threshold, i.e., |N c | / r≥R c ;

[0036] N c ={e i ||dist(e i ,p c -r|≤δ,e i ∈E}, center p c =(x c ,y c ), where δ is the defined threshold.

[0037] The beneficial effects of this invention are that, compared with the prior art, the algorithm involved in this invention converts two-dimensional image data into one-dimensional power histogram processing, which occupies less memory space, improves real-time data processing volume, has high detection accuracy, and is fast. It effectively reduces the error in the quantity of medicines during the packaging process and plays an important role in realizing intelligent, efficient, and pollution-free oral medication packaging and distribution in wards.

[0038] The method of this invention uses multiple power histogram cross-validation to eliminate false peak points and false pixel points, thereby obtaining accurate edge points on the circle and improving the accuracy of circle detection. Attached Figure Description

[0039] Figure 1 This is a flowchart of the circular drug quantity detection method based on the power theorem described in this invention;

[0040] Figure 2 This is a flowchart of the image preprocessing process;

[0041] Figure 3 It is a grayscale image of a medical packaging patch;

[0042] Figure 4 This is a schematic diagram of a median filter.

[0043] Figure 5 This is an image filtering flowchart;

[0044] Figure 6 The images show the results of adding salt and pepper noise and performing median filtering. Image a is a schematic diagram after adding salt and pepper noise, and image b is a schematic diagram after performing median filtering.

[0045] Figure 7 This is a schematic diagram of the Ostu algorithm;

[0046] Figure 8 This is a schematic diagram of the image segmentation results from the Ostu algorithm;

[0047] Figure 9 This is a schematic diagram of the edge detection results using the Sobel operator;

[0048] Figure 10 This is a schematic diagram of the power theorem;

[0049] Figure 11 Figure a shows a peak detection diagram. The Gaussian operator smooths the histogram, and the difference operator transforms a peak in the power histogram into a positive or negative peak that is symmetrical about the horizontal axis, as shown in Figure b. According to the zero-crossing intersection principle, the zero point where the positive and negative changes cross in Figure b is a peak point, as shown in Figure c. Detailed Implementation

[0050] The technical solution of the present invention will be further described below with reference to the accompanying drawings and embodiments. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and should not be used to limit the scope of protection of this application.

[0051] like Figure 1 As shown, the method for detecting the quantity of circular medicines based on the power theorem of the present invention specifically includes the following steps:

[0052] (1) Acquire images of pharmaceutical packaging, perform image preprocessing, including image enhancement, image filtering, image segmentation, and edge extraction, and obtain the edge points of circular pharmaceutical packaging.

[0053] like Figure 2 As shown, images of medical packaging are acquired and preprocessed, including image grayscale conversion, median filtering, image segmentation, and edge extraction. The extracted edge points are then stored in point set E.

[0054] (1.1) Perform image grayscale conversion to achieve image enhancement;

[0055] Image grayscale conversion transforms a color image into a grayscale image. Grayscale images separate the target tablet from the background while preserving the tablet's edge information to the greatest extent possible. Therefore, when enhancing an image with grayscale, only a preliminary grayscale conversion is needed to preserve the tablet's edge information, facilitating subsequent filtering and denoising processes.

[0056] grayscale image of medical packaging Figure 3 As shown.

[0057] (1.2) Median filtering is used to filter the grayscale image to improve image quality;

[0058] Median filtering works by replacing the pixel's value with the median gray value within the pixel's neighborhood. Its working principle can be described using... Figure 4 It is represented by five parts.

[0059] Wander the template across the image until its center coincides with a certain pixel position; read the grayscale values ​​of the corresponding pixels under the template; arrange these grayscale values ​​in ascending order; find the middle value among these values; assign this middle value to the pixel at the center position of the template.

[0060] Median filtering can modify pixels whose gray values ​​differ too much from those of surrounding pixels, making them closer to the values ​​of surrounding pixels, thus eliminating isolated noise points.

[0061] like Figure 5 As shown, before performing median filtering, salt-and-pepper noise needs to be added to the grayscale image. Adding noise can fill in the holes in the image, making the filtering effect of median filtering more obvious.

[0062] Figure 6 This is the result of adding salt and pepper noise and performing median filtering. Comparing the median filtering result with the grayscale image, we can see that the noise in the background has been smoothed more finely. In this process, the edges of the pills are not damaged, and the background grayscale is unified more deeply, which will make the subsequent image segmentation process more accurate.

[0063] (1.3) Perform image segmentation to divide the filtered image into foreground pills and background;

[0064] like Figure 7 As shown, the Otsu thresholding method is used to first obtain the optimal threshold, and then the image is binarized to segment the filtered tablet image into foreground tablets and background.

[0065] This method uses a threshold t to segment the image into foreground μ0 and background μ1, with probabilities ω0 and ω1 respectively. The total mean of the image is μ. According to the following formula (1), the minimum gray value of the image is traversed to the maximum value t. The t when g is the maximum is the optimal threshold for segmentation.

[0066]

[0067] In terms of variance, g can measure the uniformity of gray-level distribution. The larger the variance value, the greater the difference between the background and the target that make up the image.

[0068] The Ostu algorithm segmentation results are as follows: Figure 8 As shown.

[0069] (1.4) Edge extraction to obtain the edge points of the circular drug;

[0070] Based on the step property of an image at its edges, the gradient of an image exhibits a maximum value in the edge region. Operators that utilize this property for edge detection are called first-order differential edge detection operators, also known as gradient edge operators. The Sobel operator is capable of performing both horizontal and vertical edge detection; its template includes a horizontal template G. x and a longitudinal gradient template G y Formula (2) is the template details.

[0071]

[0072] By performing a convolution operation between this operator and the original image, we can obtain the gradient values ​​of the horizontal direction G(x) and the vertical direction G(y). When the gradient value is greater than a certain threshold, the point can be considered an edge point.

[0073] like Figure 9 The image shows the edge detection effect of the Sobel operator. Edge detection preserves the edge information of the tablet more effectively. Therefore, edge detection is currently the most suitable algorithm for detecting tablet edges, and the extracted edge points are stored in the point set E.

[0074] (2) Select any point P in the image of the medicine package and draw a ray from it to the edge of the medicine. Record the set of intersection points of the ray and the edge points. Calculate the product of the line segments of point P and each intersection point. According to the power theorem, calculate the set of edge points with the same product and construct a power histogram.

[0075] like Figure 10 As shown, suppose a circle has center O and radius r. Define the power of any point P with respect to the center O as: Simultaneously satisfy Where A, B, M, and N are the intersection points of the rays emanating from point P into the circle and the circle itself. According to the power theorem, the power calculated for edge points on the same circle, i.e., pow(O,P), should be approximately equal. Edge points on different circles will have different pow(O,P) values. Therefore, this clustering principle can be used to determine which points belong to the edge points of the same circle.

[0076] For an image I, let E be the set of edge pixels, P be a point outside the edge pixels, and R be M rays emanating from point P into the circle, where R = {r j |j=1,2,......,M}, for one of the rays r j ∈R, define a subset E of E j ={e k |k=1,2,......,K}, the points in this subset are ray r j The point of intersection with the edge of the circle.

[0077] Calculate E jFind the distances between all points and point P, calculate the product of the distances between all pairs of points, and store the product in set p. j :

[0078]

[0079] Define the power histogram of the image as PH, count the set of intersections of the M rays emanating from point P with edge points, and calculate the product set p. j ,get:

[0080]

[0081] Where, r k The product order value, which is the x-axis of the power histogram, n k The product is r k The number of edge pixels, p(r) k ) represents the product order value r k The probability of occurrence is represented by the ordinate of the power histogram, for each product order value r. k Each of these corresponds to a series of edge pixels, forming a pixel set. According to the power theorem, the vertices in the pixel set of the power histogram represent the pixels on several circles.

[0082] Before constructing the power histogram based on the above principles, two specific steps are required:

[0083] The first step is to perform preprocessing to obtain a set of refined edge pixels e, and increase the number of reference points;

[0084] First, for incomplete circles, it's impossible to collect two intersection points from certain reference points; therefore, increasing the number of reference points can overcome this positional bias problem. Second, for small circles far from the reference points, it may be impossible to collect enough intersection points, leading to missed detections. Third, since the power of a point associated with a circle is equal to the square of the tangent segment from that point to the circle, different circles may share the same circle power from different reference points. For real images with noise and quantization errors, using multiple reference points with multiple histograms and cross-validation checks will reduce the randomness of the system and increase the confidence of the detection peaks.

[0085] The second step is to consider the density of rays emitted from each reference point;

[0086] Sparse rays are likely to cause missed intersections during the construction of the power histogram, while dense rays may lead to duplicate counting and increased computational costs. Therefore, under optimal conditions, the number of rays required will vary depending on the size and complexity of the image.

[0087] For example, given an image of size n×m pixels, we obtain an edge set E. The following strategy is used to determine the number of rays to be taken: if the size of the edge set is smaller than the size of the boundary pixels (i.e., |E| < 2mn-4), then the rays from point P are directed to each edge pixel in E; otherwise, the rays are emitted from point P to the boundary pixels (i,j), i∈{1,m},j∈{1,m}.

[0088] (3) Peak detection is performed on the power histogram. A certain peak value represents the cumulative points on the same circle, which can effectively detect circular areas.

[0089] When working with digital images, edge pixels rarely lie on a circle. Therefore, circle detection aims to detect a group of edge pixels that may be roughly located on either side of an edge. Similarly, points on the same circle do not precisely correspond to the same power value. Consequently, peaks found in the power histogram tend to spread out and may exhibit mixed clusters, posing problems for the peak detection stage.

[0090] Therefore, this invention proposes using a zero-crossing peak detection operator to detect histogram peaks. A peak detection operator k is defined. σ Peak detection results That is, using the detection operator and the power histogram to perform convolution, with peak detection operator k. σ It is the difference operator Γ and the Gaussian filter operator g σ The product of.

[0091]

[0092]

[0093] like Figure 11 As shown, peak detection includes the following three processes: First, the histogram is smoothed by the Gaussian operator (Figure a); second, a peak in the power histogram is transformed into a positive and negative peak that is symmetrical about the horizontal axis using the difference operator (Figure b); and third, according to the zero-crossing intersection principle, the zero point where the positive and negative changes cross in Figure b is a peak point (Figure c).

[0094] (4) Perform cross-validation on each detected peak point, calculate and output the circle parameter, which is the detected circular area, and count the number of circles, which is the number of drugs.

[0095] In step (3), a series of peaks were detected, each corresponding to a series of edge points. However, there are false points among the edge points, meaning that some points do not belong to the same circle. Furthermore, some of the detected peaks are false. Therefore, the cross-validation method is used to exclude edge points and obtain accurate edge points on the circle, thereby improving the accuracy of the circle parameter calculation.

[0096] Cross-validation involves emitting rays from N distinct points in an image towards a detection circle, resulting in N power histograms. The peak values ​​from all power histograms are then stored in a set S = {S...} i |i = 1, 2, ..., N}. Each S i The array contains a series of peaks. For one of the peaks p, ψ(p) is defined as the set of edge points corresponding to peak point p, including real edge points and false edge points. χ(p) is defined as a subset of ψ(p), where each point in χ(p) is a real edge point on the same circle corresponding to peak p.

[0097] χ(p) needs to satisfy the following cross-validation conditions:

[0098] (1)χ(p) is a common subset of the point sets corresponding to multiple peaks representing the same circle in N power histograms.

[0099] χ(p)=ψ(p)∩ψ(p1)∩ψ(p2)......ψ(p n p1, p2, ..., p n These are different peaks, and they come from other power histograms that are different from the P peak.

[0100] (2) χ(p) must have a certain scale, and the number of elements in the point set must be greater than a certain threshold, |χ(p)|≥λ n .

[0101] (3) The points on the detection circle are required to have a certain density. Based on the pixels in χ(p), the least squares method is used to fit the circle C to obtain the circle parameters (x). c ,y c Let N be the number of points on circle C. c The density distribution of points on the circle must be greater than a certain set threshold, i.e., |N c | / r≥R c .

[0102] N c ={e i ||dist(e i ,p c -r|≤δ,e i ∈E}, center p c =(x c ,y c ), where δ is the defined threshold.

[0103] Based on the above conditions, a peak point can be cross-validated using multiple power histograms. This eliminates false peak points and false pixels, improving the accuracy of circle detection.

[0104] The beneficial effects of this invention are that, compared with the prior art, the algorithm involved in this invention converts two-dimensional image data into one-dimensional power histogram processing, which occupies less memory space, improves real-time data processing volume, has high detection accuracy, and is fast. It effectively reduces the error in the quantity of medicines during the packaging process and plays an important role in realizing intelligent, efficient, and pollution-free oral medication packaging and distribution in wards.

[0105] The method of this invention uses multiple power histogram cross-validation to eliminate false peak points and false pixel points, thereby obtaining accurate edge points on the circle and improving the accuracy of circle detection.

[0106] The applicant of this invention has provided a detailed description of the embodiments of the invention in conjunction with the accompanying drawings. However, those skilled in the art should understand that the above embodiments are merely preferred embodiments of the invention. The detailed description is only intended to help readers better understand the spirit of the invention and is not intended to limit the scope of protection of the invention. On the contrary, any improvements or modifications made based on the inventive spirit of the invention should fall within the scope of protection of the invention.

Claims

1. A method for detecting the number of circular medicine based on the theorem of power of a circle, characterized in that, The method comprises the steps of: (1) collecting a medical package image, performing image preprocessing, including image enhancement, image filtering, image segmentation, and edge extraction to obtain circular medicine edge points; Specifically comprising: (1.1) performing image gray scale conversion to achieve the purpose of image enhancement; (1.2) filtering the gray scale image using median filtering to eliminate isolated noise points; Median filtering is to roam the template in the image, the center of the template coincides with a certain pixel position; read the gray scale values of each corresponding pixel under the template; arrange these gray scale values from small to large into a column; find out the one in the middle of these values; assign this middle value to the pixel corresponding to the center position of the template; (1.3) image segmentation, segmenting the filtered image into foreground tablets and background; (1.4) edge extraction to obtain circular medicine edge points; (2) preprocessing to obtain a set of thinned edge pixels; increase the number of reference points and set the density of the rays emitted from each reference point; Select any point P in the medical package image to make a ray to the medicine edge, record the intersection set of the ray and the edge point, calculate the line segment product of P point and each intersection point, calculate the number of edge points with the same product according to the power theorem of circle, and construct a power histogram; Specifically comprising: (2.1) define E as the set of edge pixels of the image, P as a point in the image other than an edge point, and R as M rays emanating from P toward the circle, For one of the rays define a subset of E The points in this subset are the intersection points of the ray with the edge of the circle; (2.2) Compute the distance of all points in P and compute the product of the distances of all pairs of points and store the product in the set : (2.3) define the power histogram of the image as PH, count the intersection set of M rays from P points and edge points, calculate the product set, and get: wherein, is a product magnitude value, is an abscissa of the power histogram, is a product magnitude value is a number of edge pixel points with a product is a product magnitude value is a probability of occurrence, is an ordinate of the power histogram, for each product magnitude value corresponds to a series of edge pixel points, forming a pixel point set; (3) peak detection on the power histogram, a certain peak value represents the cumulative points on the same circle; A zero-crossing peak detection operator is used to detect the histogram peaks, and a peak detection operator is defined The peak detection result is the convolution of the detection operator and the power histogram, and the peak detection operator is the product of a difference operator and a Gaussian filter operator ; (4) cross-validation of each detected peak point, calculating and outputting the circle parameters, i.e. the detected circular region, and counting the number of circles, i.e. the number of medicines; Cross-validation means that N rays are sent from N different points in the image to the detected circle, and N power histograms are obtained. The peak values in all the power histograms are stored in a set ; Each of contains a series of peaks, for one of the peaks , define as the edge point set corresponding to the peak point p, which includes the real edge point and the false edge point; define as subset of , which satisfies the cross-validation condition The point in is a real edge point on the same circle corresponding to the peak p.

2. The circular medicine quantity detection method based on the power theorem of circle according to claim 1, wherein Step (1.2), before median filtering, add salt and pepper noise to the gray scale image of the image to supplement the hollow points in the image.

3. The circular medicine quantity detection method based on the power theorem of circle according to claim 1, wherein Step (1.3), using Otsu threshold segmentation method to perform image segmentation, first obtaining the optimal threshold, then performing binary processing on the image, and dividing the image into foreground tablets and background.

4. The circular medicine quantity detection method based on the power theorem of circle according to claim 1, wherein Step (1.4), using Sobel operator for edge detection, performing convolution operation on the operator and the original picture to obtain horizontal and vertical gradient values, and considering that the point is an edge point when the gradient value is greater than a certain threshold.

5. The circular medicine quantity detection method based on the power theorem of circle according to claim 1, wherein Step (4), The following cross-validation conditions need to be met: (1) is a common subset of the point sets corresponding to the plurality of peaks representing the same circle in the power-of-N histogram; , are different peaks and come from other power histograms than the P peak; (2) To have a certain size, the number of elements in the point set is greater than a certain threshold, ; (3) the points on the circle are required to have a certain density, according to least square method to fit the circle C, and obtain the circle parameters , define the number of points on the circle C as the density distribution of the points on the circle needs to be greater than a certain set threshold, that is ; , center of the circle , is a defined threshold value.

Images