A microbial microscopic image target feature recognition method and device and a storage medium

Abstract

The present application relates to a kind of microbial microscopic image target feature identification method, device and storage medium, applied to image processing technical field, comprising: by marking the image on the target to be identified, convert the target to be identified into different scale space and construct multiple groups of Gaussian pyramid and convert into feature pyramid, according to feature pyramid, obtain feature point, obtain the feature parameter of each feature point, when the target needs to be identified, convert the image to be identified into multiple groups of Gaussian pyramid, and the pixel point in Gaussian pyramid is matched with feature point, if the number of pixel point on the image to be identified is matched exceeds the number of pre-set, then consider that there is identification target on the image;By the present application, without using convolutional neural network, microorganism image features can be quickly and accurately automatically identified, and then automatically identify white blood cells, clue cells, spores, blastospores, hyphae, trichomonas and other microorganisms.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of image processing, in particular to a microorganism microscopic image target feature recognition method and device and storage medium. BACKGROUND

[0002] When a physician performs a microorganism diagnosis on a body fluid sample or a skin sample to be detected, a microorganism microscopic image is obtained by observing and imaging a smear of the body fluid sample or the skin sample through a microscope, and a microorganism flora type is recognized, marked, scaled and counted. The most important thing is how to quickly and accurately recognize the microorganism, so that the corresponding microorganism type can be used for marking, scaling and counting of the microorganism flora type. Moreover, quickly and effectively recognizing the microorganism type in the sample can greatly reduce the workload of the clinician. For example, recognizing leukocytes, clue cells, spores, blastospores, hyphae and trichomonas. Specifically, effectively recognizing clue cells is a necessary condition for clinical diagnosis of bacterial vaginosis, because clue cells are an important specific and sensitive diagnostic indicator of bacterial vaginosis. Recognizing trichomonas can effectively distinguish trichomonas vaginitis. Recognizing spores, blastospores and pseudohyphae can distinguish whether it is fungal / mold vaginitis. Recognizing leukocytes can effectively distinguish inflammatory reactions.

[0003] At present, most of the auxiliary recognition devices for body fluid samples and skin samples mainly recognize through a convolutional neural network based on artificial intelligence. However, the recognition through the convolutional neural network has a high requirement for the hardware devices of the hospital, a large amount of manpower needs to be invested in the early stage, and it cannot timely change the clinical demand changes, and the flexibility is relatively poor. SUMMARY

[0004] Therefore, the present application aims to provide a microorganism microscopic image target feature recognition method and device and storage medium, so as to solve the problem in the prior art that the recognition of a specific target through a convolutional neural network has a high requirement for the hardware devices of the hospital, a large amount of manpower needs to be invested in the early stage, and it cannot timely change the clinical demand changes, and the flexibility is relatively poor. Through the present application, the features of leukocytes, clue cells, spores, blastospores, hyphae and trichomonas in the body fluid sample and the skin sample can be efficiently and accurately recognized, the microorganism flora type can be accurately recognized, and the microorganism flora type can be marked, scaled and counted.

[0005] According to a first aspect of an embodiment of the present application, a microorganism microscopic image target feature recognition method is provided, and the method comprises:

[0006] obtaining a fluorescently stained microorganism microscopic image containing a target to be recognized, and labeling the target to be recognized in the microorganism microscopic image;

[0007] constructing a plurality of first Gaussian pyramids of the labeled to-be-recognized target in the microbial microscopic image by a scale-variable Gaussian function, the first Gaussian pyramid having a plurality of layers of scale spaces with different scale space factors, and each layer of scale space containing a Gaussian image of the to-be-recognized target;

[0008] converting the first Gaussian pyramid into a feature pyramid, each group of first Gaussian pyramids corresponding to a group of feature pyramids;

[0009] obtaining a feature point set according to the feature pyramid, each group of feature pyramids corresponding to a feature point set;

[0010] obtaining the modulus and direction of each feature point in the feature point set, and taking the modulus and direction of the feature point as the feature of the feature point;

[0011] obtaining a to-be-recognized fluorescently stained microbial microscopic image, and constructing a plurality of second Gaussian pyramids of the to-be-recognized fluorescently stained microbial microscopic image by a scale-variable Gaussian function;

[0012] matching the pixel points in the plurality of second Gaussian pyramids with the features of the feature points, and if the matched pixel points in the plurality of second Gaussian pyramids exceed a preset number, determining that the to-be-recognized fluorescently stained microbial microscopic image contains the to-be-recognized target.

[0013] Preferably,

[0014] the constructing of the plurality of first Gaussian pyramids of the labeled to-be-recognized target in the microbial microscopic image by the scale-variable Gaussian function comprises:

[0015] firstly enlarging the original resolution of the to-be-recognized target by one time as the resolution of a first group of first Gaussian pyramids, the scale factor of the first layer of scale space of the first group of first Gaussian pyramids being σ, and the ratio between the scale factors of the scale spaces of adjacent layers being k, to obtain a first group of first Gaussian pyramids with a plurality of layers of scale spaces;

[0016] secondly reducing the resolution of the first group of first Gaussian pyramids by a preset ratio as the resolution of a second group of first Gaussian pyramids, the scale factor of the first layer of scale space of the second group of first Gaussian pyramids being 2σ, and the ratio between the scale factors of the scale spaces of adjacent layers being k, to obtain a second group of first Gaussian pyramids with a plurality of layers of scale spaces, until a preset number of first Gaussian pyramids are obtained.

[0017] Preferably,

[0018] the converting of the first Gaussian pyramid into a feature pyramid comprises:

[0019] Subtracting the Gaussian images of two adjacent layers in each group of the first Gaussian pyramid and adding e, a feature pyramid is obtained, and each group of the first Gaussian pyramid corresponds to a group of the feature pyramid.

[0020] Preferably,

[0021] The feature point set is obtained according to the feature pyramid.

[0022] In the feature pyramid, if the feature value of any pixel point at any layer is greater than or less than the feature values of the 8-neighborhood pixel points, and the feature values of the corresponding pixel points at the adjacent layer are also greater than or less than the feature values of the 8-neighborhood pixel points, the pixel point is considered as a feature point, and all the feature points in each group of the feature pyramid form a feature point set.

[0023] Preferably,

[0024] The feature of the feature point further includes:

[0025] The modulus and direction of the most similar pixel point of the feature point are obtained from the adjacent pixel points of the feature point, and the modulus and direction of the most similar pixel point of the feature point are taken as the feature of the feature point.

[0026] Preferably,

[0027] The most similar pixel point of the feature point includes:

[0028] After the direction of the feature point is obtained, a region image is obtained with the feature point as the center and a preset radius, the directions of all the pixel points in the region image are obtained, and the pixel points with a direction difference less than a preset direction difference from the direction of the feature point are taken as the most similar pixel points of the feature point.

[0029] Preferably,

[0030] The pixel points in the multiple groups of the second Gaussian pyramid are matched with the feature of the feature point, and if the matched pixel points in the multiple groups of the second Gaussian pyramid exceed a preset number, it is determined that the to-be-identified fluorescently dyed microbial microscopic image contains the to-be-identified target, which includes:

[0031] After the second Gaussian pyramid with the same number of groups as the first Gaussian pyramid is obtained, in the first group of the second Gaussian pyramid, the pixel points with the same feature as the feature points in the feature point set of the first group of the first Gaussian pyramid are obtained, and the number of the matched pixel points in the first group of the second Gaussian pyramid is counted.

[0032] In the second group of the second Gaussian pyramid, pixel points having the same feature as the feature points in the feature point set of the first group of the second Gaussian pyramid are obtained, the number of the matched pixel points in the second group of the second Gaussian pyramid is counted, until the pixel points in the second Gaussian pyramid of all groups are matched, the number of the matched pixel points of all groups is added, and if the number is greater than a preset number, it is determined that the to-be-identified fluorescently dyed microbial microscopic image contains the to-be-identified target.

[0033] According to a second aspect of the embodiment of the present application, a microbial microscopic image target feature identification device is provided, and the device comprises:

[0034] The labeling module is configured to obtain a fluorescently dyed microbial microscopic image containing a to-be-identified target, and label the to-be-identified target in the microbial microscopic image.

[0035] The image conversion module is configured to construct a plurality of first Gaussian pyramids of the labeled to-be-identified target in the microbial microscopic image by using a scale-variable Gaussian function, wherein the first Gaussian pyramids have a plurality of layers of scale spaces with different scale space factors, and each layer of scale space contains a Gaussian image of the to-be-identified target.

[0036] The feature pyramid acquisition module is configured to convert the first Gaussian pyramids into feature pyramids, wherein each group of first Gaussian pyramids corresponds to a group of feature pyramids.

[0037] The feature point acquisition module is configured to acquire a feature point set according to the feature pyramids, wherein each group of feature pyramids corresponds to a feature point set.

[0038] The feature acquisition module is configured to acquire the magnitude and direction of each feature point in the feature point set, and take the magnitude and direction of the feature point as the feature of the feature point.

[0039] The input module is configured to acquire a to-be-identified fluorescently dyed microbial microscopic image, and construct a plurality of second Gaussian pyramids of the to-be-identified fluorescently dyed microbial microscopic image by using a scale-variable Gaussian function.

[0040] The identification module is configured to match the pixel points in the plurality of second Gaussian pyramids with the features of the feature points, and if the number of the matched pixel points in the plurality of second Gaussian pyramids exceeds a preset number, it is determined that the to-be-identified fluorescently dyed microbial microscopic image contains the to-be-identified target.

[0041] According to a third aspect of the embodiment of the present application, a storage medium is provided, and the storage medium stores a computer program, wherein the computer program is executed by a host computer to implement each step in the above method.

[0042] The technical solution provided by the embodiment of the present application can have the following beneficial effects:

[0043] The application converts the to-be-identified target into different scale spaces by labeling and cutting the to-be-identified target on the image, converts the to-be-identified target into different scale spaces by constructing multiple groups of Gaussian pyramids and converting them into feature pyramids, obtains feature points according to the feature pyramids, obtains the feature parameters of each feature point, that is, obtains the feature parameters of the to-be-identified target, and when subsequent target recognition is needed, the image also needs to be converted into multiple groups of Gaussian pyramids, and the pixel points in the Gaussian pyramids are compared with the feature points, if the feature parameters of the pixel points meet the feature parameters of the feature points, it is considered that the pixel points match the feature points, and if the number of matched feature points on the image exceeds a preset number, it is considered that the image contains an identified target. Through the application, the target can be recognized without using a convolutional neural network, saving the time for training the neural network, and the hardware requirement is low. When the identified target needs to be replaced, only the feature parameters of the new identified target need to be obtained, and the flexibility is high. Through the application, the features of white blood cells, clue cells, spores, budding spores, hyphae, trichomonas in body fluid samples and skin samples can be efficiently and accurately recognized, the types of microbial flora can be accurately and automatically recognized, and the types of microbial flora can be marked, scaled and counted.

[0044] It should be understood that the above general description and the following detailed description are only exemplary and explanatory, and cannot limit the application. BRIEF DESCRIPTION OF DRAWINGS

[0045] The accompanying drawings, which are incorporated into the specification and constitute a part of the specification, illustrate embodiments consistent with the application and, together with the specification, serve to explain the principles of the application.

[0046] Figure 1 is a flowchart of a microbial microscopic image target feature recognition method according to an exemplary embodiment;

[0047] Figure 2 is a system diagram of a microbial microscopic image target feature recognition device according to an exemplary embodiment;

[0048] In the drawings: 1 - labeling module, 2 - image conversion module, 3 - feature pyramid acquisition module, 4 - feature point acquisition module, 5 - feature acquisition module, 6 - input module, 7 - recognition module DETAILED DESCRIPTION

[0049] The exemplary embodiments will be described in detail herein with reference to the attached drawings. The description below refers to the accompanying drawings, which show, by way of example, specific embodiments with which this application can be practiced. The following detailed description is not intended to limit the application. Rather, the following detailed description includes specific details for the purpose of providing a thorough understanding of the inventive concepts. However, it will be apparent to those skilled in the art that the inventive concepts can be practiced without these specific details. In some instances, well-known structures and functions have not been described in detail in order to avoid obscuring the inventive concepts.

[0050] Embodiment One

[0051] Figure 1 FIG. 1 is a flowchart of a method for identifying a target feature in a microscopic image of microorganisms according to an exemplary embodiment. As shown in FIG. 1, the method includes the following steps: Figure 1

[0052] S1, obtaining a fluorescently stained microscopic image of microorganisms containing a target to be identified, and labeling the target to be identified in the microscopic image of microorganisms;

[0053] The fluorescent staining can be double fluorescent staining, multiple fluorescent staining, or can be replaced by Gram staining. The present application does not limit the method of staining.

[0054] S2, constructing a plurality of first Gaussian pyramids of the labeled target to be identified in the microscopic image of microorganisms by a scale- variable Gaussian function, the first Gaussian pyramid having a plurality of layers of scale spaces with different scale space factors, and each layer of scale space containing a Gaussian image of the target to be identified;

[0055] S3, converting the first Gaussian pyramid into a feature pyramid, each group of first Gaussian pyramids corresponding to a feature pyramid;

[0056] S4, obtaining a set of feature points according to the feature pyramid, each group of feature pyramids corresponding to a set of feature points;

[0057] S5, obtaining the magnitude and direction of each feature point in the set of feature points, and taking the magnitude and direction of the feature point as the feature of the feature point;

[0058] S6, obtaining a fluorescently stained microscopic image to be identified, and constructing a plurality of second Gaussian pyramids of the fluorescently stained microscopic image to be identified by a scale- variable Gaussian function;

[0059] S7, matching the features of the feature points with the pixel points in the plurality of second Gaussian pyramids, and if the number of matched pixel points in the plurality of second Gaussian pyramids exceeds a predetermined number, determining that the fluorescently stained microscopic image to be identified contains the target to be identified;

[0060] ​It can be understood that in the present application, a fluorescently stained microbial microscopic image containing a to-be-identified target is acquired, and the to-be-identified target is labeled in the microbial microscopic image, wherein the to-be-identified target includes spores, blastospores, and pseudohyphae, etc. In the range visible to the human eye, the human eye can distinguish objects regardless of their size, but when the picture is put into a computing device, the device cannot provide the size of the object. Therefore, in order to effectively identify the to-be-identified target, a scale space is first constructed, because the Gaussian convolution kernel is the only linear kernel to realize scale change. A plurality of first Gaussian pyramids are constructed through a scale-variable Gaussian function. In the present embodiment and the following embodiments, the number of groups of the Gaussian pyramids is 6, and the number of layers of each Gaussian pyramid is 3. Each first Gaussian pyramid is converted into a feature pyramid, and feature points are screened through the feature pyramid. Each first Gaussian pyramid contains a plurality of feature points, and the feature points of each first Gaussian pyramid are collected to obtain a feature point set. The magnitude and direction of each feature point in the feature point set are calculated, and the magnitude and direction of each feature point are taken as the feature of the feature point, that is, the feature parameter of the feature point. In this way, the feature parameters of the identification targets such as spores, blastospores, and pseudohyphae in the image are obtained. Then, when target identification is needed, the image to be identified is converted into a multi-scale space using the same Gaussian function to construct 6 second Gaussian pyramids. It is worth emphasizing that when pixel point matching is performed, it needs to be one-to-one, that is, the feature points of the first group of first Gaussian pyramids are matched with the pixel points of the first group of second Gaussian pyramids, the feature points of the second group of first Gaussian pyramids are matched with the pixel points of the second group of second Gaussian pyramids, and if the number of pixel points matched in the 6 groups is more than 1000 points, it is considered that the image contains spores, blastospores, and pseudohyphae, etc. In the present application, the to-be-identified target is labeled and cut on the image, and the to-be-identified target is converted into different scale spaces. A plurality of Gaussian pyramids are constructed and converted into feature pyramids, and feature points are obtained according to the feature pyramids. The feature parameters of each feature point are obtained, that is, the feature parameters of the to-be-identified target are obtained. When target identification is needed in the future, the image also needs to be converted into a plurality of Gaussian pyramids, and the pixel points in the Gaussian pyramids are compared with the feature points. If the feature parameters of the pixel points meet the feature parameters of the feature points, it is considered that the pixel points are matched with the feature points. If the number of matched feature points on the image exceeds a predetermined number, it is considered that the image contains the identification target. Through the present application, target identification can be realized without using a convolutional neural network, which saves the time for training the neural network and has low requirements for hardware. When the identification target needs to be replaced, only the feature parameters of the new identification target need to be obtained, which has high flexibility.

[0061] Preferably,

[0062] The labeled to-be-recognized target in the microbial microscopic image is constructed into a plurality of first Gaussian pyramids through a scale-variable Gaussian function.

[0063] The original resolution of the to-be-recognized target is first enlarged by one time as the resolution of a first group of first Gaussian pyramids, the scale factor of the first layer scale space of the first group of first Gaussian pyramids is σ, and the ratio between the scale factors of the scale spaces of adjacent layers is k, to obtain a plurality of layers of scale spaces of the first group of first Gaussian pyramids.

[0064] The resolution of the first group of first Gaussian pyramids is reduced by a preset ratio as the resolution of a second group of first Gaussian pyramids, the scale factor of the first layer scale space of the second group of first Gaussian pyramids is 2σ, and the ratio between the scale factors of the scale spaces of adjacent layers is k, to obtain a plurality of layers of scale spaces of the second group of first Gaussian pyramids, until a preset number of first Gaussian pyramids is obtained.

[0065] It can be understood that the present application first defines a scale-variable Gaussian function:

[0066]

[0067] According to the formula, the scale space of a two-dimensional image of fluorescent staining can be defined as:

[0068] L(x,y,σ)=G(x,y,σ)*I(x,y)

[0069] (x,y) is not only a spatial coordinate, but also a scale coordinate, σ is a scale space factor, which is a standard deviation of a Gaussian normal distribution, reflecting the degree of image blurring, the greater the value, the more blurred the image, and the corresponding scale is also larger, and I(x,y) represents the intensity of the pixel in the gray image; before feature extraction, a Gaussian scale space needs to be constructed, and the construction is performed in the following manner: the image is first pyramid down-sampled, and then Gaussian filtering is performed, and the specific operation manner is that different σ is used for each layer of image pyramid to make the image more blurred, and after this operation, each layer of pyramid has a plurality of images blurred by Gaussian, and the specific operation is as follows: because the fluorescent gynecological image is prevented from edge distortion, only 1024*512 of each image is used, so the number of groups of the pyramid used is Log2min(1024, 512)-3, the number of groups of the pyramid constructed is 6, and then the number of layers of each group is taken as 3 to start construction, first, the 0th group is increased by one time, and the width and height of the image are doubled to become 2048*1024(I0), so that more detailed information in the image can be obtained, the 0th layer is I0*G(x,y,σ0), the 1st layer is I0*G(x,y,kσ0), and the 2nd layer is I0*G(x,y,k 2k is a scale factor of two adjacent Gaussian scale spaces, the first group is down-sampling I0 to 1024*512 (I1), the 0th layer is I1*G(x, y, 2σ0), the 1st layer is I1*G(x, y, 2kσ0), the second layer is I1*G(x, y, 2k 2 σ0), and so on, until the sixth group of I0 becomes 64*32 (I5), the 0th layer of the sixth group is I5*G(x, y, 6σ0), the 1st layer of the sixth group is I5*G(x, y, 6kσ0), the second layer of the sixth group is I5*G(x, y, 6k 2 σ0).

[0070] Preferably,

[0071] The converting the first Gaussian pyramid into a feature pyramid comprises:

[0072] Subtracting the Gaussian images of two adjacent layers in each group of the first Gaussian pyramid and adding e to obtain a feature pyramid, each group of the first Gaussian pyramid corresponding to a group of the feature pyramid;

[0073] It can be understood that for feature point extraction in an image, the following formula is used:

[0074] D(x, y, σ) = L(x, y, kσ) - L(x, y, σ) + e

[0075] In the formula, k is a scale factor of two adjacent Gaussian scale spaces, so the feature extraction scheme of the present application is to subtract the images of two adjacent Gaussian spaces and add e, which can convert the three-layer Gaussian pyramid into a two-layer feature pyramid.

[0076] Preferably,

[0077] The obtaining a feature point set according to the feature pyramid comprises:

[0078] In the feature pyramid, any one pixel point is selected at any layer, if the feature value of the pixel point is greater than or less than the feature values of the pixel points in the surrounding 8-neighborhood, and the feature values of the corresponding pixel points in the adjacent layers are also greater than or less than the feature values of the pixel points in the surrounding 8-neighborhood, the pixel point is considered as a feature point, and all the feature points of each group of the feature pyramid are a feature point set;

[0079] It can be understood that in the feature pyramid, taking the two-layer feature pyramid as an example, if the D value of any pixel point in one layer is greater than or less than the D value of the 8-neighborhood pixel points around the pixel point, and the D value of the corresponding pixel point in another layer is also greater than or less than the D value of the 8-neighborhood pixel points around the pixel point, the pixel point is considered as a feature point, and 6 feature point sets are obtained by screening the feature points of each feature pyramid. After obtaining the feature points, the invariance of image rotation needs to be realized, so the direction of the feature point needs to be assigned. The gradient distribution characteristics of the neighborhood pixels of the feature point are used to determine the direction parameter of the feature point, and then the stable direction of the local structure of the key point is obtained by using the gradient histogram of the image. Specifically, first, the scale σ of the feature point is determined after the feature point is found. Next, the modulus and direction of the region image with 3*1.5σ as the radius and the feature point as the center are calculated, and the calculation formula is as follows:

[0080]

[0081]

[0082] In the formula, m(x, y) represents the modulus of the feature point, and θ(x, y) represents the direction of the feature point.

[0083] Preferably,

[0084] The features of the feature point further include:

[0085] The modulus and direction of the most similar pixel point of the feature point are obtained, and the modulus and direction of the most similar pixel point of the feature point are taken as the features of the feature point.

[0086] It can be understood that in order to improve the matching accuracy and reduce the probability of false matching in the subsequent pixel point matching process, the most similar pixel point of the feature point also needs to be used to participate in the matching of the pixel point, that is, the modulus and direction of the most similar pixel point are taken as the feature parameters of the feature point.

[0087] Preferably,

[0088] The most similar pixel point of the feature point includes:

[0089] After obtaining the direction of the feature point, a region image is obtained with the feature point as the center and a predetermined radius, the directions of all pixel points in the region image are obtained, and the pixel points with a direction difference less than a predetermined direction difference from the direction of the feature point are taken as the most similar pixel points of the feature point.

[0090] It can be understood that, in addition to the modulus and direction of the feature point, the feature parameters of the feature point also include the gradient direction and amplitude of other pixel points in the region image. In the gradient histogram, when the difference between the direction of a pixel point and the direction of the feature point is not more than 20%, the direction and modulus of the pixel point are used as the feature parameters of the feature point, which are used for subsequent pixel point matching.

[0091] Preferably,

[0092] The matching of the pixel points in the multiple sets of second Gaussian pyramids with the features of the feature points is performed. If the number of matched pixel points in the multiple sets of second Gaussian pyramids exceeds a preset number, it is determined that the to-be-identified fluorescently dyed microbial microscopic image contains the to-be-identified target.

[0093] After obtaining the second Gaussian pyramids with the same number of sets as the first Gaussian pyramids, in the first set of second Gaussian pyramids, pixel points with the same features as the feature points in the feature point set of the first set of first Gaussian pyramids are obtained, and the number of matched pixel points in the first set of second Gaussian pyramids is counted.

[0094] In the second set of second Gaussian pyramids, pixel points with the same features as the feature points in the feature point set of the second set of first Gaussian pyramids are obtained, and the number of matched pixel points in the second set of second Gaussian pyramids is counted. Until the pixel point matching of all sets of second Gaussian pyramids is completed, the numbers of matched pixel points of all sets are added. If the number is greater than a preset number, it is determined that the to-be-identified fluorescently dyed microbial microscopic image contains the to-be-identified target.

[0095] It can be understood that, when the to-be-identified target needs to be identified, the to-be-identified image is converted into a multi-scale space by the Gaussian function defined above, and six sets of Gaussian pyramids are constructed according to the above scheme. In the first set of second Gaussian pyramids, it is determined whether there is a pixel point with the same modulus and direction as a feature point in the feature point set in the first set of first Gaussian pyramids. If there is, it is further determined whether there is a most similar pixel point in the region image of the pixel point, which has the same modulus and direction as the most similar pixel point in the feature point set. If there is, it is considered that the pixel point matches the feature point. After the matching of the first set of second Gaussian pyramids is completed, the matching of the second set of second Gaussian pyramids is continued. After the matching of the six sets of second Gaussian pyramids is completed, the numbers of matched pixel points of the six sets are added. If the number is greater than 1000 pixel points, it is considered that the to-be-identified target (spores, germ spores, and pseudohyphae, etc.) exists in the image.

[0096] Example Two

[0097] Figure 2 is a system schematic diagram of a microorganism microscopic image target feature recognition device according to an exemplary embodiment, comprising:

[0098] The labeling module 1 is configured to obtain a fluorescently stained microorganism microscopic image containing a target to be recognized, and label the target to be recognized in the microorganism microscopic image.

[0099] The image conversion module 2 is configured to construct a plurality of groups of first Gaussian pyramids for the labeled target to be recognized in the microorganism microscopic image by using a scale-variable Gaussian function, wherein the first Gaussian pyramids have a plurality of layers of scale spaces with different scale space factors, and each layer of scale space contains a Gaussian image of the target to be recognized.

[0100] The feature pyramid acquisition module 3 is configured to convert the first Gaussian pyramids into feature pyramids, wherein each group of first Gaussian pyramids corresponds to a group of feature pyramids.

[0101] The feature point acquisition module 4 is configured to acquire a feature point set according to the feature pyramids, wherein each group of feature pyramids corresponds to a feature point set.

[0102] The feature acquisition module 5 is configured to acquire the magnitude and direction of each feature point in the feature point set, and take the magnitude and direction of the feature point as the feature of the feature point.

[0103] The input module 6 is configured to acquire a fluorescently stained microorganism microscopic image to be recognized, and construct a plurality of groups of second Gaussian pyramids for the fluorescently stained microorganism microscopic image to be recognized by using a scale-variable Gaussian function.

[0104] The recognition module 7 is configured to match the pixel points in the plurality of groups of second Gaussian pyramids with the features of the feature points, and if the matched pixel points in the plurality of groups of second Gaussian pyramids exceed a preset number, it is determined that the fluorescently stained microorganism microscopic image to be recognized contains the target to be recognized.

[0105] It can be understood that the application is used for obtaining a fluorescently stained microorganism microscopic image containing a to-be-identified target through the labeling module 1, labeling the to-be-identified target in the microorganism microscopic image; the image conversion module 2 is used for constructing a plurality of groups of first Gaussian pyramids of the to-be-identified target labeled in the microorganism microscopic image through a scale-variable Gaussian function, the first Gaussian pyramid has a plurality of layers of scale spaces with different scale space factors, and each layer of scale space contains a Gaussian image of the to-be-identified target; the feature pyramid acquisition module 3 is used for converting the first Gaussian pyramid into a feature pyramid, each group of first Gaussian pyramids corresponds to a feature pyramid; the feature point acquisition module 4 is used for acquiring a feature point set according to the feature pyramid, and each group of feature pyramids corresponds to a feature point set; the feature acquisition module 5 is used for acquiring the magnitude and direction of each feature point in the feature point set, and taking the magnitude and direction of the feature point as the feature of the feature point; the input module 6 is used for acquiring a to-be-identified fluorescently stained microorganism microscopic image, and constructing a plurality of groups of second Gaussian pyramids of the to-be-identified fluorescently stained microorganism microscopic image through a scale-variable Gaussian function; the recognition module 7 is used for matching the pixel points in the plurality of groups of second Gaussian pyramids with the features of the feature points, and if the matched pixel points in the plurality of groups of second Gaussian pyramids exceed a preset number, it is determined that the to-be-identified fluorescently stained microorganism microscopic image contains the to-be-identified target; the application cuts and labels the to-be-identified target on the image, converts the to-be-identified target into different scale spaces, constructs a plurality of groups of Gaussian pyramids and converts them into feature pyramids, acquires feature points according to the feature pyramids, and acquires the feature parameters of each feature point, that is, the feature parameters of the to-be-identified target are obtained. When the target needs to be recognized later, the image also needs to be converted into a plurality of groups of Gaussian pyramids, and the pixel points in the Gaussian pyramids are compared with the feature points. If the feature parameters of the pixel points meet the feature parameters of the feature points, it is considered that the pixel points match the feature points. If the matched feature points on the image exceed a preset number, it is considered that the image contains the to-be-identified target. Through the application, the target can be recognized without using a convolutional neural network, the time for training the neural network is saved, the requirement for hardware is low, and when the to-be-identified target needs to be replaced, only the feature parameters of the new to-be-identified target need to be acquired, and the flexibility is high.

[0106] By installing the software of the microbial microscopic image target feature recognition method, the features of leukocytes, clue cells, spores, blastospores, hyphae and trichomonads in the body fluid sample and the skin sample can be efficiently and accurately recognized, the accurate recognition of the microbial flora types is realized, the images with microbial targets are screened out and provided to the clinical doctors or the testing personnel, the final confirmation is made by the clinical doctors or the testing personnel, the interpretation of the microorganisms in the sample is assisted, the analysis results are output to the software report result module, printing and archiving are performed by the user, and finally the purpose of auxiliary diagnosis is achieved.

[0107] Embodiment three:

[0108] The embodiment provides a storage medium, and the storage medium stores a computer program.

[0109] It can be understood that the storage medium mentioned above can be a read-only memory, a disk or an optical disk.

[0110] It can be understood that the same or similar parts in the above embodiments can be mutually referred to, and the content not described in detail in some embodiments can be referred to the same or similar content in other embodiments.

[0111] It should be noted that, in the description of the present application, the terms "first", "second" and the like are only used for the purpose of description and should not be understood as indicating or implying relative importance. In addition, in the description of the present application, unless otherwise specified, the meaning of "a plurality of" is at least two.

[0112] Any process or method descriptions in flow charts or otherwise described herein represent embodiments that can be managed as one or more modules of executable instructions coded to perform specified logic functions or processes, and the scope of preferred embodiments of the present application includes additional implementations that can not be precisely shown or described herein, but that are nevertheless within the scope of the embodiments of the present application, including those that can be performed in a different order, in a substantially simultaneous manner, or in an opposite order, depending on the functionality involved, as would be understood by persons skilled in the art of the embodiments of the present application.

[0113] It should be understood that each part of the present application can be realized by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, a plurality of steps or methods can be realized by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if realized by hardware, and as in another embodiment, it can be realized by any one or a combination of the following technologies known in the art: discrete logic circuit with logic gate circuit for implementing logic function on data signal, application specific integrated circuit with suitable combination logic gate circuit, programmable gate array (PGA), field programmable gate array (FPGA) and the like.

[0114] Those skilled in the art of the present technology can understand that all or part of the steps carried out by the above-mentioned embodiment method can be completed by a program instructing the relevant hardware, and the program can be stored in a computer readable storage medium, and when executed, includes one or a combination of steps of the embodiment method.

[0115] In addition, each functional unit in each embodiment of the present application can be integrated into one processing module, or each unit can exist physically alone, or two or more units can be integrated into one module. The above-mentioned integrated module can be realized in the form of hardware or in the form of software functional module. The integrated module, if realized in the form of software functional module and sold or used as an independent product, can also be stored in a computer readable storage medium.

[0116] The above-mentioned storage medium can be a read-only memory, a magnetic disk or an optical disk, etc.

[0117] In the description of the present specification, the description of the terms "one embodiment", "some embodiments", "example", "specific example" or "some examples" means that the specific features, structures, materials or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present application. In the present specification, the illustrative description of the above-mentioned terms does not necessarily mean the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.

[0118] Although the embodiments of the present application have been shown and described above, it should be understood that the above-mentioned embodiments are exemplary and cannot be understood as limiting the present application, and those skilled in the art can make changes, modifications, replacements and variations to the above-mentioned embodiments within the scope of the present application.

Claims

1. A method for identifying target features in microscopic images of microorganisms, characterized in that, The method includes: Acquire fluorescently stained microscopic images of microorganisms containing the target to be identified, and annotate the target to be identified in the microscopic images of microorganisms; The target to be identified in the microscopic image of the microorganism is marked with a scale-variable Gaussian function to construct multiple sets of first Gaussian pyramids. The first Gaussian pyramid has multiple scale spaces with different scale spatial factors, and each scale space contains a Gaussian image of the target to be identified. The first Gaussian pyramid is converted into a feature pyramid, and each set of first Gaussian pyramids corresponds to a set of feature pyramids. The feature point set is obtained based on the feature pyramid, and each feature pyramid corresponds to a feature point set. Obtain the magnitude and orientation of each feature point in the feature point set, and use the magnitude and orientation of the feature point as the feature of that feature point; Acquire microscopic images of fluorescently stained microorganisms to be identified, and construct multiple sets of second Gaussian pyramids from the microscopic images of fluorescently stained microorganisms to be identified using a scale-variable Gaussian function; The pixels in multiple sets of second Gaussian pyramids are matched with the features of the feature points. If the number of matched pixels in multiple sets of second Gaussian pyramids exceeds a preset number, it is determined that the fluorescently stained microscopic image of the microorganism to be identified contains the target to be identified.

2. The method according to claim 1, characterized in that, The step of constructing multiple sets of first Gaussian pyramids by using a scale-variable Gaussian function to identify the marked targets in the microbial microscopic image includes: The original resolution of the target to be identified is first magnified by one time to obtain the resolution of the first group of first Gaussian pyramids. The scale factor of the first layer of the first group of first Gaussian pyramids is σ, and the ratio between the scale factors of adjacent layers is k, thus obtaining the first group of first Gaussian pyramids with multiple scale spaces. The resolution of the first Gaussian pyramid in the first group is reduced by a preset ratio to become the resolution of the second Gaussian pyramid. The scale factor of the first layer of the scale space of the second Gaussian pyramid is 2σ, and the ratio between the scale factors of adjacent layers is k. This process is repeated until a preset number of Gaussian pyramids are obtained.

3. The method according to claim 2, characterized in that, The process of converting the first Gaussian pyramid into a feature pyramid includes: Subtract the Gaussian images of two adjacent layers in each group of the first Gaussian pyramid and add e to obtain the feature pyramid. Each group of the first Gaussian pyramid corresponds to a set of feature pyramids.

4. The method according to claim 3, characterized in that, The step of obtaining the feature point set based on the feature pyramid includes: In the feature pyramid, if any pixel is selected in any layer and its feature value is greater than or less than the feature values ​​of its 8 neighboring pixels, and the feature value of the corresponding pixel in the adjacent layer is also greater than or less than the feature values ​​of its 8 neighboring pixels, then the pixel is considered a feature point. All the feature points in each feature pyramid constitute a feature point set.

5. The method according to claim 4, characterized in that, The features of the feature points also include: Obtain the magnitude and orientation of the pixel that is most similar to the feature point among the neighboring pixels of the feature point, and use the magnitude and orientation of the pixel that is most similar to the feature point as the feature of the feature point.

6. The method according to claim 5, characterized in that, The pixels most similar to this feature point include: After obtaining the direction of the feature point, a region image is obtained with the feature point as the center and a preset radius. The directions of all pixels in the region image are obtained respectively. The pixel whose direction difference with the feature point is less than the preset direction difference is taken as the most similar pixel of the feature point.

7. The method according to claim 6, characterized in that, The process of matching pixels in multiple sets of second Gaussian pyramids with the features of the feature points, and determining that the fluorescently stained microscopic image of the microorganism to be identified contains the target to be identified if the number of matched pixels in multiple sets of second Gaussian pyramids exceeds a preset number, includes: After obtaining the second Gaussian pyramid with the same number of groups as the first Gaussian pyramid, in the first group of the second Gaussian pyramid, obtain the pixels with the same features as the feature points in the feature point set of the first group of the first Gaussian pyramid, and count the number of matching pixels in the first group of the second Gaussian pyramid. In the second group of second Gaussian pyramids, pixels with the same features as the feature points in the feature point set of the first group of second Gaussian pyramids are obtained. The number of matching pixels in the second group of second Gaussian pyramids is counted until the pixels in the second Gaussian pyramids of all groups are matched. The number of matching pixels in all groups is added together. If it is greater than a preset number, it is determined that the fluorescent stained microscopic image of the microorganism to be identified contains the target to be identified.

8. A device for identifying target features in microscopic images of microorganisms, characterized in that, The device includes: Annotation module: used to acquire fluorescently stained microscopic images of microorganisms containing targets to be identified, and to annotate the targets to be identified in the microscopic images of microorganisms; Image conversion module: used to construct multiple sets of first Gaussian pyramids for the targets to be identified in the microscopic images of microorganisms using a scale-variable Gaussian function. The first Gaussian pyramid has multiple scale spaces with different scale spatial factors, and each scale space contains a Gaussian image of the target to be identified. Feature pyramid acquisition module: used to convert the first Gaussian pyramid into a feature pyramid, with each set of first Gaussian pyramids corresponding to a set of feature pyramids; Feature point acquisition module: used to acquire feature point sets based on the feature pyramid, with each feature pyramid corresponding to a feature point set; Feature acquisition module: used to acquire the magnitude and orientation of each feature point in the feature point set, and use the magnitude and orientation of the feature point as the feature of that feature point; Input module: used to acquire microscopic images of fluorescently stained microorganisms to be identified, and to construct multiple sets of second Gaussian pyramids from the microscopic images of fluorescently stained microorganisms to be identified using a scale-variable Gaussian function; Identification module: used to match the pixels in multiple sets of second Gaussian pyramids with the features of the feature points. If the number of matched pixels in multiple sets of second Gaussian pyramids exceeds a preset number, it is determined that the fluorescently stained microscopic image of the microorganism to be identified contains the target to be identified.

9. A storage medium, characterized in that, The storage medium stores a computer program, which, when executed by the main controller, implements each step of the method for identifying target features in microscopic images of microorganisms as described in any one of claims 1-7.

Images