An image recognition-based injury data analysis system and method

Abstract

The application discloses an image recognition-based injury data analysis system and method, relates to the technical field of image recognition and analysis, and is characterized in that the system and method are used for acquiring images of parts with injuries, setting an injury area in the images as a first type of feature area and setting other areas as a second type of feature area, pre-processing the acquired images, starting a shooting device, setting a cause-of-injury object in a target image as a target cause-of-injury object, identifying trace features in the target image, taking a first reference vector as a first evaluation dimension and taking an actual area corresponding to a first reference image as a second evaluation dimension, respectively evaluating the first reference image, calculating a time interval according to relevant evaluation values, reminding relevant personnel to acquire images of the injury parts again after the relevant interval time, comparing the acquired images with the images acquired last time, marking differences between the images acquired twice, and providing a reference for doctors.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of image recognition and analysis, in particular to a wound condition data analysis system and method based on image recognition. BACKGROUND

[0002] In people's daily life, it is inevitable to encounter sudden situations that cause physical injury. After the occurrence of the injury, the patient may not be able to be sent to the hospital in time due to the limitation of external conditions, or the patient subjectively thinks that it is inconvenient to go to the hospital, resulting in not wanting to go to the hospital for treatment. However, if the injury is not treated in time, the injury will have the risk of further deterioration, which may pose a greater threat to the health of the patient.

[0003] In actual situations, there is another situation that the patient does not pay enough attention to the injury at the time of the occurrence of the injury, and realizes the severity of the injury only when the injury has not improved for a long time or even continues to worsen, which has delayed the treatment time and increased the difficulty of the subsequent diagnosis and treatment process. SUMMARY

[0004] The purpose of the present application is to provide a wound condition data analysis system and method based on image recognition to solve the problems raised in the background.

[0005] In order to solve the above technical problems, the present application provides the following technical scheme: a wound condition data analysis system and method based on image recognition.

[0006] Step S100: obtaining the historical diagnosis and treatment records of doctors for external injury cases, and importing the image information of each external injury case into a first database, wherein the image information includes: an image of an injury part of the external injury and injury-causing object information corresponding to the external injury;

[0007] Step S200: performing image recognition on the part with the injury in the image in the first database, setting the injury area in the image as a first type of feature area, and setting other areas as a second type of feature area;

[0008] Step S300: calculating the divergence vector and image gradient vector of each color layer of the image in the first database, and calculating the vector difference between the color channel divergence vector and the gradient vector of each image;

[0009] Step S400: obtaining the injury-causing object corresponding to each image in the first database, classifying the images in the first database according to the types of the injury-causing objects, obtaining the trace features of the injuries caused by each type of injury-causing object, obtaining the actual area of the injury corresponding to the image from the diagnosis and treatment records, and establishing a corresponding relationship between the trace features and the actual area of the injury;

[0010] Step S500: Start the shooting device, acquire the target image of the injured area, divide the first type of feature region and the second type of feature region in the target image, compare and correct the second type of feature region in the target image with the preset reference image, and process the first type of feature region in the target image based on the comparison and correction result.

[0011] Step S600: The injury-causing object that causes the injury in the target image is set as the target injury-causing object. The target injury-causing object information is obtained, the trace features in the target image are identified, and the actual area corresponding to the first reference image is calculated based on the correspondence between the trace features and the actual area of ​​the injury.

[0012] Step S700: Calculate the image gradient vector of the first reference image, set the image gradient vector of the first reference image as the first reference vector, use the first reference vector as the first evaluation dimension, use the actual area corresponding to the first reference image as the second evaluation dimension, and evaluate the first reference image respectively.

[0013] Step S800: Obtain the first visit interval of the trauma case caused by the type of injury corresponding to the target injury agent from the historical medical records, adjust the first visit interval according to the area features in the first reference image, and obtain the observation interval of the next visit according to the changes in the first type of feature area between two adjacent visits.

[0014] Step S900: After the relevant interval, remind the relevant personnel to re-acquire images of the injured area, compare them with the previously acquired images, and mark the differences between the two images for the doctor's reference.

[0015] Furthermore, step S300 includes:

[0016] Step S301: Establish a planar coordinate system in the plane where the image is located, and separate the color channels of the image to obtain a red channel layer, a green channel layer and a blue channel layer. The planar coordinate systems corresponding to the red channel layer, the green channel layer and the blue channel layer are the same.

[0017] Step S302: Obtain the gradient vector corresponding to each color layer, use the gradient vector of each color layer as a component vector of the image gradient vector, and perform vector summation on the component vectors corresponding to all color layers to obtain the image gradient vector V. c V c =A c ·x+B c ·y, where x represents the unit vector in the first direction of the planar coordinate system, y represents the unit vector in the second direction of the planar coordinate system, A c B represents the coefficient of x.c Represents the coefficient of y;

[0018] The gradient vector represents the orientation of the change in the injury image. Based on image comparison, an injury image with the same orientation of change as the first image is found for comparison.

[0019] By calculating the separation of color channels, the separation degree of the image reflects the shape characteristics of the wound. Doctors need to make different diagnoses of injuries based on different wound shape characteristics. However, if users cannot accurately describe the shape characteristics of the wound, it will affect the doctor's judgment.

[0020] The combined vector represents the superposition of the gradient vectors of the three color channels. The angle between the individual vectors of each color channel and the image gradient vector indicates the mutual influence of colors. The greater the degree of separation, the more complex the injury, and the more necessary it is for doctors to observe.

[0021] Step S303: In the same image, obtain the angle between each component vector and the image gradient vector, and calculate the vector difference, where the vector difference of the i-th image is k. i , Where, θ j It represents the angle between the component vector corresponding to the j-th color channel layer and the image gradient vector, where a is a coefficient that satisfies the condition 0 < a < 0.5;

[0022] In a plane, the angle between two vectors ranges from 0° to 180°. The larger the angle, the greater the difference in the directions of the two vectors. In order to reflect the difference in the directions of the vectors, the larger the angle between the vectors, the greater the result of the separation degree calculation.

[0023] When 0 < a < 0.5, the sine function is monotonically increasing. The closer a is to 0.5, the greater the increase. The value of a should be selected according to the actual situation.

[0024] Furthermore, step S400 includes:

[0025] In a certain image category, the common image features of each image in the certain image category are recorded as the trace features of the certain image category, and the first corresponding group R1(INC, IMC) is obtained, where INC represents the type of a certain injury-causing substance, and IMC represents the trace features corresponding to the type of the certain injury-causing substance.

[0026] Step S402: Obtain image M from the image category corresponding to INC, and obtain image region M1 in M ​​through the first type of feature region recognition method, where M1 has a length of p1 pixels and a width of q1 pixels;

[0027] Step S403: Segment and save the trace feature IMC portion in M1 as M2, where M2 is p2 pixels long and q2 pixels wide;

[0028] Step S404: Obtain the actual area D of the first type of feature region in image M. real Calculate the actual area D corresponding to the trace feature IMC in M. M IMC ,in,

[0029] Step S405: In the image category corresponding to INC, calculate the trace area corresponding to the trace feature IMC of each image, and calculate the average trace area D. INC One image is selected from the image categories corresponding to INC as the trace feature reference image. The first type of feature region in the trace feature reference image is D. * Record the image region P corresponding to the trace features in the reference image. IMC , where D * The length is p d pixels, width q d Pixel P IMC Given a length of p0 pixels and a width of q0 pixels, calculate the feature ratio η: η = (p0 × q0) / (p d ×q d Get the second corresponding group R2(INC, P) IMC D INC ,η).

[0030] Furthermore, step S500 includes:

[0031] Step S501: Denote the first type of feature region in the target image as image Q1, and denote the second type of feature region in the target image as image Q2;

[0032] Step S502: Obtain the preset reference image Q * Q2 is preprocessed to achieve the same result as Q. * The same specifications, but the preprocessing method is recorded ω;

[0033] Step S503: Preprocess the image Q1 using the preprocessing method ω, and record the preprocessed image as the first reference image.

[0034] Furthermore, step S600 includes:

[0035] Step S601: Set the trace features corresponding to the target damaging object as target trace features. In the first reference image, identify the target trace features and save the image region corresponding to the target trace features as P. r , where P r Length pr pixels, width q r Pixel;

[0036] Step S602: Calculate the actual area D corresponding to the target trace features. R ,in Among them, D 1 INC P represents the average trace area corresponding to the target trace features in the first database. 1 INC This represents the image region in the reference image corresponding to the trace features of the target trace.

[0037] Step S603: Calculate the actual area D1 corresponding to the first reference image, where D1 = D R / η1, where η1 represents the feature ratio corresponding to the target trace feature.

[0038] Furthermore, step S700 includes:

[0039] Step S701: Let the first reference vector be V1, obtain the image whose image gradient vector is in the same direction as V1 from the first database, and import the image into the first comparison set;

[0040] Step S702: Obtain the actual area corresponding to each first-class feature region in the first comparison set, and obtain the maximum value D of the actual area. max The minimum value D of the actual area min Calculate the first evaluation value γ1, where,

[0041] The selection of γ1 is related to the size of the injured area and the direction of change of the injured image. The resultant vector represents the direction of change of the injured image. Under the condition of the same direction of change, the larger the area of ​​the injury, the greater the severity of the injury.

[0042] Step S703: Obtain the type of the target injury agent, extract the image corresponding to the type of the injury agent from the first database and import it into the second database, and obtain the image of the injury area with a size equal to D1 from the second database and import it into the second comparison set;

[0043] Step S704: Obtain the difference degree corresponding to each image in the second comparison set, and obtain the maximum value k of the vector difference degree. max The average value of the vector difference k mid The vector difference k1 between the first reference image and the second evaluation value γ2 is calculated, where, k1 represents the vector difference degree of the first reference image;

[0044] By separating and calculating color channels, the vector difference reflects the shape characteristics of the wound. Doctors need to make different diagnoses of injuries based on different wound shape characteristics. However, if users cannot accurately describe the shape characteristics of the wound, it will affect the doctor's judgment.

[0045] The image gradient vector represents the superposition of the gradient vectors of the three color channels. The angle between the gradient vector of each color channel and the image gradient vector indicates the mutual influence of colors. The greater the difference between the vectors, the more complex the injury is, and the more observation by the doctor is required.

[0046] Furthermore, step S800 includes:

[0047] Step S801: Obtain the first visit interval T from the historical medical records. D Calculate the adjustment value T1 for the first visit interval, where T1 = γ1 × T D ;

[0048] Step S802: Extract the medical records with the actual area of ​​the first type of feature region D1 from the medical records, calculate the average total recovery time t1 of the medical records, compare it with the average number of visits x1 of the medical records, and calculate the unit reference interval t. r = t1 / x1;

[0049] Step S803: Calculate the area change rate ρ of the injured area, ρ = (D x -D x+1 ) / D x , where D x D represents the actual area of ​​the first type of feature region obtained in the x-th acquisition. x+1 This represents the actual area of ​​the first type of feature region obtained in the (x+1)th iteration;

[0050] Step S804: Calculate the adjusted adjustment unit value T a T a =γ2×ρ1×t r , where ρ1 represents the rate of change of the area of ​​the currently acquired injury area compared to the area of ​​the previous injury area;

[0051] Furthermore, step S900 includes:

[0052] Step S901: When the relevant person is seeking medical treatment for the first time, the adjusted value of the first visit interval is used as the observation interval. When the relevant person seeks medical treatment multiple times for the same injury, the next observation interval T is calculated. b , among which, T b =T0+n×T aWhere n represents the manually entered adjustment parameter, and T0 represents the observation interval before this visit. After the next observation interval, the relevant personnel will be reminded to visit the doctor.

[0053] When D x >D x+1 When the injury improves as expected, or even better than expected, the output T indicates a change in condition. b The value will increase, and the patient will not need to be diagnosed by a doctor so frequently. x <D x+1 When this occurs, it indicates that the injury has worsened; at this time, D... x -D x+1 <0, the output T b The value will decrease, and the patient's condition will require another doctor's diagnosis in the short term;

[0054] Step S902: For the image acquired again, extract the second reference image corresponding to the first reference image using the method in step S500, and extract the second reference vector V2 corresponding to the second reference image;

[0055] Step S903: Calculate the image difference H between the first reference image and the second reference image, where H = cos<V1,V2> cos<V1,V2> Represents the cosine values ​​of vectors V1 and V2;

[0056] Step S904: The doctor examines the second reference image, marks the areas in the second reference image that are different from the first reference image, and provides the doctor with the degree of difference between the first reference image and the second reference image for reference. The doctor makes a judgment on the injury of the target user based on the first reference image and the second reference image.

[0057] To better implement the above methods, an image recognition-based injury data analysis system is also proposed, which includes:

[0058] The system comprises the following modules: historical record management module, injury area identification module, gradient vector calculation module, trace feature management module, image acquisition module, area calculation module, evaluation module, observation interval calculation module, and information reminder module. Specifically, the historical record management module manages historical medical records for trauma cases; the injury area identification module identifies first-class and second-class feature regions in images; the gradient vector calculation module calculates the color layer sub-vectors and image gradient vectors for each image; the trace feature management module establishes the correspondence between trace features and the actual area of ​​the injury; the image acquisition module acquires current image information; the area calculation module calculates the actual area corresponding to the first reference image; the evaluation module evaluates the current image; the observation interval calculation module calculates the observation interval for the next visit; and the information reminder module sends notifications to relevant personnel.

[0059] Compared with the prior art, the beneficial effects achieved by the present invention are as follows: The present invention acquires injury images, extracts the area features and color change features of the injury images, compares them in a database, and acquires the causative agent that caused the injury. It calculates the time of acquiring the second injury image based on the image features of the injury area and the features of the causative agent, compares the two injury images, and marks the changes in the injury over a period of time. This assists doctors in making remote judgments about the injury, simplifies the patient consultation process, and improves the efficiency of doctors in handling injuries. Attached Figure Description

[0060] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0061] Figure 1 This is a schematic diagram of the structure of an injury data analysis system based on image recognition, as per this invention patent.

[0062] Figure 2 This is a schematic diagram of the first part of an image recognition-based injury data analysis method according to the present invention patent;

[0063] Figure 3 This is a schematic diagram of the second part of an image recognition-based injury data analysis method according to the present invention patent;

[0064] Figure 4 This is a schematic diagram of image comparison for an injury data analysis method based on image recognition, as described in this invention patent. Detailed Implementation

[0065] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0066] Please see Figure 1 , Figure 2 , Figure 3 and Figure 4 The present invention provides the following technical solution:

[0067] Step S100: Obtain the doctor's historical diagnosis and treatment records of trauma cases, and import the image information of each trauma case into the first database. The image information includes: images of the injured parts and information on the causative agent corresponding to the trauma.

[0068] Step S200: In the images in the first database, perform image recognition on the parts with injuries, set the injured areas in the images as the first type of feature areas, and set the other areas as the second type of feature areas;

[0069] Step S300: Calculate the color layer vectors and image gradient vectors of each image in the first database, and calculate the vector difference between the color channel vectors and gradient vectors of each image.

[0070] Step S300 includes:

[0071] Step S301: Establish a planar coordinate system in the plane where the image is located, and separate the color channels of the image to obtain a red channel layer, a green channel layer and a blue channel layer. The planar coordinate systems corresponding to the red channel layer, the green channel layer and the blue channel layer are the same.

[0072] Step S302: Obtain the gradient vector corresponding to each color layer, use the gradient vector of each color layer as a component vector of the image gradient vector, and perform vector summation on the component vectors corresponding to all color layers to obtain the image gradient vector V. c V c =A c ·x+B c ·y, where x represents the unit vector in the first direction of the planar coordinate system, y represents the unit vector in the second direction of the planar coordinate system, A c B represents the coefficient of x. c Represents the coefficient of y;

[0073] In this embodiment, the extreme points in each channel layer are obtained, and the extreme points are substituted into the gradient calculation formula to calculate the gradient vector of each layer.

[0074] Where A = α r +α g +α b B = β r +β g +β b α r α represents the coefficient of the x-component in the red channel layer. g α represents the coefficient of the x-component in the green channel layer. b β represents the coefficient of the x-component in the blue channel layer. r β represents the coefficient of the y-component in the red channel layer. g β represents the coefficient of the y-component in the green channel layer. b Represents the coefficients of the y-component in the blue channel layer;

[0075] Step S303: In the same image, obtain the angle between each component vector and the image gradient vector, and calculate the vector difference, where the vector difference of the i-th image is k. i , Where, θ j Let represent the angle between the component vector corresponding to the j-th color channel layer and the image gradient vector, where 'a' is a coefficient satisfying the condition 0 < a < 0.5.

[0076] Step S400: Obtain the injury-causing agent corresponding to each image in the first database, classify the images in the first database according to the type of injury-causing agent, obtain the trace features of injuries caused by various types of injury-causing agents, obtain the actual area of ​​the injury corresponding to the image from the medical records, and establish the correspondence between the trace features and the actual area of ​​the injury.

[0077] In practice, the injury can be classified by acquiring an image of the injury-causing object, performing image recognition on the object, or by having the user add a description of the injury-causing object, such as its name or size, when uploading a skin image of the injured area; or by pre-establishing a data model of the injury-causing object, which the user can then select from when uploading a skin image of the injured area.

[0078] Step S400 includes:

[0079] Step S401: In a certain image category, the common image features of each image in the certain image category are recorded as the trace features of the certain image category, and the first corresponding group R1(INC, IMC) is obtained, where INC represents the type of a certain injury agent, and IMC represents the trace features corresponding to the type of the certain injury agent.

[0080] Step S402: Obtain image M from the image category corresponding to INC, and obtain image region M1 in M ​​through the first type of feature region recognition method, where M1 has a length of p1 pixels and a width of q1 pixels;

[0081] Step S403: Segment and save the trace feature IMC portion in M1 as M2, where M2 is p2 pixels long and q2 pixels wide;

[0082] Step S404: Obtain the actual area D of the first type of feature region in image M. real Calculate the actual area D corresponding to the trace feature IMC in M. M IMC ,in,

[0083] Step S405: In the image category corresponding to INC, calculate the trace area corresponding to the trace feature IMC of each image, and calculate the average trace area D. INC One image is selected from the image categories corresponding to INC as the trace feature reference image. The first type of feature region in the trace feature reference image is D. * Record the image region P corresponding to the trace features in the reference image. IMC , where D * The length is p d pixels, width q d Pixel P IMC Given a length of p0 pixels and a width of q0 pixels, calculate the feature ratio η: η = (p0 × q0) / (p d ×q d Get the second corresponding group R2(INC, P) IMC D INC ,η).

[0084] Step S500: Start the shooting device, acquire the target image of the injured area, divide the first type of feature region and the second type of feature region in the target image, compare and correct the second type of feature region in the target image with the preset reference image, and process the first type of feature region in the target image based on the comparison and correction result.

[0085] During implementation, remote data transmission is carried out via the Internet, which consists of user terminals, servers, and doctor terminals. After the user terminal takes an image, the image information is uploaded to the server via the Internet for data analysis and comparison. The processed image is then sent to the doctor terminal.

[0086] After making a preliminary diagnosis, the relevant doctors upload the diagnosis results to the server, and the server forwards the diagnosis results to the corresponding user terminals;

[0087] Users need to register on the user terminal. In this embodiment, it is necessary to obtain the image information of the patient's skin in a normal state, that is, when there is no injury. The skin information in a normal state is bound to the user's identity information. For example, after the injury, if the patient is unable to take a picture himself, people around the patient can also obtain the image of the patient's skin through the user terminal. After the patient's identity is verified, the image of the patient's normal skin is compared with the image of the skin after the injury.

[0088] Step S500 includes:

[0089] Step S501: Denote the first type of feature region in the target image as image Q1, and denote the second type of feature region in the target image as image Q2;

[0090] Step S502: Obtain the preset reference image Q * Q2 is preprocessed to achieve the same result as Q. * The same specifications, but the preprocessing method is recorded ω;

[0091] Step S503: Preprocess the image Q1 using the preprocessing method ω, and record the preprocessed image as the first reference image.

[0092] Step S600: The injury-causing object that causes the injury in the target image is set as the target injury-causing object. The target injury-causing object information is obtained, the trace features in the target image are identified, and the actual area corresponding to the first reference image is calculated based on the correspondence between the trace features and the actual area of ​​the injury.

[0093] Step S600 includes:

[0094] Step S601: Set the trace features corresponding to the target damaging object as target trace features. In the first reference image, identify the target trace features and save the image region corresponding to the target trace features as P. r , where P r Length p r pixels, width q r Pixel;

[0095] Step S602: Calculate the actual area D corresponding to the target trace features. R ,in Among them, D 1 INC P represents the average trace area corresponding to the target trace features in the first database. 1 INC This represents the image region in the reference image corresponding to the trace features of the target trace.

[0096] Step S603: Calculate the actual area D1 corresponding to the first reference image, where D1 = D R / η1, where η1 represents the feature ratio corresponding to the target trace feature.

[0097] Step S700: Calculate the image gradient vector of the first reference image, set the image gradient vector of the first reference image as the first reference vector, use the first reference vector as the first evaluation dimension, use the actual area corresponding to the first reference image as the second evaluation dimension, and evaluate the first reference image respectively.

[0098] Step S700 includes:

[0099] Step S701: Let the first reference vector be V1, obtain the image whose image gradient vector is in the same direction as V1 from the first database, and import the image into the first comparison set;

[0100] Step S702: Obtain the actual area corresponding to each first-class feature region in the first comparison set, and obtain the maximum value D of the actual area. max The minimum value D of the actual area min Calculate the first evaluation value γ1, where,

[0101] Step S703: Obtain the type of the target injury agent, extract the image corresponding to the type of the injury agent from the first database and import it into the second database, and obtain the image of the injury area with a size equal to D1 from the second database and import it into the second comparison set;

[0102] Step S704: Obtain the difference degree corresponding to each image in the second comparison set, and obtain the maximum value k of the vector difference degree. max The average value of the vector difference k mid The vector difference k1 between the first reference image and the second evaluation value γ2 is calculated, where, k1 represents the vector difference of the first reference image.

[0103] Step S800: Obtain the first visit interval of the trauma case caused by the type of injury corresponding to the target injury agent from the historical medical records, adjust the first visit interval according to the area features in the first reference image, and obtain the observation interval of the next visit according to the changes in the first type of feature area between two adjacent visits.

[0104] Step S800 includes:

[0105] Step S801: Obtain the first visit interval T from the historical medical records. D Calculate the adjustment value T1 for the first visit interval, where T1 = γ1 × TD ;

[0106] Step S802: Extract the medical records with the actual area of ​​the first type of feature region D1 from the medical records, calculate the average total recovery time t1 of the medical records, compare it with the average number of visits x1 of the medical records, and calculate the unit reference interval t. r = t1 / x1;

[0107] Step S803: Calculate the area change rate ρ of the injured area, ρ = (D x -D x+1 ) / D x , where D x D represents the actual area of ​​the first type of feature region obtained in the x-th acquisition. x+1 This represents the actual area of ​​the first type of feature region obtained in the (x+1)th iteration;

[0108] Step S804: Calculate the adjusted adjustment unit value T a T a =γ2×ρ1×t r , where ρ1 represents the rate of change of the area of ​​the currently acquired injury area compared to the area of ​​the previous injury area.

[0109] Step S900: After the relevant interval, remind the relevant personnel to re-acquire images of the injured area, compare them with the previously acquired images, and mark the differences between the two images for the doctor's reference;

[0110] Step S900 includes:

[0111] Step S901: When the relevant person is seeking medical treatment for the first time, the adjusted value of the first visit interval is used as the observation interval. When the relevant person seeks medical treatment multiple times for the same injury, the next observation interval T is calculated. b , among which, T b =T0+n×T a Where n represents the manually entered adjustment parameter, and T0 represents the observation interval before this visit. After the next observation interval, the relevant personnel will be reminded to visit the doctor.

[0112] Step S902: For the image acquired again, extract the second reference image corresponding to the first reference image using the method in step S500, and extract the second reference vector V2 corresponding to the second reference image;

[0113] Step S903: Calculate the image difference H between the first reference image and the second reference image, where H = cos<V1,V2> cos<V1,V2> Represents the cosine values ​​of vectors V1 and V2;

[0114] Figure 4 This represents a comparison diagram of a first reference image and a second reference image, where D1 represents the first reference image, D2 represents the second reference image, V1 represents the first reference vector, and V2 represents the second reference vector.

[0115] Step S904: The doctor examines the second reference image, marks the areas in the second reference image that are different from the first reference image, and provides the doctor with the degree of difference between the first reference image and the second reference image for reference. The doctor makes a judgment on the injury of the target user based on the first reference image and the second reference image.

[0116] The system includes:

[0117] Image management module, image correction module, first comparison module, second comparison module, and information reminder module;

[0118] The image management module is used to manage and preprocess injury images;

[0119] The image correction module is used to correct images including injuries. The image correction module includes an image acquisition unit, an image recognition unit, an image comparison unit, and a correction application unit. The image acquisition unit is used to acquire images, the image recognition unit is used to identify the injury area in the image, the image comparison unit is used to compare the currently acquired image with images of normal skin in historical data records to obtain an image correction method, and the correction application unit is used to apply the correction method to the currently acquired image to obtain a first reference image.

[0120] The first comparison module is used to compare the acquired image with images in the database. The first comparison module includes: a channel separation unit, a first reference vector acquisition unit, a first evaluation value calculation unit, an image feature comparison unit, a vector difference calculation unit, and a second evaluation value calculation unit. The channel separation unit is used to separate the color channels of the image, the first reference vector acquisition unit is used to acquire a first reference vector, the first evaluation value calculation unit is used to calculate a first evaluation value, the image feature comparison unit is used to compare the injury features caused by the injuring object, the vector difference calculation unit is used to calculate the color separation degree of the image, and the second evaluation value calculation unit is used to calculate a second evaluation value.

[0121] The second comparison module is used to compare the two acquired images. The second comparison module includes: a time interval calculation unit, a second reference image acquisition unit, and an image difference calculation unit. The time interval calculation unit is used to calculate the observation time using a first evaluation value and a second evaluation value. The second reference image acquisition unit is used to acquire a second reference image. The image difference calculation unit is used to calculate the image difference.

[0122] The information reminder module is used to send information reminders to target users and doctors.

[0123] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0124] Finally, it should be noted that the above descriptions are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for injury data analysis based on image recognition, characterized in that, The method includes the following steps: Step S100: Obtain the doctor's historical diagnosis and treatment records of trauma cases, and import the image information of each trauma case into the first database. The image information includes: images of the injured parts and information on the causative agent corresponding to the trauma. Step S200: In the images in the first database, perform image recognition on the parts with injuries, set the injured areas in the images as the first type of feature areas, and set the other areas as the second type of feature areas; Step S300: Calculate the color layer vectors and image gradient vectors of each image in the first database, and calculate the vector difference between the color channel vectors and gradient vectors of each image. Step S400: Obtain the injury-causing agent corresponding to each image in the first database, classify the images in the first database according to the type of injury-causing agent, obtain the trace features of injuries caused by various types of injury-causing agents, obtain the actual area of ​​the injury corresponding to the image from the medical records, and establish the correspondence between the trace features and the actual area of ​​the injury. Step S500: Start the shooting device, acquire the target image of the injured area, divide the first type of feature region and the second type of feature region in the target image, compare and correct the second type of feature region in the target image with the preset reference image, and process the first type of feature region in the target image based on the comparison and correction result. Step S600: The injury-causing object that causes the injury in the target image is set as the target injury-causing object. The target injury-causing object information is obtained, the trace features in the target image are identified, and the actual area corresponding to the first reference image is calculated based on the correspondence between the trace features and the actual area of ​​the injury. Step S700: Calculate the image gradient vector of the first reference image, set the image gradient vector of the first reference image as the first reference vector, use the first reference vector as the first evaluation dimension, use the actual area corresponding to the first reference image as the second evaluation dimension, and evaluate the first reference image respectively. Step S700 includes: Step S701: Let the first reference vector be V1, obtain the image whose image gradient vector is in the same direction as V1 from the first database, and import the image into the first comparison set; Step S702: Obtain the actual area corresponding to each first-class feature region in the first comparison set, and obtain the maximum value D of the actual area. max The minimum value D of the actual area min Calculate the first evaluation value γ1, where, ; Step S703: Obtain the type of the target injury agent, extract the image corresponding to the type of the injury agent from the first database and import it into the second database, and obtain the image of the injury area with a size equal to D1 from the second database and import it into the second comparison set; Step S704: Obtain the difference degree corresponding to each image in the second comparison set, and obtain the maximum value k of the vector difference degree. max The average value of the vector difference k mid The vector difference k1 between the first reference image and the second evaluation value γ2 is calculated, where, k1 represents the vector difference degree of the first reference image; Step S800: Obtain the first visit interval of the trauma case caused by the type of injury corresponding to the target injury agent from the historical medical records, adjust the first visit interval according to the area features in the first reference image, and obtain the observation interval of the next visit according to the changes in the first type of feature area between two adjacent visits. Step S900: After the relevant interval, remind the relevant personnel to re-acquire images of the injured area, compare them with the previously acquired images, and mark the differences between the two images for the doctor's reference.

2. The injury data analysis method based on image recognition according to claim 1, characterized in that: Step S300 includes: Step S301: Establish a planar coordinate system in the plane where the image is located, and separate the color channels of the image to obtain a red channel layer, a green channel layer and a blue channel layer. The planar coordinate systems corresponding to the red channel layer, the green channel layer and the blue channel layer are the same. Step S302: Obtain the gradient vector corresponding to each color layer, use the gradient vector of each color layer as a component vector of the image gradient vector, and perform vector summation on the component vectors corresponding to all color layers to obtain the image gradient vector V. c V c =A c ·x+B c ·y, where x represents the unit vector in the first direction of the planar coordinate system, y represents the unit vector in the second direction of the planar coordinate system, A c B represents the coefficient of x. c Represents the coefficient of y; Step S303: In the same image, obtain the angle between each component vector and the image gradient vector, and calculate the vector difference, where the vector difference of the i-th image is k. i , , where θ j Let represent the angle between the component vector corresponding to the j-th color channel layer and the image gradient vector, where 'a' is a coefficient satisfying the condition 0 < a < 0.

5.

3. The injury data analysis method based on image recognition according to claim 2, characterized in that: Step S400 includes: Step S401: In a certain image category, the common image features of each image in the certain image category are recorded as the trace features of the certain image category, and the first corresponding group R1 (INC, IMC) is obtained, where INC represents the type of a certain injury agent, and IMC represents the trace features corresponding to the type of the certain injury agent. Step S402: Obtain image M from the image category corresponding to INC, and obtain image region M1 in M ​​through the first type of feature region recognition method, where M1 has a length of p1 pixels and a width of q1 pixels; Step S403: Segment and save the trace feature IMC portion in M1 as M2, where M2 is p2 pixels long and q2 pixels wide; Step S404: Obtain the actual area D of the first type of feature region in image M. real Calculate the actual area D corresponding to the trace feature IMC in M. M IMC ,in, ; Step S405: In the image category corresponding to INC, calculate the trace area corresponding to the trace feature IMC of each image, and calculate the average trace area D. INC One image is selected from the image categories corresponding to INC as the trace feature reference image. The first type of feature region in the trace feature reference image is D. * Record the image region P corresponding to the trace features in the reference image. IMC , where D * The length is p d pixels, width q d Pixel P IMC Given a length of p0 pixels and a width of q0 pixels, calculate the feature ratio η: η = (p0 × q0) / (p d ×q d Obtain the second corresponding group R2(INC, P) IMC D INC ,η).

4. The injury data analysis method based on image recognition according to claim 3, characterized in that: Step S500 includes: Step S501: Denote the first type of feature region in the target image as image Q1, and denote the second type of feature region in the target image as image Q2; Step S502: Obtain the preset reference image Q * Q2 is preprocessed to achieve the same result as Q. * The same specifications, but the preprocessing method is recorded ω; Step S503: Preprocess the image Q1 using the preprocessing method ω, and record the preprocessed image as the first reference image.

5. The injury data analysis method based on image recognition according to claim 4, characterized in that: Step S600 includes: Step S601: Set the trace features corresponding to the target damaging object as target trace features. In the first reference image, identify the target trace features and save the image region corresponding to the target trace features as P. r , where P r Length p r pixels, width q r Pixel; Step S602: Calculate the actual area D corresponding to the target trace features. R ,in , where D 1 INC P represents the average trace area corresponding to the target trace features in the first database. 1 INC This represents the image region in the reference image corresponding to the trace features of the target trace. Step S603: Calculate the actual area D1 corresponding to the first reference image, where D1 = D R / η1, where η1 represents the feature ratio corresponding to the target trace feature.

6. The injury data analysis method based on image recognition according to claim 5, characterized in that: Step S800 includes: Step S801: Obtain the first visit interval T from the historical medical records. D Calculate the adjustment value T1 for the first visit interval, where T1 = γ1 × T D ; Step S802: Extract the medical records with the actual area of ​​the first type of feature region D1 from the medical records, calculate the average total recovery time t1 of the medical records, compare it with the average number of visits x1 of the medical records, and calculate the unit reference interval t. r =t1 / x1; Step S803: Calculate the rate of change of the injured area ρ, ρ = (D x -D x+1 ) / D x , where D x D represents the actual area of ​​the first type of feature region obtained in the x-th acquisition. x+1 This represents the actual area of ​​the first type of feature region obtained in the (x+1)th iteration; Step S804: Calculate the adjusted adjustment unit value T a T a =γ2×ρ1×t r , where ρ1 represents the rate of change of the area of ​​the currently acquired injury area compared to the area of ​​the previous injury area.

7. The injury data analysis method based on image recognition according to claim 6, characterized in that: Step S900 includes: Step S901: When the relevant person is seeking medical treatment for the first time, the adjusted value of the first visit interval is used as the observation interval. When the relevant person seeks medical treatment multiple times for the same injury, the next observation interval T is calculated. b , among which, T b =T0+n×T a Where n represents the manually input adjustment parameter, and T0 represents the observation interval before this visit. After the next observation interval, the relevant personnel will be reminded to visit the doctor. Step S902: For the image acquired again, extract the second reference image corresponding to the first reference image using the method in step S500, and extract the second reference vector V2 corresponding to the second reference image; Step S903: Calculate the image difference H between the first reference image and the second reference image, where H = cos<V1,V2> cos<V1,V2> Represents the cosine values ​​of vectors V1 and V2; Step S904: The doctor examines the second reference image, marks the areas in the second reference image that are different from the first reference image, and provides the doctor with the degree of difference between the first reference image and the second reference image for reference. The doctor makes a judgment on the injury of the target user based on the first reference image and the second reference image.

8. A injury data analysis system for performing the injury data analysis method based on image recognition as described in any one of claims 1-7, characterized in that: The system includes the following modules: a history record management module, a wound area identification module, a gradient vector calculation module, a trace feature management module, an image acquisition module, an area calculation module, an evaluation module, an observation interval calculation module, and an information reminder module. Specifically, the history record management module manages the historical medical records of trauma cases; the wound area identification module identifies wound areas in images as first-class and second-class feature regions; the gradient vector calculation module calculates the color layer sub-vectors and image gradient vectors for each image; the trace feature management module establishes the correspondence between trace features and the actual area of ​​the injury; the image acquisition module acquires current image information; the area calculation module calculates the actual area corresponding to the first reference image; the evaluation module evaluates the current image; the observation interval calculation module calculates the observation interval for the next visit; and the information reminder module sends message reminders to relevant personnel.

Images