Pet health monitoring method and device

Abstract

The invention provides a pet health monitoring method and device, and is suitable for the technical field of image processing, and the method comprises the steps: carrying out the feature extraction of pet image information based on a plurality of pet image feature extraction matrixes, and obtaining initial pet image feature information; performing decomposition transformation processing on the initial pet image feature information to obtain target pet image feature information; and generating multiple pieces of pet health monitoring information according to the target pet image feature information and the multi-source pet clinical case information. Multi-dimensional pet image features such as expressions, oral cavities and skin are accurately extracted and recognized, health monitoring requirements of various pets are rapidly met, and the accuracy and reliability of pet health monitoring are improved.

Description

Technical Field

[0001] The present application belongs to the field of image processing technology, and in particular to a pet health monitoring method and device. Background Art

[0002] With the continuous growth in the number of pets and the increasing attention paid by pet owners to their pets' health, the pet medical industry has ushered in a rapid development momentum.

[0003] Existing technologies often focus on using single-modality data or relatively simple algorithmic models to diagnose pet diseases. For example, by building a machine learning model based on a large amount of case data, the symptom text described by pet owners is analyzed to provide possible disease diagnosis results; or by using convolutional neural networks in deep learning to process pet medical images (such as X-rays and ultrasound images) to identify abnormal areas and assist in diagnosis.

[0004] However, existing technologies rely solely on text descriptions to obtain comprehensive information about a pet's condition. Identifying a pet's condition based solely on medical images is not suitable for daily household pet diagnosis, and it is difficult to effectively monitor early symptoms. Therefore, the accuracy and reliability of pet health monitoring cannot be guaranteed. Summary of the Invention

[0005] In view of this, the embodiments of the present application provide a pet health monitoring method and device, which aims to solve the problems in the prior art that pet health monitoring is prone to misdiagnosis and missed diagnosis, and the accuracy and effectiveness of monitoring the health status of pets are low.

[0006] A first aspect of an embodiment of the present application provides a pet health monitoring method, comprising:

[0007] Acquire multiple pet image information and multi-source pet clinical case information;

[0008] Based on a plurality of preset pet image feature extraction matrices, performing feature extraction processing on the plurality of pet image information to obtain a plurality of initial pet image feature information;

[0009] Decomposing and transforming the plurality of initial pet image feature information to obtain a plurality of target pet image feature information;

[0010] A plurality of pet health monitoring information is generated based on the plurality of target pet image feature information and multi-source pet clinical case information.

[0011] A second aspect of an embodiment of the present application provides a pet health monitoring device, comprising:

[0012] An information acquisition module, used to acquire multiple pet image information and multi-source pet clinical case information;

[0013] an initial pet image feature information generating module, configured to perform feature extraction processing on the plurality of pet image information based on a plurality of preset pet image feature extraction matrices to obtain a plurality of initial pet image feature information;

[0014] a target pet image feature information generating module, configured to perform decomposition and transformation processing on the plurality of initial pet image feature information to obtain a plurality of target pet image feature information;

[0015] The pet health monitoring information generation module is used to generate multiple pet health monitoring information based on the multiple target pet image feature information and multi-source pet clinical case information.

[0016] The third aspect of an embodiment of the present application provides a terminal device, which includes a memory and a processor, wherein the memory stores a computer program that can be run on the processor, and when the processor executes the computer program, it implements the steps of the pet health monitoring method described in the first aspect above.

[0017] Compared with the prior art, the beneficial effects of the embodiments of the present application are as follows: the present application builds a cross-modal pet consultation foundation, avoids the limitations of single-modal data for pet health monitoring, utilizes multiple pet image feature extraction matrices to extract semantic features of pet image information, and further separates global features and local features by decomposing the extracted pet image semantic features, facilitating in-depth correlation analysis and processing with multi-source pet clinical case information to realize monitoring of pet health status, thereby improving the accuracy and effectiveness of identifying pet disease risks and health trends. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following briefly introduces the drawings required for use in the embodiments or descriptions of the prior art. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0019] Figure 1 This is a schematic diagram of the implementation process of the pet health monitoring method provided in Example 1 of the present application;

[0020] Figure 2 This is a schematic diagram of the implementation process of the pet health monitoring method provided in Example 2 of the present application;

[0021] Figure 3This is a schematic diagram of the implementation process of the pet health monitoring method provided in Example 3 of the present application;

[0022] Figure 4 This is a schematic diagram of the implementation process of the pet health monitoring method provided in Example 4 of the present application;

[0023] Figure 5 This is a schematic diagram of the implementation process of the pet health monitoring method provided in Example 5 of the present application;

[0024] Figure 6 This is a schematic diagram of the implementation process of the pet health monitoring method provided in Example 6 of the present application;

[0025] Figure 7 This is a schematic diagram of the implementation process of the pet health monitoring method provided in Example 7 of the present application;

[0026] Figure 8 This is a schematic diagram of the implementation process of the pet health monitoring method provided in Example 8 of the present application;

[0027] Figure 9 is a structural diagram of a pet health monitoring device provided in an embodiment of the present application;

[0028] Figure 10 It is a schematic diagram of the terminal device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0029] In the following description, specific details such as specific system structures and techniques are provided for purposes of illustration rather than limitation to facilitate a thorough understanding of the embodiments of the present application. However, it will be apparent to those skilled in the art that the present application may be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to avoid obscuring the description of the present application with unnecessary detail.

[0030] In order to illustrate the technical solution described in this application, specific embodiments are provided below.

[0031] Figure 1 The following is a flowchart of the implementation of the pet health monitoring method provided in Example 1 of the present application, which is detailed as follows:

[0032] Step S101: Acquire multiple pet image information and multi-source pet clinical case information.

[0033] In this embodiment, pet image information may refer to visual data of various parts of a pet's body captured by a photographic device. This data may be captured statically under natural light using a mobile phone camera or high-definition camera. This data may include full-body images, facial images, skin images, oral images, ear canal images, feces images, and vomitus images of the pet. This facilitates subsequent facial expression analysis, breed identification, oral assessment, feces, vomitus, skin examinations, ear canal examinations, and eye examinations, enabling health monitoring of pets in the home. Multi-source pet clinical case information may refer to structured and unstructured data related to a pet's health status, encompassing multiple dimensions such as medical history, physiological indicators, and diagnostic records. Specifically, multi-source pet clinical case information may include basic attribute data such as breed, age, weight, gender, and allergy history; as well as medical history and symptom data, such as the pet's onset time, symptom descriptions, medication records, and past medical history, as well as text or voice information. Multi-source pet clinical case information can be obtained by accessing a pet hospital's case database.

[0034] Step S102 : performing feature extraction processing on the plurality of pet image information based on a plurality of preset pet image feature extraction matrices to obtain a plurality of initial pet image feature information.

[0035] In this embodiment, the preset pet image feature extraction matrix can optionally be manually set and can be multiple matrices of the same dimension or multiple matrices of different dimensions, and is used to extract features from pet image information. The pet image information can be first converted into a matrix form using the PIL library, OpenCV library, Matplotlib library, or NumPy library in Python to generate pet image matrix information. Then, multiple preset pet image feature extraction matrices can be convolved with the pet image matrix information. Then, the convolution results can be weighted summed based on preset weights, and the weighted summed result is used as the initial pet image feature information.

[0036] In this embodiment, optionally, the pet image information can be standardized first, and the standardization process can include grayscale conversion, histogram equalization, and Gaussian filtering. Then, the pet contour in the pet image information is identified based on a preset pet contour mask, and then the pet oral image information, pet skin image information, pet hair image information, pet eye image information, pet ear image information, pet excrement image information and pet vomit image information are extracted respectively. Then, the pet oral image information, pet skin image information, pet hair image information, pet eye image information, pet ear image information, pet excrement image information and pet vomit image information are convolved with multiple preset pet image feature extraction matrices respectively, and then the results after convolution of each pet image feature extraction matrix are output respectively, and the separately output results are spliced, so that the spliced ​​results are used as the initial pet image feature information.

[0037] In this embodiment, optionally, multiple preset pet image feature extraction matrices can be manually set, and the pet image feature extraction matrix can include a preset pet image edge feature matrix, a preset pet image texture feature matrix, and a preset pet image scale feature matrix, wherein the pet image edge feature matrix is ​​used to perform convolution operations on areas such as the mouth and eyes, extract contour edge information, and generate an edge response map; the pet image texture feature matrix is ​​used to calculate local pixel grayscale difference patterns for skin images, and quantify microscopic texture features such as dandruff and erythema; the pet image scale feature matrix is ​​used to perform multi-level decomposition on excrement images, separate the solid morphological features and bloodshot texture features therein, and then standardize the feature maps output by each pet image feature extraction matrix and splice them, and use the spliced ​​feature maps as the initial pet image feature information.

[0038] Step S103 : performing decomposition and transformation processing on the multiple pieces of initial pet image feature information to obtain multiple pieces of target pet image feature information.

[0039] In this embodiment, it is understandable that the initial pet image feature information contains a large number of detail features and therefore has a lot of repeated or minor or noise information. Therefore, it is necessary to separate the key information from the redundant information and the noise information in the initial pet image feature information to filter out the redundant information and the noise information in the pet image feature information. A set of basic pet image feature units can be pre-set. It is understandable that a set of basic pet image feature units includes multiple basic pet image feature sub-units. The initial pet image feature information can be decomposed into a linear combination for representing the basic pet image feature unit, thereby generating multiple groups of basic pet image feature sub-unit weight information, and then performing weighting on multiple basic pet image feature sub-units. The basic pet image feature subunits are weighted summed, and then the basic pet image feature subunits and weight information combination whose weighted summation result is closest to the initial pet image feature information are determined, and then the weight values ​​in the weight information combination are arranged from large to small, the weight values ​​of the first half are increased, the weight values ​​of the first half of the second half are reduced, and the weight values ​​of the second half of the second half are adjusted to 0, so as to highlight the effective information in the pet image feature information, suppress and filter out the redundant information and noise information in the pet image feature information, and then the basic pet image feature subunits are weighted summed through the adjusted weight values, and the result of the weighted summation is used as the target pet image feature information.

[0040] Step S104 : generating a plurality of pet health monitoring information based on the plurality of target pet image feature information and multi-source pet clinical case information.

[0041] In this embodiment, the marked pet image feature information and the multi-source pet clinical case information may be aligned first, the pet breed information in the multi-source pet clinical case information may be uniquely encoded, the text information and numerical data in the multi-source pet clinical case information may be standardized, and then the uniquely encoded pet breed information and the standardized text information and numerical data may be mapped to the high-dimensional space by the Sigmoid function, so that the pet breed information and the standardized text information and numerical data may be mapped to the high-dimensional space, and then the diseases, symptoms, and other related information may be mapped to the high-dimensional space. , breeds as nodes, "disease-symptom association" and "breed-susceptibility association" as edges, target pet image feature information and multi-source pet clinical case information as node attributes respectively, by calculating the similarity between node attributes and nodes and edges, and artificially presetting a similarity threshold, nodes and edges with similarity greater than the similarity threshold are extracted, and then the calculated similarity is used as the weight value, and the extracted nodes and edges are weighted summed, and then the result of the weighted sum is used as the probability of the pet in the image information suffering from various diseases. It can be understood that the generated multiple pet health monitoring information may include probability values ​​of pets suffering from multiple diseases.

[0042] The pet health monitoring method provided in the embodiment of the present application builds a cross-modal pet consultation foundation, avoids the limitations of single-modal data for pet health monitoring, utilizes multiple pet image feature extraction matrices to extract semantic features of pet image information, and further separates global features and local features by decomposing the extracted pet image semantic features, facilitating in-depth correlation analysis and processing with multi-source pet clinical case information to realize monitoring of the pet's health status, thereby improving the accuracy and effectiveness of identifying pet disease risks and health trends.

[0043] Figure 2 The flowchart of the implementation of the pet health monitoring method provided in the second embodiment of the present application is shown. The difference between the second embodiment and the first embodiment is that the step S102 specifically includes:

[0044] Step S201 : Based on preset pet image block size information, a plurality of pet image information is subjected to block processing to obtain a plurality of pet image block information.

[0045] In this embodiment, the preset pet image block size information may be manually set and may include pet image block length information and pet image block width information, where both the pet image block length information and the pet image block width information may be expressed in pixels. The pet image information may be segmented based on the preset pet image block size information, and the segmented pet sub-image information may be used as the pet image block information.

[0046] Step S202 : extracting pixel information of a plurality of pet image block information to obtain a plurality of pet image block pixel information and pet image block pixel quantity information.

[0047] In this embodiment, it can be understood that after the pet image information is divided into blocks, each pet image block information has multiple pixel information, and the pixel information in each pet image block information is extracted and processed to obtain the content information and coordinate information of each pixel as the pet image block pixel information, and the total number of pixels in a single pet image block information is obtained as the pet image block pixel number information.

[0048] Step S203 , selecting and processing the pet image block pixel information according to a preset pet image block pixel information selection order to obtain pet image block pixel selection information and a plurality of pet image block pixel non-selected information.

[0049] In this embodiment, the preset order of selecting the pixel information of the pet image blocks can be manually set. The selection can be started from the first pixel information of the first row and first column, and then the pixel information of the first row and second column is selected. Then the pixel information of the first row and third column is selected. After the pixel information of the first row is selected, the selection starts from the pixel information of the second row and first column, and then the pixel information of the second row and second column is selected. And so on, until all the pixel information of the pet image blocks has been selected. It can be understood that each time the selection process is performed, the number of selections is increased by 1. Before all the pixel information of the pet image blocks is selected, the number of selections is updated and the next pixel information is automatically selected according to the selection order of the pet image block pixel information. It can be understood that each time one pet image block pixel information is selected, the other multiple pixel information that has not been selected is regarded as the unselected pet image block pixel information.

[0050] Step S204 , obtaining a plurality of initial pet image feature information according to a plurality of preset pet image feature extraction matrices, pet image block pixel selection information, and a plurality of pet image block pixel non-selection information.

[0051] In this embodiment, the preset pet image feature extraction matrix can be manually set and may include an edge feature matrix, a texture feature matrix, and a scale feature matrix. The edge feature matrix can be used to accurately extract the contour information of organs such as the pet's mouth and eyes; the texture feature matrix can be used to quantify microscopic changes on the skin surface, such as erythema granularity; and the scale feature matrix can be used to separate the ratio of solid and liquid components in feces. The method can first convert the selected pixel information of the pet image blocks into a one-dimensional vector, then perform a tensor product operation with the edge feature matrix, texture feature matrix, and scale feature matrix to generate multiple pet image feature response vectors. For the unselected pixel information of the pet image blocks, the correlation between the unselected pixel information of the pet image blocks and the selected pixel information of the pet image blocks can be calculated. The Euclidean distance can be calculated to quantify the correlation information. The correlation information is then element-wise multiplied by the pet image feature response vector to obtain a weighted pet image feature vector. All weighted pet image feature vectors are then concatenated, and the concatenated vector is used as the initial pet image feature information. By performing pixel-level processing on pet image block information, the interference of background noise such as pet hair and shooting environment on feature extraction is effectively suppressed. For example, when analyzing pet oral image information, the dark background area in the pixel is not selected. For example, the tongue information is given a low weight through Euclidean distance calculation, while the gum edge area in the selected pixel information is enhanced, thereby improving the detection sensitivity of early oral ulcers in pets.

[0052] The pet health monitoring method provided in the embodiment of the present application adaptively blocks pet image information through preset pet image block size information, and can dynamically adjust the block size for specific parts of the pet, such as the ear canal, excrement, etc., to ensure that the pet's subtle pathological features, such as crystallization of ear canal secretions and blood in excrement, are not missed. By performing pixel-level processing on the pet image block information, the interference of background noise such as pet hair and shooting environment on feature extraction is effectively suppressed. The pet image feature extraction matrix is ​​used to accurately extract subtle pixel features to capture subtle changes in the pet's health status, so that the texture changes in the early stage of skin fungal infection, abnormal excrement morphology caused by digestive system diseases, and other pet health change characteristics can be detected in time, thereby improving the accuracy and effectiveness of pet health monitoring.

[0053] Figure 3 The flowchart of the implementation of the pet health monitoring method provided in the third embodiment of the present application is shown. The difference between the third embodiment and the second embodiment is that the step S204 specifically includes:

[0054] Step S301 : calculating contrast information between pixels in a plurality of pet image blocks according to the selected pixel information of the pet image blocks and the non-selected pixel information of a plurality of pet image blocks.

[0055] In this embodiment, pixel values ​​of the selected pixel information for a pet image block can be differentiated point by point from pixel values ​​of the unselected pixel information for the pet image block to obtain pixel difference values ​​between adjacent pixels. Pixel values ​​can be grayscale values ​​or brightness values. Furthermore, by calculating the absolute difference between each selected pixel information for a pet image block and multiple unselected pixel information for the pet image blocks, and performing Gaussian weighting on the absolute difference, inter-pixel contrast information for the pet image blocks is generated to reflect the severity of local grayscale changes. For example, in pet skin image information, the inter-pixel contrast in healthy areas is low, while the inter-pixel contrast in diseased areas, such as erythema and dandruff, is high.

[0056] Step S302 : calculating local contrast information of pixels in the pet image blocks based on the contrast information between the pixels in the plurality of pet image blocks.

[0057] In this embodiment, spatial aggregation processing may be performed on the contrast information between pixels of each pet image block, median filtering processing may be performed on the contrast information between pixels of multiple pet image blocks, and then smoothing processing is performed using a Gaussian kernel function. Subsequently, based on manually set weight information, weighted averaging processing is performed on the contrast information between pixels of the smoothed pet image blocks. The result of the weighted averaging processing is used as the local contrast information of the pet image block pixels, which is used to characterize the texture complexity of the local area of ​​the pet image information, such as the wrinkled texture of the pet's oral mucosa, the granular structure of the pet's excrement, etc.

[0058] Step S303, determining whether the local contrast information of the pet image block pixels is greater than a preset local contrast threshold information of the pet image block pixels; if so, generating the pet image receptive field information based on the pet image block pixel selection information and the multiple pet image block pixel non-selected information; if not, skipping the pet image block pixel selection information and the multiple pet image block pixel non-selected information.

[0059] In this embodiment, the preset local contrast threshold information for the pet image block pixels can be manually set. When the local contrast information for the pet image block pixels is greater than the preset local contrast threshold information for the pet image block pixels, the local region is considered to contain potential pathological features, such as skin damage, corneal opacity, etc., and the current pet image block pixel selection information and the multiple pet image block pixel non-selected information are combined and processed to generate pet image receptive field information for subsequent feature extraction and disease probability identification calculation. When the local contrast information for the pet image block pixels is less than or equal to the preset local contrast threshold information for the pet image block pixels, the local region is considered to contain no potential pathological features, and the pet image block pixel selection information and the multiple pet image block pixel non-selected information are skipped, and the pet image receptive field information does not need to be generated.

[0060] Step S304 , counting the number of times the pet image block pixel information is selected to obtain pet image block pixel selection number information.

[0061] In this embodiment, each time the pet image block pixel information is selected once, the number of selections is increased by 1, so that the pet image block pixel selection information currently used for calculation and the number of selections corresponding to the non-selected information of multiple pet image block pixels can be determined as the pet image block pixel selection number information.

[0062] Step S305 , determining whether the number of times the pet image block pixel selection is less than the number of pet image block pixel information; if so, returning to step S203 ; if not, proceeding to step S306 .

[0063] In this embodiment, the sum of the pet image block pixel selection times information and the pet image block pixel quantity information are compared. When the pet image block pixel selection times information is less than the pet image block pixel quantity information, it means that the pet image block information has not been fully resolved, so it is necessary to continue to select and process the pet image block pixel information; when the pet image block pixel selection times information is equal to the pet image block pixel quantity information, it means that the pet image block information has been fully resolved, so there is no need to continue to select and process the pet image block pixel information, so as to ensure that all pixel information is resolved and avoid missing key features in the pet image information.

[0064] Step S306 : Mapping the pet image receptive field information based on a plurality of preset pet image feature extraction matrices to obtain a plurality of initial pet image feature information.

[0065] In this embodiment, multiple preset pet image feature extraction matrices can be manually set and may include an edge detection matrix and a texture analysis matrix. The pet image receptive field information can be first converted into a two-dimensional tensor, and then convolved with the edge detection matrix and the texture analysis matrix, respectively. The edge detection matrix can be used to extract the pet's oral contour information, while the texture matrix can be used to extract skin texture information. The convolution results are then concatenated and used as the independent variable of a ReLU function. After the ReLU function is calculated, the function value is output as the multiple initial pet image feature information.

[0066] The pet health monitoring method provided in the embodiment of the present application calculates the contrast information between the pixels of the pet image blocks to accurately identify the areas with large differences between the pixels. Combined with the local contrast information of the pixels of the pet image blocks, it effectively filters the interference of noise information and highlights the potential lesion areas with high contrast. By generating the receptive field information of the pet image, it fully captures the abnormal details of the potential lesion areas. By counting and judging the number of times the pixels of the pet image blocks are selected, it ensures that the pixel information of each pet image block is fully analyzed in multiple iterations to avoid missing the key features in the pet image information, thereby achieving accurate capture and efficient analysis of subtle lesions of the pet, thereby reducing the probability of misdiagnosis and missed diagnosis of pet diseases and improving the accuracy and effectiveness of pet health monitoring.

[0067] Figure 4 The flowchart of the pet health monitoring method provided in the fourth embodiment of the present application is shown. The difference between the fourth embodiment and the third embodiment is that:

[0068] The pet image receptive field information includes the central pixel information of the pet image receptive field and the peripheral pixel information of the plurality of pet image receptive fields;

[0069] The plurality of preset pet image feature extraction matrices include a preset pet image horizontal feature extraction weight matrix and a preset pet image vertical feature extraction weight matrix;

[0070] The step S306 specifically includes:

[0071] Step S401 : Based on a preset pet image level feature extraction weight matrix, weighted processing is performed on a plurality of peripheral pixel information of the pet image receptive fields to generate a plurality of peripheral pixel level enhancement information of the pet image receptive fields.

[0072] In this embodiment, the preset pet image horizontal feature extraction weight matrix can be manually set, such as a manually set one-dimensional vector. The weight values ​​can be symmetrically distributed along the horizontal direction to highlight horizontal pixel changes in the pixel information surrounding the pet image receptive field. A dot product operation can be performed on the pet image horizontal feature extraction weight matrix and the pixel information surrounding the pet image receptive field. The result of the dot product operation is used as the horizontal enhancement information of the pixels surrounding the pet image receptive field to enhance horizontal edge or texture features. For example, this can highlight the horizontal direction of the gum margin in pet oral images or enhance the horizontal diffusion of erythema in pet skin images.

[0073] Step S402 : Based on a preset pet image vertical feature extraction weight matrix, weighted processing is performed on a plurality of peripheral pixel information of the pet image receptive fields to generate a plurality of vertical enhancement information of the peripheral pixels of the pet image receptive fields.

[0074] In this embodiment, the preset weight matrix for extracting vertical features from pet images can be manually set, such as a one-dimensional vector. The weight values ​​can be symmetrically distributed along the vertical direction, thereby emphasizing vertical pixel variations in the peripheral pixel information of the pet image receptive field. A dot product operation can be performed between the weight matrix for extracting vertical features from pet images and the peripheral pixel information of the pet image receptive field. The result of the dot product operation can be used as vertical enhancement information for the peripheral pixels of the pet image receptive field, thereby enhancing vertical edge or texture features. For example, in pet ear canal images, the vertical texture of pet secretion deposits can be highlighted, or the longitudinal distribution of solid particles in pet excrement images can be enhanced.

[0075] Step S403 , obtaining a plurality of initial pet image feature information based on a plurality of horizontal enhancement information of peripheral pixels of the pet image receptive field, a plurality of vertical enhancement information of peripheral pixels of the pet image receptive field, and information of a central pixel of the pet image receptive field.

[0076] In this embodiment, the central pixel information of the pet image receptive field can be first converted into a vector of the same dimension as the peripheral pixel information of the pet image receptive field, and then the converted central pixel information of the pet image receptive field is spliced ​​with the vertical enhancement information of the peripheral pixels of the pet image receptive field and the central pixel information of the pet image receptive field to form a three-dimensional feature vector including the central pixel intensity, horizontal features, and vertical features, and the three-dimensional feature vector is used as the initial pet image feature information.

[0077] In this embodiment, optionally, the average grayscale value of the local area can be represented by the central pixel information, the horizontal edge response can be reflected by the horizontal enhancement information, and the vertical edge response can be reflected by the vertical enhancement information. Then, the values ​​of the central pixel intensity, horizontal features, and vertical features can be divided by the maximum value of the pixel value to eliminate the numerical difference, and the central pixel intensity, horizontal features, and vertical features after eliminating the numerical difference are spliced ​​to generate the initial pet image feature information.

[0078] The pet health monitoring method provided in the embodiment of the present application introduces a pet image horizontal feature extraction weight matrix and a pet image vertical feature extraction weight matrix to directionally enhance the edge features and texture features in specific directions in the pet image information, so as to finely capture the directional features in the pet image information, thereby deeply analyzing the structural information of the pet's lesion area, and is used to determine the lateral range of the pet's gingival redness and swelling, identify the depth of the pet's oral mucosa wrinkles, monitor irregular patches on the pet's skin surface, etc., thereby fully analyzing the pet image information to effectively improve the accuracy and effectiveness of pet health monitoring.

[0079] Figure 5 The flowchart of the pet health monitoring method provided in the fifth embodiment of the present application is shown. The difference between the fifth embodiment and the first embodiment is that the step S103 specifically includes:

[0080] Step S501: Sampling the plurality of initial pet image feature information according to a preset pet image feature low-pass sampling matrix and a preset pet image feature high-pass sampling matrix to obtain a plurality of initial pet image feature low-frequency information and a plurality of initial pet image feature high-frequency information.

[0081] In this embodiment, both the preset low-pass sampling matrix and the preset high-pass sampling matrix for pet image features can be manually set. The preset low-pass sampling matrix can be a mean filter kernel, used to extract low-frequency components from the initial pet image feature information, namely, global structural features, such as the overall oral contour and large areas of erythema on the skin. The preset high-pass sampling matrix can be a Laplacian kernel, used to extract high-frequency components, namely, local detail features, such as subtle jagged edges on the gum line and the dotted distribution of dandruff. The initial pet image feature information can be decomposed into initial low-frequency and high-frequency information by convolving the low-pass and high-pass sampling matrices with the initial pet image feature information. For example, in pet excrement image information, the low-frequency information can reflect the overall morphology, such as the liquid / solid ratio, while the high-frequency information can capture subtle abnormalities such as blood streaks and mucosal debris.

[0082] Step S502: Sampling and processing the plurality of initial pet image feature low-frequency information and the plurality of initial pet image feature high-frequency information according to a preset pet image feature low-pass sampling matrix and a preset pet image feature high-pass sampling matrix to obtain a plurality of target pet image feature information.

[0083] In this embodiment, a convolution calculation can be performed on the preset pet image feature low-pass sampling matrix and the initial pet image feature low-frequency information to achieve re-low-pass sampling of the initial pet image feature low-frequency information, further compress the resolution of the global features, and generate a simpler low-frequency representation, such as simplifying the pet's oral contour into a geometric shape; a convolution calculation can be performed on the preset pet image feature high-pass sampling matrix and the initial pet image feature high-frequency information to achieve re-high-pass sampling of the initial pet image feature high-frequency information, thereby enhancing the sharpness of the detail features, such as highlighting the texture differences at the edges of pet skin lesions.

[0084] The pet health monitoring method provided in the embodiment of the present application performs multi-frequency decomposition processing on the initial pet image feature information to characterize the macroscopic morphology and microscopic lesions of the pet body, avoids mutual interference between different pet image features, and improves the distinguishability of the pet image feature information. The decomposed initial pet image feature low-frequency information and the initial pet image feature high-frequency information are secondary sampled to filter out noise information and interference information in the pet image feature information, so as to effectively meet the needs of pet health monitoring in home scenarios and improve the accuracy and reliability of pet health monitoring.

[0085] Figure 6The flowchart of the pet health monitoring method provided in Example 6 of the present application is shown. The difference between the method and Example 5 above is that:

[0086] The initial pet image feature low-frequency information includes initial pet image horizontal feature low-frequency information, initial pet image vertical feature low-frequency information, and initial pet image diagonal feature low-frequency information;

[0087] The initial pet image feature high-frequency information includes initial pet image horizontal feature high-frequency information, initial pet image vertical feature high-frequency information, and initial pet image diagonal feature high-frequency information;

[0088] The step S502 specifically includes:

[0089] Step S601 : Sampling the initial pet image horizontal feature low-frequency information according to a preset pet image feature low-pass sampling matrix to obtain target pet image horizontal feature low-frequency information.

[0090] In this embodiment, the preset pet image feature low-pass sampling matrix can be manually set, and a large-scale mean kernel can be used. A convolution calculation can be performed on the pet image feature low-pass sampling matrix with the initial pet image horizontal feature low-frequency information, and the result of the convolution calculation can be used as the target pet image horizontal feature low-frequency information to achieve subsampling of the initial pet image horizontal feature low-frequency information, thereby further weakening local fluctuations and enhancing the horizontal image change trend characteristics.

[0091] Step S602 : Sampling the initial pet image horizontal feature high-frequency information according to a preset pet image feature high-pass sampling matrix to obtain target pet image horizontal feature high-frequency information.

[0092] In this embodiment, the preset high-pass sampling matrix for the pet image features can be manually set, and can use a sharpening kernel or a convolution kernel with a center of 3 and a neighborhood of -1. A convolution calculation can be performed on the high-pass sampling matrix for the pet image features and the initial high-frequency information of the pet image horizontal features. The result of the convolution calculation is used as the target high-frequency information of the pet image horizontal features, thereby enhancing the initial high-frequency information of the pet image horizontal features to highlight the high-frequency details in the horizontal direction.

[0093] Step S603 : Sampling the initial pet image vertical feature low-frequency information according to a preset pet image feature low-pass sampling matrix to obtain target pet image vertical feature low-frequency information.

[0094] In this embodiment, the preset pet image feature low-pass sampling matrix can be manually set. The preset pet image feature low-pass sampling matrix can be convolved with the initial pet image vertical feature low-frequency information, and the convolution result is used as the target pet image vertical feature low-frequency information to extract vertical macro-level information.

[0095] Step S604 : Sampling the high-frequency information of the vertical features of the initial pet image according to a preset high-pass sampling matrix of the pet image features to obtain the high-frequency information of the vertical features of the target pet image.

[0096] In this embodiment, the preset high-pass sampling matrix for pet image features can be manually set. The preset high-pass sampling matrix for pet image features can be convolved with the high-frequency information of the vertical features of the initial pet image, and the result of the convolution calculation is used as the high-frequency information of the vertical features of the target pet image to enhance the detail contrast in the vertical direction.

[0097] Step S605, performing fusion processing based on the initial pet image horizontal feature low-frequency information, the initial pet image vertical feature low-frequency information, the target pet image horizontal feature low-frequency information, the pet image diagonal feature low-frequency information, the initial pet image horizontal feature high-frequency information, the initial pet image vertical feature high-frequency information, the target pet image vertical feature high-frequency information and the pet image diagonal feature high-frequency information to obtain the target pet image feature information.

[0098] In this embodiment, the initial pet image horizontal feature low-frequency information, the initial pet image vertical feature low-frequency information, and the pet image diagonal feature low-frequency information are spliced ​​to generate the pet image low-frequency structural feature information, and then the initial pet image horizontal feature high-frequency information, the initial pet image vertical feature high-frequency information, and the pet image diagonal feature high-frequency information are spliced ​​to generate the pet image high-frequency structural feature information, and then the target pet image horizontal feature low-frequency information is dot-producted with the initial pet image horizontal feature high-frequency information, and the result of the dot-product calculation is used as the pet image horizontal feature high-frequency suppression information, and the low-frequency structural feature suppression information is used. Suppress high-frequency noise in healthy areas, such as the uniform texture high-frequency response of normal pet skin, and enhance lesion-related high-frequency signals, such as the irregular high-frequency fluctuations at the edge of pet rashes; perform dot product calculation on the high-frequency information of the vertical features of the target pet image and the low-frequency information of the vertical features of the initial pet image, and use the result of the dot product calculation as the high-frequency suppression information of the vertical features of the pet image; then fuse the low-frequency structural feature information of the pet image, the high-frequency structural feature information of the pet image, the high-frequency suppression information of the horizontal features of the pet image, and the high-frequency suppression information of the vertical features of the pet image to generate the target pet image feature information, thereby reducing misdiagnosis caused by environmental interference or other noise reasons.

[0099] The pet health monitoring method provided in the embodiment of the present application performs secondary sampling processing on the initial pet image feature information through a preset pet image feature low-pass sampling matrix and a preset pet image feature high-pass sampling matrix to extract the global structural features of the pet body. Through deep analysis processing of horizontal feature information, vertical feature information and diagonal features, the potential pathological features distributed in the pet body are fully captured, thereby improving the accuracy and effectiveness of monitoring pet health through pet image information.

[0100] Figure 7 The flowchart of the pet health monitoring method provided in the seventh embodiment of the present application is shown. The difference between the seventh embodiment and the first embodiment is that the step S104 specifically includes:

[0101] Step S701 , performing semantic information extraction on the multi-source pet clinical case information to obtain pet clinical disease information, pet clinical symptom information, pet case breed information, and pet disease age information.

[0102] In this embodiment, natural language processing technology can be used to perform word segmentation and named entity recognition on text information in multi-source pet clinical case information, thereby realizing semantic information extraction. For example, from "poodle, 8 years old, vomiting for 3 days, accompanied by diarrhea, and blood in stool", the pet case breed information as poodle, the pet's age information as 8 years old, and pet clinical symptom information such as vomiting, diarrhea, and blood in stool can be extracted. Then, a preset pet clinical disease dictionary is used to map the information to standardized pet clinical disease information, such as digestive system diseases.

[0103] Step S702 : extracting semantic association information from the multi-source pet clinical case information to obtain pet clinical disease symptom association information, pet breed disease association information, and pet age disease information.

[0104] In this embodiment, historical case data can be analyzed through an association rule mining algorithm to construct a probabilistic association network of "disease-symptoms", "breed-disease", and "age-disease". For example, statistics show that the association strength of "Golden Retriever-hip dysplasia" is 85%, the association probability of "puppy-parasitic infection" is 70%, and the co-occurrence frequency of "oral ulcer-bad breath" is 60%, so as to form pet clinical disease symptom association information, pet breed disease association information, and pet age disease information.

[0105] Step S703, generating a plurality of pet clinical case characteristic information based on the pet clinical disease information, pet clinical symptom information, pet case breed information, pet illness age information, pet clinical disease symptom association information, pet breed disease association information and pet age disease information.

[0106] In this embodiment, pet case breed information can first be one-hot encoded, and the pet age information can be normalized (mapped to the interval [0, 1]). The pet clinical disease information and pet clinical symptom information can then be converted into a Boolean vector. This Boolean vector is then concatenated with the pet clinical disease symptom association information, the pet breed disease association information, and the pet age disease information to form multidimensional pet clinical case feature information containing both semantic and association features. For example, for a poodle diarrhea case, a vector containing features such as "Breed - Poodle," "Age - Adult," "Symptoms - Diarrhea + Bloodshot," "Association - Poodle - Gastroenteritis (0.7)," and "Age - Adult - Gastroenteritis (0.6)" is generated.

[0107] Step S704: Generate multiple pet health monitoring information based on the multiple target pet image feature information and pet clinical case feature information.

[0108] In this embodiment, the cosine similarity of the target pet image feature information and the pet clinical case feature information can be calculated, and then the cosine similarity can be used as the probability of the pet being sick. The probability of the pet being sick can be combined with the pet clinical disease information and the pet clinical symptom information to output pet health monitoring information including the disease name, probability value, and associated symptoms to characterize the probability of the pet being sick.

[0109] The pet health monitoring method provided in the embodiment of the present application converts non-structured multi-source pet clinical case information into standardized features such as pet clinical disease information through semantic information extraction, and quantifies the correlation strength between pet clinical disease information, pet clinical symptom information, pet case breed information and pet disease age information through semantic association extraction, thereby generating pet clinical case feature information and integrating multi-dimensional information to calculate monitoring confidence, thereby avoiding the missed detection of early minor pet lesions by single-dimensional information, thereby reducing the probability of missed detection of abnormal pet health conditions in home scenarios, and improving the accuracy, robustness and timeliness of pet health monitoring.

[0110] Figure 8 The flowchart of the implementation of the pet health monitoring method provided in the eighth embodiment of the present application is shown. The difference between the eighth embodiment and the seventh embodiment is that the step S704 specifically includes:

[0111] Step S801 : Mapping the plurality of target pet image feature information based on a preset pet image feature mapping matrix to generate target pet image feature mapping information.

[0112] In this embodiment, the preset pet image feature mapping matrix can be manually set. The pet image feature mapping matrix can be multiplied by the target pet image feature information, and the multiplication result is used as the target pet image feature mapping information to map the target pet image feature information into a high-dimensional space.

[0113] Step S802 : Based on a preset pet clinical case feature index mapping matrix and a preset pet clinical case feature representation mapping matrix, the pet clinical case feature information is mapped to generate pet clinical case feature index mapping information and pet clinical case feature representation mapping information.

[0114] In this embodiment, the preset pet clinical case feature index mapping matrix and the preset pet clinical case feature representation mapping matrix can both be set manually, and can be multiplied by the preset pet clinical case feature index mapping matrix and the preset pet clinical case feature representation mapping matrix with the pet clinical case feature information respectively, and the multiplication results are used as the pet clinical case feature index mapping information and the pet clinical case feature representation mapping information respectively.

[0115] Step S803 : generating pet image-case interaction mapping feature information based on the target pet image feature mapping information and the pet clinical case feature index mapping information.

[0116] In this embodiment, the target pet image feature mapping information and the pet clinical case feature index mapping information may be multiplied, and the multiplication result is used as the pet image-case interaction mapping feature information.

[0117] Step S804 : generating pet image-case interaction mapping feature representation information based on the pet image-case interaction mapping feature information and the pet clinical case feature representation mapping information.

[0118] In this embodiment, the pet image-case interaction mapping feature information and the pet clinical case feature representation mapping information may be multiplied, and the multiplication result is used as the pet image-case interaction mapping feature representation information.

[0119] Step S805 : generating target pet image feature information to be fused based on the target pet image feature information and the pet image-case interactive mapping feature representation information.

[0120] In this embodiment, the target pet image feature information and the pet image-case interaction mapping feature representation information may be multiplied, and the multiplication result is used as the target pet image feature information to be fused.

[0121] Step S806 , generating pet health monitoring feature fusion information based on the target pet image feature information to be fused and the pet clinical case feature information.

[0122] In this embodiment, the target pet image feature information to be fused and the pet clinical case feature information may be multiplied together, and the multiplication result may be used as the pet health monitoring feature fusion information.

[0123] Step S807 , calculating pet health monitoring matching probability information based on the pet health monitoring feature fusion information and a preset pet health monitoring feature matching probability calculation function.

[0124] In this embodiment, the preset pet health monitoring feature matching probability calculation function can be manually set, and a preset Softmax function can be used. Alternatively, the pet health monitoring feature fusion information can be used as an independent variable of the pet health monitoring feature matching probability calculation function, and the function value obtained by the pet health monitoring feature matching probability calculation function can be used as the pet health monitoring matching probability information.

[0125] Step S808 : generating a plurality of pet health monitoring information according to the pet health monitoring matching probability information, the pet clinical disease information corresponding to the pet health monitoring matching probability information, and the pet clinical symptom information.

[0126] In this embodiment, the pet health monitoring matching probability information, the pet clinical disease information corresponding to the pet health monitoring matching probability information, and the pet clinical symptom information can be structured and output to generate pet health monitoring information, such as "rickets, probability of disease 0.94, associated symptoms: hind limb lameness, joint swelling".

[0127] The pet health monitoring method provided in the embodiment of the present application maps the target pet image feature information through a pet image feature mapping matrix, a pet clinical case feature index mapping matrix, and a pet clinical case feature representation mapping matrix, and then associates the mapped pet image feature information with multi-source pet clinical case information, thereby solving the problem of the separation of image information and pathological text semantics in the prior art, thereby improving the accuracy and effectiveness of pet health monitoring, and enhancing the user's trust in the monitoring results by accurately outputting pet health monitoring information, thereby meeting the pet health monitoring needs of non-professional users in home scenarios and effectively improving residents' sense of security and happiness.

[0128] Corresponding to the method of the above embodiment, Figure 9 A structural block diagram of a pet health monitoring device provided in an embodiment of the present application is shown. For ease of explanation, only the parts related to the embodiment of the present application are shown. Figure 9The exemplary pet health monitoring device may be an executing entity of the pet health monitoring method provided in the aforementioned first embodiment.

[0129] Reference Figure 9 , the pet health monitoring device comprises:

[0130] An information acquisition module 910 is used to acquire multiple pet image information and multi-source pet clinical case information;

[0131] An initial pet image feature information generating module 920 is configured to perform feature extraction processing on the plurality of pet image information based on a plurality of preset pet image feature extraction matrices to obtain a plurality of initial pet image feature information;

[0132] a target pet image feature information generating module 930 for performing decomposition and transformation processing on the plurality of initial pet image feature information to obtain a plurality of target pet image feature information;

[0133] The pet health monitoring information generating module 940 is configured to generate a plurality of pet health monitoring information based on the plurality of target pet image feature information and multi-source pet clinical case information.

[0134] The process of each module in the pet health monitoring device provided in the embodiment of the present application realizing its own function can be specifically referred to the aforementioned Figure 1 The description of the first embodiment is omitted here.

[0135] It should be understood that the size of the serial numbers of the steps in the above embodiments does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0136] It should be understood that when used in the present specification and the appended claims, the term "comprising" indicates the presence of described features, integers, steps, operations, elements and / or components, but does not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or collections thereof.

[0137] It will also be understood that the term "and / or" used in this specification and the appended claims refers to and includes any and all possible combinations of one or more of the associated listed items.

[0138] As used in this specification and the appended claims, the term "if" can be interpreted as "when" or "upon" or "in response to determining" or "in response to detecting," depending on the context. Similarly, the phrase "if it is determined" or "if [described condition or event] is detected" can be interpreted as meaning "upon determination" or "in response to determining" or "upon detection of [described condition or event]" or "in response to detecting [described condition or event]," depending on the context.

[0139] In addition, in the description of the present specification and the appended claims, the terms "first", "second", "third", etc. are only used to distinguish descriptions and should not be understood as indicating or implying relative importance. It should also be understood that although the terms "first", "second", etc. are used in the text to describe various elements in some embodiments of the present application, these elements should not be limited by these terms. These terms are only used to distinguish one element from another. For example, a first table can be named a second table, and similarly, a second table can be named a first table without departing from the scope of the various described embodiments. Both the first table and the second table are tables, but they are not the same table.

[0140] References to "one embodiment" or "some embodiments" in this specification mean that a particular feature, structure, or characteristic described in conjunction with that embodiment is included in one or more embodiments of the present application. Thus, phrases such as "in one embodiment," "in some embodiments," "in other embodiments," and "in other embodiments" appearing in various places in this specification do not necessarily refer to the same embodiment, but rather mean "one or more but not all embodiments," unless otherwise specifically emphasized. The terms "including," "comprising," "having," and variations thereof all mean "including but not limited to," unless otherwise specifically emphasized.

[0141] The pet health monitoring method provided in the embodiments of the present application can be applied to terminal devices such as mobile phones, tablet computers, wearable devices, vehicle-mounted devices, augmented reality (AR) / virtual reality (VR) devices, laptop computers, ultra-mobile personal computers (UMPCs), netbooks, and personal digital assistants (PDAs). The embodiments of the present application do not impose any restrictions on the specific type of terminal device.

[0142] For example, the terminal device can be a station (STAION, ST) in a WLAN, a cellular phone, a cordless phone, a Session Initiation Protocol (SIP) phone, a Wireless Local Loop (WLL) station, a Personal Digital Assistant (PDA) device, a handheld device with wireless communication function, a computing device or other processing device connected to a wireless modem, a vehicle-mounted device, a vehicle networking terminal, a computer, a laptop computer, a handheld communication device, a handheld computing device, a satellite wireless device, a wireless modem card, a TV set-top box (STB), customer premise equipment (CPE) and / or other devices for communicating on a wireless system and a next-generation communication system, such as a mobile terminal in a 5G network or a mobile terminal in a future evolved Public Land Mobile Network (PLMN) network.

[0143] As an example and not a limitation, when the terminal device is a wearable device, the wearable device can also be a general term for wearable devices that are intelligently designed and developed using wearable technology for daily wear, such as glasses, gloves, watches, clothing and shoes. A wearable device is a portable device that is worn directly on the body or integrated into the user's clothes or accessories. Wearable devices are not only hardware devices, but also achieve powerful functions through software support, data interaction, and cloud interaction. Broadly speaking, wearable smart devices include those that are full-featured, large in size, and can achieve complete or partial functions without relying on smartphones, such as smart watches or smart glasses, as well as those that only focus on a certain type of application function and need to be used in conjunction with other devices such as smartphones, such as various smart bracelets and smart jewelry for vital sign monitoring.

[0144] Figure 10 This is a schematic diagram of the structure of a terminal device provided by an embodiment of the present application. Figure 10 As shown, the terminal device 100 of this embodiment includes: at least one processor 1000 ( Figure 10 Only one is shown), a memory 1001, wherein the memory 1001 stores a computer program 1002 that can be run on the processor 1000. When the processor 1000 executes the computer program 1002, the steps in the above-mentioned embodiments of the pet health monitoring method are implemented, such as Figure 1 Alternatively, when the processor 1000 executes the computer program 1002, the functions of the modules / units in the above-mentioned device embodiments are realized, for example, Figure 9 Functions of modules 910 to 940 are shown.

[0145] The terminal device 100 may be a computing device such as a desktop computer, a notebook, a PDA, or a cloud server. The terminal device may include, but is not limited to, a processor 1000 and a memory 1001. Those skilled in the art will understand that Figure 10 It is only an example of the terminal device 100 and does not constitute a limitation of the terminal device 100. It may include more or fewer components than shown in the figure, or a combination of certain components, or different components. For example, the terminal device may also include an input and sending device, a network access device, a bus, etc.

[0146] The processor 1000 may be a central processing unit (CPU), other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA), other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor.

[0147] In some embodiments, the memory 1001 may be an internal storage unit of the terminal device 100, such as a hard disk or memory of the terminal device 100. The memory 1001 may also be an external storage device of the terminal device 100, such as a plug-in hard disk, a Smart Media Card (SMC), a Secure Digital (SD) card, a Flash Card, etc. equipped on the terminal device 100. Furthermore, the memory 1001 may include both an internal storage unit of the terminal device 100 and an external storage device. The memory 1001 is used to store an operating system, application programs, a boot loader, data, and other programs, such as the program code of the computer program. The memory 1001 may also be used to temporarily store data that has been sent or is to be sent.

[0148] In addition, the functional units in the various embodiments of the present application may be integrated into a single processing unit, or each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or software functional units.

[0149] An embodiment of the present application also provides a terminal device, which includes at least one memory, at least one processor, and a computer program stored in the at least one memory and executable on the at least one processor. When the processor executes the computer program, the terminal device implements the steps of any of the above-mentioned method embodiments.

[0150] An embodiment of the present application further provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps in the above-mentioned various method embodiments can be implemented.

[0151] An embodiment of the present application provides a computer program product. When the computer program product is run on a terminal device, the terminal device can implement the steps in the above-mentioned method embodiments when executing the computer program product.

[0152] If the integrated module / unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present application can implement all or part of the processes in the above-mentioned embodiment method by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium. When the computer program is executed by a processor, it can implement the steps of each of the above-mentioned method embodiments. The computer program includes computer program code, which can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium can include: any entity or device that can carry the computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signal, telecommunication signal and software distribution medium.

[0153] In the above embodiments, the description of each embodiment has its own focus. For parts that are not described or recorded in detail in a certain embodiment, reference can be made to the relevant description of other embodiments.

[0154] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0155] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0156] The above-described embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present application, and should all be included in the scope of protection of the present application.

Claims

1. A pet health monitoring method, characterized in that: include: Acquire multiple pet image information and multi-source pet clinical case information; Based on a plurality of preset pet image feature extraction matrices, performing feature extraction processing on the plurality of pet image information to obtain a plurality of initial pet image feature information; Decomposing and transforming the plurality of initial pet image feature information to obtain a plurality of target pet image feature information; A plurality of pet health monitoring information is generated based on the plurality of target pet image feature information and multi-source pet clinical case information.

2. The pet health monitoring method according to claim 1, wherein: The step of performing feature extraction processing on the plurality of pet image information based on a plurality of preset pet image feature extraction matrices to obtain a plurality of initial pet image feature information specifically includes: Based on preset pet image block size information, the plurality of pet image information is subjected to block processing to obtain a plurality of pet image block information; Extracting pixel information of the plurality of pet image block information to obtain pixel information of the plurality of pet image blocks and pixel quantity information of the pet image blocks; Selecting and processing the pet image block pixel information according to a preset pet image block pixel information selection order to obtain pet image block pixel selected information and a plurality of pet image block pixel unselected information; A plurality of initial pet image feature information is obtained according to a plurality of preset pet image feature extraction matrices, pet image block pixel selection information and a plurality of pet image block pixel non-selection information.

3. The pet health monitoring method according to claim 2, wherein: The step of obtaining a plurality of initial pet image feature information based on a plurality of preset pet image feature extraction matrices, pet image block pixel selection information, and a plurality of pet image block pixel non-selection information specifically includes: Calculating contrast information between pixels in multiple pet image blocks based on the pet image block pixel selection information and non-selected pixel information in multiple pet image blocks; Calculating local contrast information of the pet image block pixels based on the contrast information between the plurality of pet image block pixels; When the local contrast information of the pet image block pixels is greater than the preset local contrast threshold information of the pet image block pixels, generating the pet image receptive field information according to the pet image block pixel selection information and the non-selected information of the plurality of pet image block pixels; Counting the number of times the pet image block pixel information is selected to obtain information on the number of times the pet image block pixel information is selected; Determine whether the number of times the pixel selections of the pet image blocks are less than the number of pixels of the pet image blocks; If so, returning to the step of selecting and processing the pet image block pixel information according to the preset pet image block pixel information selection order to obtain pet image block pixel selected information and a plurality of pet image block pixel unselected information; If not, mapping processing is performed on the pet image receptive field information based on a plurality of preset pet image feature extraction matrices to obtain a plurality of initial pet image feature information.

4. The pet health monitoring method according to claim 3, wherein: The pet image receptive field information includes the central pixel information of the pet image receptive field and the peripheral pixel information of the plurality of pet image receptive fields; The plurality of preset pet image feature extraction matrices include a preset pet image horizontal feature extraction weight matrix and a preset pet image vertical feature extraction weight matrix; The step of mapping the pet image receptive field information based on a plurality of preset pet image feature extraction matrices to obtain a plurality of initial pet image feature information specifically includes: Based on a preset pet image level feature extraction weight matrix, weighted processing is performed on a plurality of peripheral pixel information of the pet image receptive field to generate a plurality of peripheral pixel level enhancement information of the pet image receptive field; Based on a preset pet image vertical feature extraction weight matrix, weighted processing is performed on a plurality of peripheral pixel information of the pet image receptive field to generate a plurality of vertical enhancement information of the peripheral pixels of the pet image receptive field; A plurality of initial pet image feature information is obtained based on the horizontal enhancement information of the peripheral pixels of the receptive field of the pet image, the vertical enhancement information of the peripheral pixels of the receptive field of the pet image, and the central pixel information of the receptive field of the pet image.

5. The pet health monitoring method according to claim 1, wherein: The step of performing decomposition and transformation processing on the plurality of initial pet image feature information to obtain the plurality of target pet image feature information specifically includes: Sampling the plurality of initial pet image feature information according to a preset pet image feature low-pass sampling matrix and a preset pet image feature high-pass sampling matrix to obtain a plurality of initial pet image feature low-frequency information and a plurality of initial pet image feature high-frequency information; According to a preset pet image feature low-pass sampling matrix and a preset pet image feature high-pass sampling matrix, a plurality of the initial pet image feature low-frequency information and a plurality of the initial pet image feature high-frequency information are sampled and processed to obtain a plurality of target pet image feature information.

6. The pet health monitoring method according to claim 5, wherein: The initial pet image feature low-frequency information includes initial pet image horizontal feature low-frequency information, initial pet image vertical feature low-frequency information, and initial pet image diagonal feature low-frequency information; The initial pet image feature high-frequency information includes initial pet image horizontal feature high-frequency information, initial pet image vertical feature high-frequency information, and initial pet image diagonal feature high-frequency information; The step of sampling and processing the plurality of initial pet image feature low-frequency information and the plurality of initial pet image feature high-frequency information according to a preset pet image feature low-pass sampling matrix and a preset pet image feature high-pass sampling matrix to obtain the plurality of target pet image feature information specifically includes: Sampling the initial pet image horizontal feature low-frequency information according to a preset pet image feature low-pass sampling matrix to obtain target pet image horizontal feature low-frequency information; Sampling the initial pet image horizontal feature high-frequency information according to a preset pet image feature high-pass sampling matrix to obtain target pet image horizontal feature high-frequency information; Sampling the initial pet image vertical feature low-frequency information according to a preset pet image feature low-pass sampling matrix to obtain the target pet image vertical feature low-frequency information; Sampling the high-frequency information of the vertical features of the initial pet image according to a preset high-pass sampling matrix of the pet image features to obtain the high-frequency information of the vertical features of the target pet image; The target pet image feature information is obtained by performing fusion processing on the initial pet image horizontal feature low-frequency information, the initial pet image vertical feature low-frequency information, the target pet image horizontal feature low-frequency information, the pet image diagonal feature low-frequency information, the initial pet image horizontal feature high-frequency information, the initial pet image vertical feature high-frequency information, the target pet image vertical feature high-frequency information and the pet image diagonal feature high-frequency information.

7. The pet health monitoring method according to claim 1, wherein: The step of generating a plurality of pet health monitoring information based on the plurality of target pet image feature information and multi-source pet clinical case information specifically includes: Extracting semantic information from the multi-source pet clinical case information to obtain pet clinical disease information, pet clinical symptom information, pet case breed information, and pet disease age information; Extracting semantic association information from the multi-source pet clinical case information to obtain pet clinical disease symptom association information, pet breed disease association information, and pet age disease information; Generate multiple pet clinical case feature information based on the pet clinical disease information, pet clinical symptom information, pet case breed information, pet illness age information, pet clinical disease symptom association information, pet breed disease association information, and pet age disease information; A plurality of pet health monitoring information is generated based on the plurality of target pet image feature information and pet clinical case feature information.

8. The pet health monitoring method according to claim 7, wherein: The step of generating a plurality of pet health monitoring information based on the plurality of target pet image feature information and pet clinical case feature information specifically includes: Based on a preset pet image feature mapping matrix, mapping the plurality of target pet image feature information to generate target pet image feature mapping information; Based on a preset pet clinical case feature index mapping matrix and a preset pet clinical case feature representation mapping matrix, mapping the pet clinical case feature information to generate pet clinical case feature index mapping information and pet clinical case feature representation mapping information; Generate pet image-case interaction mapping feature information based on the target pet image feature mapping information and the pet clinical case feature index mapping information; generating pet image-case interaction mapping feature representation information based on the pet image-case interaction mapping feature information and the pet clinical case feature representation mapping information; Generate target pet image feature information to be fused based on the target pet image feature information and the pet image case interactive mapping feature representation information; Generate pet health monitoring feature fusion information based on the target pet image feature information to be fused and the pet clinical case feature information; Calculating pet health monitoring matching probability information based on the pet health monitoring feature fusion information and a preset pet health monitoring feature matching probability calculation function; A plurality of pet health monitoring information is generated according to the pet health monitoring matching probability information, the pet clinical disease information corresponding to the pet health monitoring matching probability information, and the pet clinical symptom information.

9. A pet health monitoring device, characterized in that: include: An information acquisition module, used to acquire multiple pet image information and multi-source pet clinical case information; an initial pet image feature information generating module, configured to perform feature extraction processing on the plurality of pet image information based on a plurality of preset pet image feature extraction matrices to obtain a plurality of initial pet image feature information; a target pet image feature information generating module, configured to perform decomposition and transformation processing on the plurality of initial pet image feature information to obtain a plurality of target pet image feature information; The pet health monitoring information generation module is used to generate multiple pet health monitoring information based on the multiple target pet image feature information and multi-source pet clinical case information.

10. A terminal device, characterized in that: The terminal device includes a memory and a processor, wherein the memory stores a computer program that can be run on the processor, and when the processor executes the computer program, the steps of the method according to any one of claims 1 to 8 are implemented.

Images