CN117274293A - Accurate bacterial colony dividing method based on image features - Google Patents

Abstract

The invention relates to the technical field of image analysis, in particular to an accurate bacterial colony dividing method based on image features. The method comprises the following steps: acquiring a gray image of the surface of a culture medium, and determining a colony area; carrying out connected domain analysis to determine a connected domain to be detected, dividing an initial sliding window, and selecting a target sliding window from the initial sliding window; determining a connected domain analysis index of the connected domain to be detected; screening out a target connected domain from the connected domain to be detected; taking all other target sliding windows except the outermost target sliding window in the target communication domain as sliding windows to be segmented; according to the distance between the sliding window to be segmented and other sliding windows to be segmented, selecting an edge sliding window from the sliding windows to be segmented; dividing the target communicating region according to the central point in the edge sliding window to obtain bacterial colonies, and taking each communicating region to be tested except the target communicating region as one bacterial colony. The invention can effectively improve the accuracy of bacterial colony division and ensure the accuracy and the reliability of bacterial colony division.

Description

Accurate classification method of bacterial colonies based on image features

Technical field

The invention relates to the technical field of image analysis, and in particular to a method for accurately dividing bacterial colonies based on image features.

Background technique

Bacterial colonies are bacterial aggregates formed on culture media. By observing and counting the distribution and abundance of different colonies, we can initially understand the types and relative quantities of bacteria present in the sample. However, different colonies may have edge-adhesive contact due to limitations in their growth area. Therefore, different bacterial colonies need to be divided into colonies.

In related technologies, color block analysis is used to achieve colony division. In this way, due to the influence of bacterial hyphae at the edges, the edge features of mutually adherent colonies are blurred, and the same type of bacteria in the same culture medium The color difference of bacterial colonies is small. Therefore, when dividing colonies only based on color patch analysis, connected domains with adhesion cannot be effectively divided. The accuracy and reliability of bacterial colony classification are low. Difference.

Contents of the invention

In order to solve the technical problems in related technologies that connected domains with adhesion cannot be effectively divided, the accuracy of bacterial colony classification is low, and the accuracy and reliability of bacterial colony classification are poor, the present invention provides a bacterial colony classification method based on image features. The technical solution used to accurately divide colonies is as follows:

The present invention proposes a method for accurately dividing bacterial colonies based on image features. The method includes:

Obtain a grayscale image of the surface of the culture medium, and determine the colony area of the grayscale image of the surface of the culture medium based on threshold segmentation;

Perform connected domain analysis on the colony area to determine the connected domain to be tested, segment the connected domain based on a preset size and non-overlapping sliding window, and obtain an initial sliding window. According to the initial sliding window and the connected domain to be tested The gray value of the pixel in the middle, selects the target sliding window from the initial sliding window;

According to the area of each connected domain to be measured, the areas of all target sliding windows in the connected domain to be measured, and the position distribution of the target sliding windows, determine the connected domain analysis index of the connected domain to be measured; according to the connected domain analysis The indicator selects the target connected domain from the connected domain to be measured;

All other target sliding windows in the target connected domain except the outermost target sliding window are used as the sliding windows to be divided; according to the distance between the sliding window to be divided and the nearest preset number of other sliding windows to be divided. The distance is used to select edge sliding windows from the sliding windows to be segmented;

The target connected domain is segmented according to the center point in the edge sliding window to obtain bacterial colonies, and each connected domain to be tested except the target connected domain is regarded as a bacterial colony.

Further, selecting a target sliding window from the initial sliding window based on the initial sliding window and the gray value of the pixel in the connected domain to be measured includes:

The normalized value of the gray value information entropy of all pixels in the initial sliding window is used as the sliding window confusion degree;

The mean value of the gray value of all pixels in the initial sliding window is used as the sliding window mean, and the standard deviation of the gray value of all pixels in the initial sliding window is used as the sliding window standard deviation; calculate the sliding window mean and the The product of the sliding window standard deviation is used as the sliding window characteristic parameter;

The mean value of the gray value of all pixels in the connected domain to be measured is used as the mean value to be measured, and the standard deviation of the gray value of all pixels in the connected domain to be measured is used as the standard deviation to be measured; calculate the mean value to be measured and the mean value to be measured The product of the standard deviation to be measured is used as the characteristic parameter to be measured;

Determine the sliding window mutation coefficient according to the sliding window characteristic parameter and the characteristic parameter to be measured;

Calculate the product of the sliding window chaos degree and the sliding window mutation coefficient as the sliding window screening coefficient;

A target sliding window is selected from the initial sliding windows according to the sliding window screening coefficient of each initial sliding window.

Further, determining the sliding window mutation coefficient according to the sliding window characteristic parameter and the characteristic parameter to be measured includes:

Calculate the absolute value of the difference between the sliding window characteristic parameter and the characteristic parameter to be measured as the characteristic difference;

The sliding window mutation coefficient is determined according to the characteristic difference and the characteristic parameter to be measured, wherein the characteristic difference is positively correlated with the sliding window mutation coefficient, and the characteristic parameter to be measured is negatively correlated with the sliding window mutation coefficient, so The value of the sliding window mutation coefficient is a normalized value.

Further, selecting a target sliding window from the initial sliding window according to the sliding window screening coefficient of each initial sliding window includes:

The initial sliding window with the sliding window screening coefficient greater than the preset screening coefficient threshold is used as the target sliding window.

Further, determining the connected domain analysis index of the connected domain to be measured based on the area of each connected domain to be measured, the areas of all target sliding windows in the connected domain to be measured, and the position distribution of the target sliding windows, includes: :

Calculate the ratio of the area of the connected domain to be measured to the areas of all target sliding windows in the connected domain to be measured as the area influence coefficient;

Use any target sliding window in the connected domain to be tested as the sliding window to be analyzed; determine the center point of the target sliding window that is closest to the center point of the sliding window to be analyzed, and calculate the distance value between the two center points The normalized value of is used as the distance index of the sliding window to be analyzed; the mean value of the distance index of all target sliding windows is used as the distance influence coefficient;

The normalized value of the product of the area influence coefficient and the distance influence coefficient is used as the connected domain analysis index.

Further, selecting the target connected domain from the connected domain to be tested according to the connected domain analysis index includes:

The connected domain to be tested whose connected domain analysis index is smaller than the preset indicator threshold is used as the target connected domain.

Further, selecting edge sliding windows from the sliding windows to be divided based on the distance between the sliding window to be divided and the nearest preset number of other sliding windows to be divided includes:

The sum of the distances between any sliding window to be divided and the nearest preset number of other sliding windows to be divided is used as the sum of the dividing distances;

Perform normalization processing on the segmentation distance and value to obtain an edge selection index;

Edge sliding windows are filtered out from the sliding windows to be divided according to the edge selection indicators of all the sliding windows to be divided.

Further, selecting edge sliding windows from the sliding windows to be divided based on the edge selection indicators of all the sliding windows to be divided includes:

The sliding window to be segmented whose edge selection index is smaller than the preset edge threshold is used as an edge sliding window.

Further, the step of segmenting the target connected domain according to the edge lines in the edge sliding window to obtain bacterial colonies includes:

Connect the center points of the two closest edge sliding windows, divide the target connected domain into at least two sub-regions, and treat each sub-region as a bacterial colony.

Further, the preset number is 10.

The invention has the following beneficial effects:

This invention determines the colony area by acquiring the grayscale image of the surface of the culture medium, and then performs connected domain analysis and segments the connected domains based on sliding windows of preset size and non-overlapping to obtain an initial sliding window. Through the initial sliding window and the test The gray value of the pixels in the connected domain is selected from the initial sliding window. It can be understood that the bacterial colonies have discrete colonies on the outside, and there are also multiple bacterial colonies that adhere to each other. When multiple bacterial colonies adhere to each other, due to the The distribution of filaments makes the middle edge area blurred and the corresponding edge information is less. The effect obtained by directly using color block analysis and processing is poor. Therefore, the embodiment of the present invention performs specific analysis by setting non-overlapping sliding windows, making the analysis process The degree of refinement is higher and the effect is better; since a single connected domain to be tested can be a colony or a combination of multiple colonies, the present invention combines the area information of the connected domain to be tested itself and the information of the target sliding window in the connected domain to be tested. The quantity and position distribution determine the connected domain analysis index, and the target connected domain is screened out from the connected domain to be tested, so that the target connected domain containing multiple colonies can be accurately screened, and then the distance between the sliding windows to be segmented in the target connected domain can be accurately selected. Determine the edge sliding window, divide the target connected domain based on the edge sliding window, so that the mutually adherent colonies can be effectively divided, and combine the originally individually distributed connected domains to be tested to divide the bacterial colonies. In summary, the present invention filters out the existing bacterial colonies Connected domains with adhesion can be effectively divided into connected domains with adhesion, which can more accurately classify bacterial colonies based on grayscale differences and edge features, improve the accuracy of bacterial colony classification, and ensure the accuracy and accuracy of bacterial colony classification. reliability.

Description of the drawings

In order to more clearly explain the technical solutions and advantages in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings that need to be used in the description of the embodiments or the prior art. Obviously, the drawings in the following description The drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without exerting creative efforts.

Figure 1 is a flow chart of a method for accurately dividing bacterial colonies based on image features provided by an embodiment of the present invention;

Figure 2 is a schematic diagram of the distribution of sliding windows to be divided according to an embodiment of the present invention.

Detailed ways

In order to further elaborate on the technical means and effects adopted by the present invention to achieve the predetermined inventive purpose, the following is a specific implementation of a method for accurately dividing bacterial colonies based on image features proposed by the present invention in conjunction with the drawings and preferred embodiments. The method, structure, characteristics and functions are described in detail below. In the following description, different terms "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Additionally, the specific features, structures, or characteristics of one or more embodiments may be combined in any suitable combination.

Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the technical field to which the invention belongs.

The specific scheme of the method for accurately classifying bacterial colonies based on image features provided by the present invention will be described in detail below with reference to the accompanying drawings.

Please refer to Figure 1, which shows a flow chart of a method for accurately classifying bacterial colonies based on image features provided by one embodiment of the present invention. The method includes:

S101: Obtain a grayscale image of the surface of the culture medium, and determine the colony area of the grayscale image of the surface of the culture medium based on threshold segmentation.

A usage scenario of the embodiment of the present invention can be, for example, collecting a grayscale image of the surface of the culture medium, and performing colony identification on the grayscale image of the surface of the culture medium based on image analysis technology. It is understandable that there may be mutual interactions between different colonies. In the case of adhesion and similar appearance, there may be only a small area between colonies with edge features. When performing edge detection, there is a high probability that edge features that are too small will be regarded as normal areas, thus making the effect of color block analysis Poor, the accuracy of identifying and classifying bacterial colonies is poor, making it impossible for existing methods to accurately classify bacterial colonies. Based on this, the present invention can refine the overall culture medium surface grayscale image to the corresponding grayscale image by setting an initial sliding window. In the sliding window, colonies that adhere to each other and have similar appearance can be effectively divided.

In the embodiment of the present invention, a high-resolution camera can be placed directly above the culture dish for photographing and collecting. The culture dish is placed on a horizontal table with a large color difference, and an external light source or appropriate lighting equipment is used to obtain high-quality images. Then, the collected images are preprocessed, and the resulting RGB color format image is grayscaled. At the same time, bilateral filtering is used to remove image noise, and the image corresponding to the culture medium is extracted, thereby obtaining a grayscale image of the culture medium surface. Among them, the process of image preprocessing is a well-known technology in the art, and will not be described in detail.

After obtaining the grayscale image of the culture medium surface, the colony area of the grayscale image of the culture medium surface can be determined based on threshold segmentation; it is understandable that because the petri dish is placed on a horizontal tabletop with a large color difference, that is, the tabletop, culture Bases and colonies can form obvious color stratification, and the distinction is more obvious in grayscale images. Therefore, threshold segmentation can be used to distinguish the blank culture medium area and the area where the colonies are located, that is, the colony area.

In the embodiment of the present invention, after determining the colony area, specific image analysis can be performed on the colony area. Please refer to subsequent embodiments for the image analysis process.

S102: Perform connected domain analysis on the colony area to determine the connected domain to be tested, segment the connected domain based on a preset size and non-overlapping sliding window, and obtain an initial sliding window. According to the initial sliding window and the pixels in the connected domain to be tested, Grayscale value, select the target sliding window from the initial sliding window.

The preset size is the size of the initial sliding window. It can be understood that the smaller the preset size, the higher the precision of corresponding colony image recognition. Therefore, in the embodiment of the present invention, the preset size is set to 5× 5 size, of course, can also be adjusted according to actual detection needs, there is no restriction on this.

In the embodiment of the present invention, connected domain analysis is performed on the colony area to determine the connected domain to be measured. The connected domain analysis is a method of combining adjacent pixels with the same or similar gray value as a connected domain. The connected domain analysis can To effectively divide colony areas, the connected domain analysis technology is a technology well known to those skilled in the art. Through connected domain analysis processing, colony areas with relatively discrete distribution can be identified and distinguished.

It can be understood that due to the irregular distribution of bacterial colonies in the petri dish, and the colony shape, size, and position of different bacterial colonies will be different, therefore, there may be cases of bacterial colonies adhering to each other. In the embodiment of the present invention, Set a sliding window with a preset size and non-overlapping to segment the connected domain to obtain an initial sliding window, and conduct a more refined analysis of the connected domain based on the initial sliding window.

Further, in some embodiments of the present invention, selecting the target sliding window from the initial sliding window according to the gray value of the pixels in the connected domain to be measured includes: converting the values of all pixels in the initial sliding window. The normalized value of gray value information entropy is used as the sliding window confusion degree; the mean value of the gray value of all pixels in the initial sliding window is used as the sliding window mean, and the standard deviation of the gray value of all pixels in the initial sliding window is used as the sliding window. Standard deviation; calculate the product of the sliding window mean and the sliding window standard deviation as the sliding window characteristic parameter; use the mean of the gray value of all pixels in the connected domain to be measured as the mean to be measured, and the standard of the gray value of all pixels in the connected domain to be measured The difference is used as the standard deviation to be measured; the product of the mean value to be measured and the standard deviation to be measured is calculated as the characteristic parameter to be measured; the sliding window mutation coefficient is determined based on the sliding window characteristic parameter and the characteristic parameter to be tested; the sliding window chaos degree and the sliding window mutation coefficient are calculated The product of is used as the sliding window screening coefficient; the target sliding window is selected from the initial sliding window according to the sliding window screening coefficient of each initial sliding window.

In the embodiment of the present invention, the normalized value of the information entropy of the gray value of all pixels in the initial sliding window can be used as the degree of confusion of the sliding window. The greater the information entropy, the corresponding gray distribution of the pixels in the initial sliding window. The more complex it is, the more likely it is a pixel containing edge information.

In the embodiment of the present invention, the mean value of the gray value of all pixels in the initial sliding window is used as the sliding window mean, and the standard deviation of the gray value of all pixels in the initial sliding window is used as the sliding window standard deviation; calculate the sliding window mean and sliding window The product of standard deviations is used as the sliding window characteristic parameter. The calculation formula corresponding to the sliding window characteristic parameter can be specifically, for example:

In the formula, Represents the sliding window characteristic parameters of the i-th initial sliding window, /> Represents the mean value of the gray value of all pixels in the i-th initial sliding window, that is, the sliding window mean value,/> Represents the standard deviation of the gray value of all pixels in the i-th initial sliding window.

In the embodiment of the present invention, the grayscale mean and standard deviation are used as the grayscale features of the pixels in the sliding window, so that the sliding window feature parameters are obtained by combining the grayscale mean and standard deviation, so as to facilitate subsequent feature analysis based on the sliding window feature parameters.

After determining the sliding window characteristic parameters, the product of the mean value to be measured and the standard deviation to be measured can be used as the characteristic parameter to be measured. The specific implementation process is similar to the sliding window characteristic parameters, and will not be described further.

Further, in some embodiments of the present invention, determining the sliding window mutation coefficient according to the sliding window characteristic parameter and the characteristic parameter to be measured includes: calculating the absolute value of the difference between the sliding window characteristic parameter and the characteristic parameter to be measured as the characteristic difference; according to The feature difference and the feature parameter to be measured determine the sliding window mutation coefficient. Among them, the feature difference is positively correlated with the sliding window mutation coefficient, and the characteristic parameter to be measured is negatively correlated with the sliding window mutation coefficient. The value of the sliding window mutation coefficient is normalized. value.

Among them, the positive correlation relationship means that the dependent variable will increase as the independent variable increases, and the dependent variable will decrease as the independent variable decreases. The specific relationship can be a multiplication relationship, an additive relationship, or the power of an exponential function. etc., determined by practical applications; a negative correlation relationship means that the dependent variable will decrease as the independent variable increases, and the dependent variable will increase as the independent variable decreases, which can be a subtraction relationship, a division relationship, etc. , determined by actual application.

It can be understood that generally bacterial colonies show light color characteristics, such as milky white, light yellow, etc., and there are usually grayscale mutations at the edges. The overall color of the connected domain to be tested is the color of the colony itself, while the edges are The corresponding color of the mutated area is closer to the background color. Since there is an obvious color difference between the background and the colony, the degree of sliding window mutation can be analyzed based on this color difference. The calculation formula corresponding to the sliding window mutation coefficient can be specifically, for example:

In the formula, Represents the sliding window mutation coefficient of the i-th initial sliding window, /> Indicates the connected domain to be tested in which the i-th initial sliding window is located/> The characteristic parameters to be measured;/> Represents the sliding window characteristic parameters of the i-th initial sliding window;/> Represents a constant coefficient, a safe value set to prevent the denominator from being 0, optionally/> ;/> Indicates taking the absolute value;/> Represents normalization processing. In one embodiment of the present invention, the normalization process can be specifically, for example, the maximum and minimum value normalization process, and the normalization in subsequent steps can all adopt the maximum and minimum value normalization process. In other aspects of the present invention, In the embodiment, other normalization methods can be selected according to the specific range of numerical values, which will not be described again.

in, Indicates the i-th initial sliding window and its connected domain to be tested/> feature difference, because the feature parameters to be measured and the sliding window feature parameters are both the product of the corresponding sliding window mean and the sliding window standard deviation, that is, the greater the feature difference, it can represent the difference between the initial sliding window and the pixels in the connected domain to be measured to which it belongs. The greater the difference in the mean value of the gray value and the larger the standard deviation of the gray value, that is, there is a larger gray value difference between the initial sliding window and the connected domain to be tested. Therefore, the pixels of the initial sliding window are mutation pixels. The greater the possibility of a point, the ratio of the feature difference to the feature parameter to be measured is calculated and normalized to eliminate the grayscale influence of different connected domains to be measured and improve the accuracy and reliability of the sliding window mutation coefficient.

In the embodiment of the present invention, after determining the sliding window mutation coefficient, the product of the sliding window chaos degree and the sliding window mutation coefficient can be calculated as the sliding window screening coefficient; where the sliding window screening coefficient is the possibility that the initial sliding window contains different colony edges. sexual indicator data. Because the degree of chaos of the sliding window represents the complexity of the gray value distribution of pixels in the sliding window, the greater the degree of chaos of the sliding window, the more complex the gray value distribution of the pixels in the corresponding initial sliding window, and the initial sliding window that normally contains colonies , its corresponding gray value changes less and tends to be stable. Therefore, the greater the degree of chaos of the sliding window, the more information contained in the corresponding initial sliding window contains the edge of the colony; the sliding window mutation coefficient represents the mutation of the pixels in the initial sliding window. In this case, the larger the mutation coefficient of the sliding window, it can also mean that the gray value of the corresponding pixel in the initial sliding window changes greatly, and there is a large difference from the connected domain to be tested, which in turn can mean that it contains background-related information. That is, the initial sliding window is more likely to be the sliding window corresponding to the edge of the colony. Therefore, the present invention calculates the product of the sliding window chaos degree and the sliding window mutation coefficient to obtain the sliding window screening coefficient.

Further, in some embodiments of the present invention, selecting the target sliding window from the initial sliding window according to the sliding window screening coefficient of each initial sliding window includes: selecting the initial sliding window whose sliding window screening coefficient is greater than the preset screening coefficient threshold. as target sliding window.

Among them, the target sliding window is a sliding window that includes the edge of the colony, that is, the position of the target sliding window is very likely to be the edge of the colony.

The preset filtering coefficient threshold is the threshold value of the sliding window filtering coefficient. Optionally, in the embodiment of the present invention, the preset filtering coefficient threshold can be, for example, 0.8, that is, the initial sliding window filtering coefficient is greater than 0.8. The sliding window serves as the target sliding window. Of course, the preset screening coefficient threshold can also be adjusted according to actual detection needs, and there is no restriction on this.

After the target sliding window is determined, the target sliding window can be further analyzed based on the distribution of the target sliding window. Please refer to subsequent embodiments for the specific process.

S103: Determine the connected domain analysis index of the connected domain to be measured based on the area of each connected domain to be measured, the areas of all target sliding windows in the connected domain to be measured, and the position distribution of the target sliding windows; The target connected domain is filtered out from the connected domain.

Further, in some embodiments of the present invention, the connected domain analysis of the connected domain to be measured is determined based on the area of each connected domain to be measured, the areas of all target sliding windows in the connected domain to be measured, and the position distribution of the target sliding windows. Indicators include: calculating the ratio of the area of the connected domain to be measured to the area of all target sliding windows in the connected domain to be measured as the area influence coefficient; taking any target sliding window in the connected domain to be measured as the sliding window to be analyzed; determining Analyze the center point of the sliding window that is closest to the center point of the target sliding window, and use the normalized value of the distance value between the two center points as the distance index of the sliding window to be analyzed; use the mean value of the distance index of all target sliding windows As the distance influence coefficient; the normalized value of the product of the area influence coefficient and the distance influence coefficient is used as the connected domain analysis index.

Among them, the connected domain analysis index is the analysis data of whether there are multiple adhesion edges of colonies in the connected domain to be tested. The calculation formula corresponding to the connected domain analysis index can be as follows:

In the formula, Represents the connected domain analysis index of the Lth connected domain to be tested,/> Represents the total area of the Lth connected domain to be tested,/> Represents the area of N target sliding windows in the Lth connected domain to be tested, N represents the total number of target sliding windows in the connected domain to be tested; n represents the index of the target sliding window, /> Represents the distance index of the nth target sliding window, /> Represents normalization processing.

In the formula, Represents the distance influence coefficient of the target sliding window in the connected domain to be tested. The distance influence coefficient represents the distribution of all target sliding windows. The smaller the value, the more concentrated the distribution between the target sliding windows. It is the distance influence coefficient of the connected domain to be tested in the edge area of the colony. The greater the possibility, the greater the corresponding connected domain analysis index. Because the edge area in the colony connected domain is larger and darker than the uneven growth area, the target sliding window corresponding to the colony edge area is more than the target sliding window corresponding to the uneven growth area. Therefore/> The smaller the value of , the more target sliding windows exist in the connected domain to be tested, indicating that the connected domain to be tested is more likely to have cohesive edges, that is, the larger the value of the connected domain analysis index is.

Therefore, in some embodiments of the present invention, filtering out the target connected domain from the connected domains to be tested based on the connected domain analysis index includes: using the connected domain to be tested whose connected domain analysis index is less than the preset indicator threshold as the target connected domain .

The preset indicator threshold is the threshold value of the connectivity domain analysis indicator. The preset indicator threshold in the embodiment of the present invention can be, for example, 0.5. That is to say, the connected domain to be tested will be tested if the connectivity domain analysis indicator is less than 0.5. as the target connected domain. Of course, the value of the preset indicator threshold can also be adjusted according to actual detection needs, and there is no restriction on this.

It can be understood that in the actual detection scenario of bacterial colonies, since there is less edge information between mutually adhered bacterial colonies, when performing connected domain analysis, the area composed of multiple bacterial colonies will be collectively regarded as one connected domain. In this case, the present invention will This situation is analyzed, and the connected domain to be tested with multiple different colonies is screened out through the connected domain analysis index as the target connected domain, while other connected domains to be tested only contain one colony. From this, the target connected domain can be further analyzed. analysis to ensure the reliability of colony classification.

S104: Use all target sliding windows in the target connected domain except the outermost target sliding window as the sliding windows to be divided; according to the distance between the sliding window to be divided and the nearest preset number of other sliding windows to be divided, Filter edge sliding windows from the sliding windows to be divided.

In the embodiment of the present invention, all target sliding windows except the outermost target sliding window in the target connected domain can be used as sliding windows to be segmented. It can be understood that the present invention can determine the center point of the target connected domain, and then , radiating outward from the center point of the target connected domain, thereby determining the outermost target sliding window, or alternatively, determining the outermost edge of the connected domain to be measured, and taking the target sliding window that overlaps the outermost edge as the outermost The target sliding window can be used as the sliding window to be segmented for specific analysis after removing the target sliding window from all target sliding windows.

Further, in some embodiments of the present invention, filtering out edge sliding windows from the sliding windows to be divided based on the distance between the sliding window to be divided and the nearest preset number of other sliding windows to be divided includes: The sum of the distances between a sliding window to be segmented and the nearest preset number of other sliding windows to be segmented is used as the segmentation distance and value; the segmentation distance and value are normalized to obtain the edge selection index; based on all the segments to be segmented The edge selection index of the sliding window selects edge sliding windows from the sliding windows to be divided.

The preset number is the selected number of other sliding windows to be divided. In the embodiment of the present invention, the preset number can be, for example, 10, or it can be adjusted according to actual detection requirements, and there is no limit to this.

In the embodiment of the present invention, the sliding windows to be divided are specifically analyzed based on the distance between the sliding windows to be divided, that is, the distance between any sliding window to be divided and the nearest preset number of other sliding windows to be divided is calculated. The sum value is used as the sum of segmentation distances, and is normalized to obtain the edge selection index. When the edge selection index is larger, it can mean that the corresponding sliding window to be segmented is generally far away from other sliding windows to be segmented, that is, the If the sliding window to be divided is in an isolated position, the corresponding edge selection index will be larger.

For example, as shown in Figure 2, Figure 2 is a schematic diagram of the distribution of sliding windows to be divided according to an embodiment of the present invention. In Figure 2, the sliding windows to be divided are in the adhesion between colonies, and their corresponding distribution is relatively dense. However, the correspondence of the non-edge sliding window to be segmented is relatively discrete. Therefore, a detailed analysis of the edge sliding window is carried out.

Further, in some embodiments of the present invention, filtering out edge sliding windows from sliding windows to be segmented based on edge selection indicators of all sliding windows to be segmented includes: selecting sliding windows to be segmented whose edge selection indicators are less than a preset edge threshold. As an edge sliding window.

The preset edge threshold is the threshold value of the edge selection index. In the embodiment of the present invention, the preset edge threshold can be, for example, 0.25. That is to say, when the edge selection index is less than 0.25, the sliding window to be divided is used as the edge sliding window. Window, of course, the preset edge threshold can also be adjusted according to actual detection requirements, and there is no restriction on this.

In the embodiment of the present invention, the sliding window to be divided at the edge of the colony has a relatively small distance from other sliding windows to be divided. Therefore, the present invention uses the distance between the sliding windows to be divided for effective judgment to ensure the final result. Accuracy and reliability of edge sliding window selection.

S105: Segment the target connected domain according to the center point in the edge sliding window to obtain bacterial colonies, and treat each connected domain to be tested except the target connected domain as a bacterial colony.

After confirming the edge sliding window, the present invention can analyze specific bacterial colonies according to the position of the edge sliding window.

Further, in some embodiments of the present invention, segmenting the target connected domain according to the edge lines in the edge sliding window to obtain bacterial colonies includes: connecting the center points of the two nearest edge sliding windows, dividing the target connected domain Divide it into at least two sub-regions and treat each sub-region as a bacterial colony.

In the embodiment of the present invention, the center point of the edge sliding window can be used as the edge point for corresponding colony segmentation, and then, the center points of the two closest edge sliding windows are connected to divide the target connected domain into at least two sub-regions, and each subregion as a bacterial colony.

Therefore, the connected domain to be tested except the target connected domain is regarded as a single bacterial colony, and the target connected domain is a connected domain in which multiple bacterial colonies are adhered to each other. Then the target connected domain is divided, and the connected domains that are adhered to each other are divided. , the corresponding bacterial colonies are obtained, thus the bacterial colonies can be divided more accurately and effectively.

This invention determines the colony area by acquiring the grayscale image of the surface of the culture medium, and then performs connected domain analysis and segments the connected domains based on sliding windows of preset size and non-overlapping to obtain an initial sliding window. Through the initial sliding window and the test The gray value of the pixels in the connected domain is selected from the initial sliding window. It can be understood that the bacterial colonies have discrete colonies on the outside, and there are also multiple bacterial colonies that adhere to each other. When multiple bacterial colonies adhere to each other, due to the The distribution of filaments makes the middle edge area blurred and the corresponding edge information is less. The effect obtained by directly using color block analysis and processing is poor. Therefore, the embodiment of the present invention performs specific analysis by setting non-overlapping sliding windows, making the analysis process The degree of refinement is higher and the effect is better; since a single connected domain to be tested can be a colony or a combination of multiple colonies, the present invention combines the area information of the connected domain to be tested itself and the information of the target sliding window in the connected domain to be tested. The quantity and position distribution determine the connected domain analysis index, and the target connected domain is screened out from the connected domain to be tested, so that the target connected domain containing multiple colonies can be accurately screened, and then the distance between the sliding windows to be segmented in the target connected domain can be accurately selected. Determine the edge sliding window, divide the target connected domain based on the edge sliding window, so that the mutually adherent colonies can be effectively divided, and combine the originally individually distributed connected domains to be tested to divide the bacterial colonies. In summary, the present invention filters out the existing bacterial colonies Connected domains with adhesion can be effectively divided into connected domains with adhesions, which can more accurately classify bacterial colonies based on grayscale differences and edge features, improve the precision of bacterial colony classification, and ensure the accuracy and accuracy of bacterial colony classification. reliability.

It should be noted that the above-mentioned order of the embodiments of the present invention is only for description and does not represent the advantages and disadvantages of the embodiments. The processes depicted in the figures do not necessarily require the specific order shown, or sequential order, to achieve desirable results. Multitasking and parallel processing are also possible or may be advantageous in certain implementations.

Each embodiment in this specification is described in a progressive manner. The same and similar parts between the various embodiments can be referred to each other. Each embodiment focuses on its differences from other embodiments.

Claims (10)

1. A method for accurately dividing bacterial colonies based on image features, characterized in that the method includes: Obtain a grayscale image of the surface of the culture medium, and determine the colony area of the grayscale image of the surface of the culture medium based on threshold segmentation; Perform connected domain analysis on the colony area to determine the connected domain to be tested, segment the connected domain based on a preset size and non-overlapping sliding window, and obtain an initial sliding window. According to the initial sliding window and the connected domain to be tested The gray value of the pixel in the middle, selects the target sliding window from the initial sliding window; According to the area of each connected domain to be measured, the areas of all target sliding windows in the connected domain to be measured, and the position distribution of the target sliding windows, determine the connected domain analysis index of the connected domain to be measured; according to the connected domain analysis The indicator selects the target connected domain from the connected domain to be measured; All other target sliding windows in the target connected domain except the outermost target sliding window are used as the sliding windows to be divided; according to the distance between the sliding window to be divided and the nearest preset number of other sliding windows to be divided. The distance is used to select edge sliding windows from the sliding windows to be segmented; The target connected domain is segmented according to the center point in the edge sliding window to obtain bacterial colonies, and each connected domain to be tested except the target connected domain is regarded as a bacterial colony. 2. A method for accurately dividing bacterial colonies based on image features as claimed in claim 1, characterized in that, according to the initial sliding window and the gray value of the pixels in the connected domain to be measured, from the initial Select the target sliding window in the sliding window, including: The normalized value of the gray value information entropy of all pixels in the initial sliding window is used as the sliding window confusion degree; The mean value of the gray value of all pixels in the initial sliding window is used as the sliding window mean, and the standard deviation of the gray value of all pixels in the initial sliding window is used as the sliding window standard deviation; calculate the sliding window mean and the The product of the sliding window standard deviation is used as the sliding window characteristic parameter; The mean value of the gray value of all pixels in the connected domain to be measured is used as the mean value to be measured, and the standard deviation of the gray value of all pixels in the connected domain to be measured is used as the standard deviation to be measured; calculate the mean value to be measured and the mean value to be measured The product of the standard deviation to be measured is used as the characteristic parameter to be measured; Determine the sliding window mutation coefficient according to the sliding window characteristic parameter and the characteristic parameter to be measured; Calculate the product of the sliding window chaos degree and the sliding window mutation coefficient as the sliding window screening coefficient; A target sliding window is selected from the initial sliding windows according to the sliding window screening coefficient of each initial sliding window. 3. A method for accurately dividing bacterial colonies based on image features as claimed in claim 2, characterized in that determining the sliding window mutation coefficient according to the sliding window characteristic parameters and the characteristic parameters to be measured includes: Calculate the absolute value of the difference between the sliding window characteristic parameter and the characteristic parameter to be measured as the characteristic difference; The sliding window mutation coefficient is determined according to the characteristic difference and the characteristic parameter to be measured, wherein the characteristic difference is positively correlated with the sliding window mutation coefficient, and the characteristic parameter to be measured is negatively correlated with the sliding window mutation coefficient, so The value of the sliding window mutation coefficient is a normalized value. 4. A method for accurately classifying bacterial colonies based on image features according to claim 2, wherein the target sliding window is selected from the initial sliding window according to the sliding window screening coefficient of each initial sliding window. Windows, including: The initial sliding window with the sliding window screening coefficient greater than the preset screening coefficient threshold is used as the target sliding window. 5. A method for accurately dividing bacterial colonies based on image features as claimed in claim 1, wherein the method is based on the area of each connected domain to be measured and the area of all target sliding windows in the connected domain to be measured. and the position distribution of the target sliding window to determine the connected domain analysis indicators of the connected domain to be measured, including: Calculate the ratio of the area of the connected domain to be measured to the areas of all target sliding windows in the connected domain to be measured as the area influence coefficient; Use any target sliding window in the connected domain to be tested as the sliding window to be analyzed; determine the center point of the target sliding window that is closest to the center point of the sliding window to be analyzed, and calculate the distance value between the two center points The normalized value of is used as the distance index of the sliding window to be analyzed; the mean value of the distance index of all target sliding windows is used as the distance influence coefficient; The normalized value of the product of the area influence coefficient and the distance influence coefficient is used as the connected domain analysis index. 6. A method for accurately classifying bacterial colonies based on image features as claimed in claim 1, wherein the step of screening out the target connected domain from the connected domain to be tested according to the connected domain analysis index includes: The connected domain to be tested whose connected domain analysis index is smaller than the preset indicator threshold is used as the target connected domain. 7. A method for accurately dividing bacterial colonies based on image features as claimed in claim 1, wherein the method is based on the distance between the sliding window to be divided and the nearest preset number of other sliding windows to be divided. , filter out edge sliding windows from the sliding windows to be divided, including: The sum of the distances between any sliding window to be divided and the nearest preset number of other sliding windows to be divided is used as the sum of the dividing distances; Perform normalization processing on the segmentation distance and value to obtain an edge selection index; Edge sliding windows are filtered out from the sliding windows to be divided according to the edge selection indicators of all the sliding windows to be divided. 8. A method for accurately dividing bacterial colonies based on image features as claimed in claim 7, wherein the edge selection index of all the sliding windows to be segmented is selected from the sliding windows to be segmented. Sliding windows including: The sliding window to be segmented whose edge selection index is smaller than the preset edge threshold is used as an edge sliding window. 9. A method for accurately dividing bacterial colonies based on image features as claimed in claim 1, wherein the target connected domain is segmented according to the edge line in the edge sliding window to obtain bacterial colonies, include: Connect the center points of the two closest edge sliding windows, divide the target connected domain into at least two sub-regions, and treat each sub-region as a bacterial colony. 10. A method for accurately classifying bacterial colonies based on image features according to claim 7, wherein the preset number is 10.