Animal disease diagnosis method, device, equipment, storage medium and program product
By collecting and analyzing multimodal data in the farm and inputting it into the disease diagnosis model for animal disease diagnosis, the problems of low efficiency and low accuracy of traditional diagnostic methods are solved, and more efficient and accurate diagnosis of animal disease is achieved.
Patent Information
- Application Number
- CN202510539941.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-04-27
- Publication Date
- 2025-05-30
AI Technical Summary
Traditional artificial animal disease diagnosis methods are inefficient and have low accuracy when facing large-scale farms, and rely on the subjective judgment of veterinarians, making them prone to misdiagnosis or misdiagnosis.
By obtaining multimodal data from the farm, including environmental data, physiological data, acoustic data and image data, preprocessing and feature extraction are performed, and inputting them to the trained disease diagnosis model for diagnosis.
It improves the efficiency and accuracy of animal disease diagnosis, reduces labor costs, and provides an objective and unified evaluation method.
Smart Images
Figure CN120072346A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the technical field of intelligent agriculture, and particularly to an animal disease diagnosis method, device, equipment, storage medium, and program product. Background Art
[0002] In recent years, with the booming development of large-scale animal husbandry, the breeding scale of farms has been continuously expanding, and the number of animals in a single farm has grown from hundreds to tens of thousands. Although this large-scale trend has improved animal breeding efficiency, it has also brought huge challenges to animal health management. Traditional manual animal disease diagnosis methods mainly rely on veterinarians' experience judgment and regular inspections. This method is inadequate when faced with large-scale farms. Veterinarians need to spend a lot of time on inspections, often unable to detect and handle health problems in a timely manner. The disease diagnosis efficiency of animals is low, resulting in unsatisfactory disease prevention and control effects. Moreover, traditional manual animal disease diagnosis methods overly rely on veterinarians' subjective judgments and are easily affected by personal experience and work status. Different veterinarians may have different judgments on the same symptom, lacking a unified diagnosis standard and objective evaluation method. When faced with complex or atypical symptoms, the accuracy of relying solely on experience judgment is often not ideal, and misdiagnosis or missed diagnosis is likely to occur, resulting in a low accuracy rate of animal disease diagnosis. Summary of the Invention
[0003] Based on this, it is necessary to provide an animal disease diagnosis method, device, equipment, storage medium, and program product that can improve the efficiency and accuracy of animal disease diagnosis for the above technical problems.
[0004] In a first aspect, this application provides an animal disease diagnosis method, and the method includes:
[0005] Obtain the original multi-modal breeding data collected for the farm; the original multi-modal breeding data includes the environmental data of the farm, the physiological data and acoustic data of the animals in the farm, and the image data collected for the farm;
[0006] Preprocess the original multi-modal breeding data to obtain target multi-modal breeding data;
[0007] Input the target multi-modal breeding data into the trained disease diagnosis model to extract the data features of the target multi-modal breeding data through the disease diagnosis model, and perform disease diagnosis on the animals in the farm based on the data features to obtain a disease diagnosis result.
[0008] In one embodiment, the preprocessing of the original multi-modal breeding data to obtain target multi-modal breeding data includes:
[0009] Perform anomaly detection processing on the original multimodal aquaculture data to obtain multimodal aquaculture data after removing the abnormal data;
[0010] Perform denoising processing on the multimodal aquaculture data after removing the abnormal data to obtain denoised multimodal aquaculture data;
[0011] Extract multimodal aquaculture data that effectively characterizes the health status of animals from the denoised multimodal aquaculture data to obtain target multimodal aquaculture data.
[0012] In one embodiment, the denoised multimodal aquaculture data includes denoised environmental data, denoised physiological data, denoised acoustic data, and denoised image data; the target multimodal aquaculture data includes first time series features and spatial features;
[0013] The extracting multimodal aquaculture data that effectively characterizes the health status of animals from the denoised multimodal aquaculture data to obtain target multimodal aquaculture data includes:
[0014] Use a pre-set sliding window to respectively extract features from the denoised environmental data, denoised physiological data, and denoised acoustic data to obtain the first time series features;
[0015] Use a residual network to extract features from the denoised image data to obtain the spatial features.
[0016] In one embodiment, the disease diagnosis model includes a feature extraction network, a time series modeling network, and a diagnosis prediction network;
[0017] The inputting the target multimodal aquaculture data into a trained disease diagnosis model to extract data features of the target multimodal aquaculture data through the disease diagnosis model, and performing disease diagnosis on the animals in the farm based on the data features to obtain a disease diagnosis result includes:
[0018] Input the target multimodal aquaculture data into the feature extraction network in the trained disease diagnosis model to extract data features of the target multimodal aquaculture data through the feature extraction network;
[0019] Input the data features into the time series modeling network to perform time series modeling on the data features through the time series modeling network to obtain second time series features;
[0020] Input the second time series features into the diagnosis prediction network to perform disease diagnosis on the animals in the farm based on the second time series features through the diagnosis prediction network to obtain a disease diagnosis result.
[0021] In one embodiment, the method further includes:
[0022] Obtain the sample multimodal breeding data collected for a sample breeding farm, and obtain the corresponding reference disease diagnosis results for the sample multimodal breeding data;
[0023] Input the sample multimodal breeding data into the disease diagnosis model to be trained, so as to perform disease diagnosis prediction on the animals in the sample breeding farm based on the sample multimodal breeding data through the disease diagnosis model to be trained, and obtain the predicted disease diagnosis results;
[0024] Based on the reference disease diagnosis results and the predicted disease diagnosis results, determine the classification loss value, regression loss value, and localization loss value respectively;
[0025] Perform weighted summation on the classification loss value, the regression loss value, and the localization loss value to obtain the target loss value;
[0026] Train the disease diagnosis model to be trained based on the target loss value to obtain the trained disease diagnosis model.
[0027] In one embodiment, the environmental data includes temperature, humidity, ammonia concentration, light intensity, and air quality data; the physiological data includes body temperature, heart rate, respiratory data, activity level, and food intake; the image data includes the image data collected by a visible light camera and the image data collected by an infrared thermal imager.
[0028] In a second aspect, the present application provides an animal disease diagnosis device, and the device includes:
[0029] An acquisition module, configured to acquire the original multimodal breeding data collected for a breeding farm; the original multimodal breeding data includes the environmental data of the breeding farm, the physiological data and acoustic data of the animals located in the breeding farm, and the image data collected for the breeding farm;
[0030] A processing module, configured to preprocess the original multimodal breeding data to obtain the target multimodal breeding data;
[0031] A diagnosis module, configured to input the target multimodal breeding data into the trained disease diagnosis model, so as to extract the data features of the target multimodal breeding data through the disease diagnosis model, and perform disease diagnosis on the animals in the breeding farm based on the data features to obtain the disease diagnosis results.
[0032] In a third aspect, the present application provides a computer device, including a memory and a processor, where a computer program is stored in the memory, and when the processor executes the computer program, the steps in the method embodiments of the present application are implemented.
[0033] Fourthly, the present application provides a computer-readable storage medium storing a computer program, which when executed by a processor implements the steps in the method embodiments of the present application.
[0034] Fifthly, the present application provides a computer program product including a computer program, which when executed by a processor implements the steps in the method embodiments of the present application.
[0035] The above-mentioned animal disease diagnosis method, device, equipment, storage medium and program product obtain the original multi-modal breeding data collected for the breeding farm; the original multi-modal breeding data includes the environmental data of the breeding farm, the physiological data and acoustic data of the animals in the breeding farm, and the image data collected for the breeding farm; preprocess the original multi-modal breeding data to obtain the target multi-modal breeding data; input the target multi-modal breeding data into the trained disease diagnosis model, so as to extract the data features of the target multi-modal breeding data through the disease diagnosis model, and diagnose the diseases of the animals in the breeding farm based on the data features to obtain the disease diagnosis result. Compared with the traditional manual animal disease diagnosis method, the present application automatically collects rich multi-modal breeding data of the breeding farm, including the environmental data and image data of the breeding farm, the physiological data and acoustic data of the animals in the breeding farm, inputs the rich multi-modal breeding data into the disease diagnosis model for full integration and analysis, and automatically diagnoses the diseases of the animals in the breeding farm, which can improve the efficiency and accuracy of animal disease diagnosis. Description of the Drawings
[0036] Figure 1 It is an application environment diagram of the animal disease diagnosis method in an embodiment;
[0037] Figure 2 It is a flowchart of the animal disease diagnosis method in an embodiment;
[0038] Figure 3 It is a structural block diagram of the animal disease diagnosis device in an embodiment;
[0039] Figure 4 It is an internal structure diagram of a computer device in an embodiment;
[0040] Figure 5 It is an internal structure diagram of a computer device in another embodiment. Detailed Embodiments
[0041] In order to make the objectives, technical solutions and advantages of the present application clearer, the present application will be further described in detail below with reference to the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.
[0042] The animal disease diagnosis method provided by this application can be applied to, for example, Figure 1 the application environment shown. Among them, the terminal 102 communicates with the server 104 through the network. The data storage system can be set separately and can store the data that the server 104 needs to process. The data storage system can be integrated on the server 104, or can be placed on the cloud or other servers. Among them, the terminal 102 can be, but is not limited to, various desktop computers, laptop computers, smartphones, tablet computers, etc. The server 104 can be an independent physical server, or a server cluster or distributed system composed of multiple physical servers. It can also be a cloud server that provides network security services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, cloud security, host security, CDN, as well as basic cloud computing services such as big data and artificial intelligence platforms. The terminal 102 and the server 104 can be directly or indirectly connected through wired or wireless communication methods, and this application does not make any restrictions here.
[0043] The server 104 can obtain the original multi-modal breeding data collected for the breeding farm; the original multi-modal breeding data includes the environmental data of the breeding farm, the physiological data and acoustic data of the animals located in the breeding farm, and the image data collected for the breeding farm. The server 104 can preprocess the original multi-modal breeding data to obtain the target multi-modal breeding data, and input the target multi-modal breeding data into the trained disease diagnosis model, so as to extract the data features of the target multi-modal breeding data through the disease diagnosis model, and based on the data features, diagnose the diseases of the animals in the breeding farm to obtain the disease diagnosis result, and push the disease diagnosis result to the terminal 102 for display.
[0044] It can be understood that this embodiment does not make any limitations here. It can be understood that Figure 1 the application scenarios in are only for illustrative purposes and are not limited thereto.
[0045] In one embodiment, as Figure 2 shown, a method for diagnosing animal diseases is provided. This method can be applied to a computer device, and the computer device can be a terminal or a server. That is, this method can be executed independently by the terminal or the server, or can be implemented through the interaction between the terminal and the server. This embodiment takes this method applied to a computer device as an example for description, including the following steps:
[0046] Step 202, obtain the original multi-modal breeding data collected for the breeding farm; the original multi-modal breeding data includes the environmental data of the breeding farm, the physiological data and acoustic data of the animals located in the breeding farm, and the image data collected for the breeding farm.
[0047] In one embodiment, the environmental data includes temperature, humidity, ammonia concentration, light intensity, and air quality data; the physiological data includes body temperature, heart rate, respiratory data, activity level, and feed intake; the image data includes the image data collected by a visible light camera and the image data collected by an infrared thermal imager. In this way, through the rich environmental data, physiological data, and image data, the accuracy of animal disease diagnosis can be further improved.
[0048] In one embodiment, the temperature monitoring in the farm uses a PT100 platinum resistance temperature sensor with a measurement accuracy of ±0.1 °C and a measurement range of -50 °C to 100 °C. The humidity monitoring in the farm uses a capacitive humidity sensor with an accuracy of ±2%RH and a response time of <10 seconds. The ammonia concentration in the farm uses an electrochemical sensor with a detection range of 0 - 100 ppm and a resolution of 0.1 ppm. The light intensity in the farm uses a silicon photocell sensor with a measurement range of 0 - 200000 lux. The air quality monitoring in the farm uses an integrated PM2.5 and CO 2 concentration detection unit.
[0049] In one embodiment, the monitoring of the animal's physiological data adopts a wearable design and is fixed on the animal in the form of a smart neck ring or ear tag. Among them, the body temperature monitoring uses a medical-grade infrared temperature sensor with an accuracy of ±0.2 °C. The heart rate monitoring uses a photoplethysmogram sensor with a sampling rate of 200 Hz. The respiratory monitoring uses a piezoelectric sensor that can detect the respiratory frequency and depth. The activity level uses a three-axis acceleration sensor with a sampling rate of 50 Hz. The feed intake records the feeding data of individual animals through a smart feeding system.
[0050] In one embodiment, the visible light camera uses a 4-million-pixel CMOS sensor and is equipped with an auto-zoom lens (2.8 - 12 mm). The infrared thermal imager uses a non-cooled focal plane detector with a temperature resolution of 0.05 °C.
[0051] In one embodiment, the acoustic data is collected by a microphone array. Among them, the microphone array is an 8-channel microphone array with a sampling rate of 48 kHz, 24-bit quantization, beamforming and noise suppression functions, and a pickup range that can cover an area of 20 square meters.
[0052] Step 204, preprocess the original multi-modal farming data to obtain the target multi-modal farming data.
[0053] In one embodiment, preprocessing the original multi-modal farming data can specifically be to perform data cleaning and feature extraction on the collected original multi-modal farming data to obtain the target multi-modal farming data.
[0054] Step 206: Input the target multi-modal breeding data into the trained disease diagnosis model, extract the data features of the target multi-modal breeding data through the disease diagnosis model, and perform disease diagnosis on the animals in the breeding farm based on the data features to obtain the disease diagnosis result.
[0055] In one embodiment, the computer device can input the target multi-modal breeding data into the feature extraction network in the trained disease diagnosis model to extract the data features of the target multi-modal breeding data through the feature extraction network, and input the data features into the diagnosis and prediction network in the trained disease diagnosis model to perform disease diagnosis on the animals in the breeding farm based on the data features through the diagnosis and prediction network, and output the disease diagnosis result.
[0056] In the above animal disease diagnosis method, the original multi-modal breeding data collected for the breeding farm is obtained; the original multi-modal breeding data includes the environmental data of the breeding farm, the physiological data and acoustic data of the animals in the breeding farm, and the image data collected for the breeding farm; the original multi-modal breeding data is preprocessed to obtain the target multi-modal breeding data; the target multi-modal breeding data is input into the trained disease diagnosis model to extract the data features of the target multi-modal breeding data through the disease diagnosis model, and perform disease diagnosis on the animals in the breeding farm based on the data features to obtain the disease diagnosis result. Compared with the traditional manual animal disease diagnosis method, in this application, rich multi-modal breeding data of the breeding farm, including the environmental data and image data of the breeding farm, and the physiological data and acoustic data of the animals in the breeding farm, is automatically collected, and the rich multi-modal breeding data is input into the disease diagnosis model for full integration and analysis to automatically diagnose the diseases of the animals in the breeding farm, which can improve the efficiency and accuracy of animal disease diagnosis. In addition, the labor cost of animal disease diagnosis is also reduced.
[0057] In one embodiment, preprocessing the original multi-modal breeding data to obtain the target multi-modal breeding data includes: performing anomaly detection processing on the original multi-modal breeding data to obtain the multi-modal breeding data after removing the abnormal data; performing denoising processing on the multi-modal breeding data after removing the abnormal data to obtain the denoised multi-modal breeding data; extracting the multi-modal breeding data that effectively represents the health status of the animals from the denoised multi-modal breeding data to obtain the target multi-modal breeding data.
[0058] In one embodiment, the anomaly detection processing uses an improved local outlier factor (LOF) algorithm for outlier detection. The core idea of this algorithm is to identify outliers by comparing the local density of a sample point with other points in its neighborhood. The specific calculation formula is:
[0059] Score(x) = ln(Σ(reach_k(x,o) / reach_k(o,o'))。
[0060] Among them, reach_k represents the k-reachable distance, which reflects the distance metric from point x to its k nearest neighbors. o and o' respectively represent the sample point x and its k nearest neighbor samples. The algorithm first calculates the local reachability density of each sample point, and then determines the degree of abnormality by comparing the ratio of the local reachability density of the sample point to that of its neighborhood points. To improve the adaptability of the algorithm, the k value can be dynamically adjusted according to the characteristics of different types of data. For example, a larger k value (such as k = 20) is used for environmental data to cope with its slow-changing characteristics, while a smaller k value (such as k = 10) is used for physiological data to capture rapid changes.
[0061] In one embodiment, the denoising process uses a denoising method based on wavelet transform. The calculation formula of this method is:
[0062] X_clean = IDWT(Threshold(DWT(X))。
[0063] Among them, DWT represents the discrete wavelet transform, which is used to decompose the signal into different scales; Threshold is a soft threshold function, which is used to remove the noise in the wavelet coefficients; IDWT is the inverse discrete wavelet transform, which is used to reconstruct the signal. In this application, the Daubechies wavelet can be selected as the basis function, and the decomposition level is adaptively adjusted according to the data type: 3-layer decomposition is used for environmental data, and 4-layer decomposition is used for physiological data to retain more detailed information. The threshold selection adopts the SURE (Stein's Unbiased Risk Estimate) criterion, which can retain the characteristics of the original signal to the greatest extent while denoising.
[0064] In the above embodiment, by performing data cleaning operations such as anomaly detection and denoising on the collected original multi-modal aquaculture data, the influence of abnormal data and noise data on subsequent disease diagnosis can be avoided, thereby further improving the accuracy of animal disease diagnosis.
[0065] In one embodiment, the denoised multi-modal aquaculture data includes denoised environmental data, denoised physiological data, denoised acoustic data, and denoised image data; the target multi-modal aquaculture data includes first-order time series features and spatial features; from the denoised multi-modal aquaculture data, multi-modal aquaculture data that effectively characterizes the health status of animals is extracted to obtain the target multi-modal aquaculture data, including: using a pre-set sliding window to extract features from the denoised environmental data, denoised physiological data, and denoised acoustic data respectively to obtain first-order time series features; using a residual network to extract features from the denoised image data to obtain spatial features.
[0066] In one embodiment, in order to capture the change characteristics at different time scales, the present application can maintain multiple sliding windows of different sizes simultaneously. For example, for environmental data, windows with scales of 1 hour, 6 hours, and 24 hours are used. For physiological data, windows with scales of 5 minutes, 30 minutes, and 2 hours are used to balance real-time performance and stability.
[0067] In one embodiment, the residual network can adopt an improved ResNet50 network to extract spatial features. The network structure is specially optimized to adapt to the characteristics of the farm scenario. Among them, the improvements to ResNet50 include: adding an attention module at the front end to enhance the perception of key areas; modifying the convolutional kernel size to adapt to object detection at different scales; adding a spatial pyramid pooling layer to improve the multi-scale expression ability of features; and adopting a transfer learning strategy for network training, first pre-training on a large-scale general dataset and then fine-tuning on the farm scenario data. In order to improve the discriminability of features, a contrastive learning loss is introduced, and the formula is:
[0068] L_contrast = -log(exp(s·cos(f1,f2)) / Σexp(s·cos(f1,fn)).
[0069] Among them, f1 and f2 are the feature views of the same sample from different perspectives, s is the temperature parameter, and cos represents the cosine similarity.
[0070] In the above embodiment, before using the disease diagnosis model to diagnose diseases of animals, first use a sliding window to extract the temporal features of environmental data, physiological data, and acoustic data, and use a residual network to extract the spatial features of image data. These features can better represent the health status of animals and can further improve the accuracy of subsequent disease diagnosis models for diagnosing animal diseases.
[0071] In one embodiment, the disease diagnosis model includes a feature extraction network, a temporal modeling network, and a diagnostic prediction network; input the target multi-modal farming data into the trained disease diagnosis model to extract the data features of the target multi-modal farming data through the disease diagnosis model, and diagnose the diseases of animals in the farm based on the data features to obtain a disease diagnosis result, including: input the target multi-modal farming data into the feature extraction network in the trained disease diagnosis model to extract the data features of the target multi-modal farming data through the feature extraction network; input the data features into the temporal modeling network to perform temporal modeling on the data features through the temporal modeling network to obtain second temporal features; input the second temporal features into the diagnostic prediction network to diagnose the diseases of animals in the farm through the diagnostic prediction network based on the second temporal features to obtain a disease diagnosis result.
[0072] In one embodiment, the feature extraction network can be improved based on the ResNet50 architecture, mainly including:
[0073] a) SE attention module. Add an SE (Squeeze-and-Excitation) module after each residual block, and its mathematical expression is:
[0074] z = Fse(u) = σ(W2δ(W1u)).
[0075] Where u is the input data, W1 and W2 are learnable weight matrices, δ is the ReLU activation function, and σ is the sigmoid function.
[0076] b) Multi-scale feature fusion. Adopt a feature pyramid structure to perform feature fusion at different levels:
[0077] P_l = Conv(C_l + Upsample(P_{l+1})).
[0078] Where P_l represents the feature map of the l-th layer, P_{l+1} is the computed feature map of the (l+1)-th layer (i.e., one layer deeper than the current layer) in the network structure, C_l is the output of the corresponding convolutional layer, Upsample is the upsampling operation, and Conv represents the convolutional operation.
[0079] c) Adaptive pooling strategy. Introduce an adaptive pooling layer to dynamically adjust the spatial resolution of the feature map.
[0080] In one embodiment, the temporal modeling network can adopt an improved bidirectional LSTM network. To improve the effect of temporal modeling, the following improvements are made: a) Introduce a time attention mechanism; b) Design a gated update mechanism.
[0081] In the above embodiment, by setting a feature extraction network in the disease diagnosis model, the input target multi-modal aquaculture data is encoded into data features readable by the model. By setting a temporal modeling network in the disease diagnosis model, the temporal information of the extracted data features is obtained to obtain temporal features, so as to better utilize the multi-modal aquaculture data. Furthermore, through the diagnosis and prediction network in the disease diagnosis model, disease diagnosis is performed based on the temporal features with temporal information, which can further improve the accuracy of animal disease diagnosis.
[0082] In one embodiment, the method further includes: obtaining sample multi-modal breeding data collected for a sample breeding farm, and obtaining a reference disease diagnosis result corresponding to the sample multi-modal breeding data; inputting the sample multi-modal breeding data into a disease diagnosis model to be trained, so as to perform disease diagnosis prediction on animals in the sample breeding farm based on the sample multi-modal breeding data through the disease diagnosis model to be trained, and obtaining a predicted disease diagnosis result; respectively determining a classification loss value, a regression loss value, and a localization loss value based on the reference disease diagnosis result and the predicted disease diagnosis result; performing weighted summation on the classification loss value, the regression loss value, and the localization loss value to obtain a target loss value; and training the disease diagnosis model to be trained based on the target loss value to obtain a trained disease diagnosis model.
[0083] In one embodiment, the target loss value can be calculated by the following formula:
[0084] L_total = α·L_cls + β·L_reg + γ·L_loc.
[0085] The specific forms of each sub-loss function are as follows:
[0086] Classification loss: L_cls = -Σy_i·log(p_i);
[0087] Regression loss: ;
[0088] Localization loss: ;
[0089] Wherein, p_i, are both predicted disease diagnosis results, y_i is the reference disease diagnosis result, α, β, and γ are respectively preset weight coefficients, and L_total represents the target loss value.
[0090] In the above embodiment, the target loss value is obtained by performing weighted summation on the classification loss value, the regression loss value, and the localization loss value, so as to perform multi-task joint training on the disease diagnosis model based on the target loss value, thereby further improving the disease diagnosis accuracy of the disease diagnosis model.
[0091] It should be understood that although the steps in the flowcharts of the above embodiments are shown in sequence, these steps are not necessarily executed in sequence. Unless there is a clear indication in this article, the execution of these steps has no strict sequence limit, and these steps can be executed in other sequences. Moreover, at least a part of the steps in the above embodiments may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed at the same time, but can be executed at different times. The execution sequence of these sub-steps or stages is not necessarily sequential, but can be executed alternately or in turn with at least a part of other steps or sub-steps or stages of other steps.
[0092] In one embodiment, as Figure 3 shown, an animal disease diagnosis device 300 is provided, which specifically includes:
[0093] An acquisition module 302, configured to acquire the original multi-modal breeding data collected for a breeding farm; the original multi-modal breeding data includes the environmental data of the breeding farm, the physiological data and acoustic data of the animals located in the breeding farm, and the image data collected for the breeding farm;
[0094] A processing module 304, configured to preprocess the original multi-modal breeding data to obtain target multi-modal breeding data;
[0095] A diagnosis module 306, configured to input the target multi-modal breeding data into a trained disease diagnosis model, so as to extract the data features of the target multi-modal breeding data through the disease diagnosis model, and perform disease diagnosis on the animals in the breeding farm based on the data features to obtain a disease diagnosis result.
[0096] In one embodiment, the processing module 304 is further configured to perform anomaly detection processing on the original multi-modal breeding data to obtain multi-modal breeding data after removing anomaly data; perform denoising processing on the multi-modal breeding data after removing anomaly data to obtain denoised multi-modal breeding data; extract multi-modal breeding data that effectively characterizes the health status of animals from the denoised multi-modal breeding data to obtain target multi-modal breeding data.
[0097] In one embodiment, the denoised multi-modal breeding data includes denoised environmental data, denoised physiological data, denoised acoustic data, and denoised image data; the target multi-modal breeding data includes first time series features and spatial features; the processing module 304 is further configured to use a preset sliding window to respectively extract features from the denoised environmental data, denoised physiological data, and denoised acoustic data to obtain first time series features; use a residual network to extract features from the denoised image data to obtain spatial features.
[0098] In one embodiment, the disease diagnosis model includes a feature extraction network, a temporal modeling network, and a diagnosis prediction network; the diagnosis module 306 is further configured to input the target multi-modal breeding data into the feature extraction network in the trained disease diagnosis model to extract the data features of the target multi-modal breeding data through the feature extraction network; input the data features into the temporal modeling network to perform temporal modeling on the data features through the temporal modeling network to obtain second temporal features; and input the second temporal features into the diagnosis prediction network to perform disease diagnosis on the animals in the farm based on the second temporal features through the diagnosis prediction network to obtain a disease diagnosis result.
[0099] In one embodiment, the apparatus further includes:
[0100] A training module, configured to obtain the sample multi-modal breeding data collected for a sample farm and obtain the reference disease diagnosis result corresponding to the sample multi-modal breeding data; input the sample multi-modal breeding data into the disease diagnosis model to be trained to perform disease diagnosis prediction on the animals in the sample farm based on the sample multi-modal breeding data through the disease diagnosis model to be trained to obtain a predicted disease diagnosis result; determine a classification loss value, a regression loss value, and a localization loss value respectively based on the reference disease diagnosis result and the predicted disease diagnosis result; perform weighted summation on the classification loss value, the regression loss value, and the localization loss value to obtain a target loss value; and train the disease diagnosis model to be trained based on the target loss value to obtain a trained disease diagnosis model.
[0101] In one embodiment, the environmental data includes temperature, humidity, ammonia concentration, light intensity, and air quality data; the physiological data includes body temperature, heart rate, respiratory data, activity level, and feed intake; and the image data includes the image data collected by a visible light camera and the image data collected by an infrared thermal imager.
[0102] The above animal disease diagnosis device obtains the original multi-modal breeding data collected for a breeding farm; the original multi-modal breeding data includes the environmental data of the breeding farm, the physiological data and acoustic data of the animals located in the breeding farm, and the image data collected for the breeding farm; preprocesses the original multi-modal breeding data to obtain target multi-modal breeding data; inputs the target multi-modal breeding data into a trained disease diagnosis model to extract the data features of the target multi-modal breeding data through the disease diagnosis model, and based on the data features, diagnoses the diseases of the animals in the breeding farm to obtain a disease diagnosis result. Compared with the traditional manual animal disease diagnosis method, this application automatically collects rich multi-modal breeding data of the breeding farm, including the environmental data and image data of the breeding farm, and the physiological data and acoustic data of the animals in the breeding farm, and inputs the rich multi-modal breeding data into the disease diagnosis model for full integration and analysis to automatically diagnose the diseases of the animals in the breeding farm, which can improve the efficiency and accuracy of animal disease diagnosis.
[0103] Each module in the above animal disease diagnosis device can be implemented in whole or in part by software, hardware, and their combination. Each of the above modules can be embedded in or independent of the processor in the computer device in the form of hardware, or stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to each of the above modules.
[0104] In one embodiment, a computer device is provided. The computer device can be a server, and its internal structure diagram can be as Figure 4 shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, abbreviated as I / O), and a communication interface. Among them, the processor, the memory, and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The input / output interface of the computer device is used for the processor to exchange information with external devices. The communication interface of the computer device is used to communicate with external terminals through a network connection. When the computer program is executed by the processor, it implements an animal disease diagnosis method.
[0105] In one embodiment, a computer device is provided. The computer device can be a terminal, and its internal structure diagram can be as Figure 5As shown in the figure. The computer device includes a processor, a memory, an input / output interface, a communication interface, a display unit, and an input device. Among them, the processor, the memory, and the input / output interface are connected through a system bus, and the communication interface, the display unit, and the input device are connected to the system bus through the input / output interface. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The input / output interface of the computer device is used to exchange information between the processor and external devices. The communication interface of the computer device is used to communicate with external terminals in a wired or wireless manner, and the wireless manner can be implemented through WIFI, a mobile cellular network, NFC (Near Field Communication), or other technologies. When the computer program is executed by the processor, it implements an animal disease diagnosis method. The display unit of the computer device is used to form a visually visible picture, which can be a display screen, a projection device, or a virtual reality imaging device. The display screen can be a liquid crystal display screen or an electronic ink display screen. The input device of the computer device can be a touch layer covering the display screen, or a button, a trackball, or a touchpad provided on the computer device housing, or an external keyboard, touchpad, or mouse, etc.
[0106] Those skilled in the art can understand that Figure 4 and Figure 5 the structure shown in the figure is only a block diagram of some structures related to the solution of this application, and does not constitute a limitation on the computer device to which the solution of this application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine some components, or have different component arrangements.
[0107] In one embodiment, a computer device is further provided, including a memory and a processor. A computer program is stored in the memory, and when the processor executes the computer program, the steps in the above method embodiments are implemented.
[0108] In one embodiment, a computer-readable storage medium is provided, storing a computer program, and when the computer program is executed by the processor, the steps in the above method embodiments are implemented.
[0109] In one embodiment, a computer program product is provided, including a computer program, and when the computer program is executed by the processor, the steps in the above method embodiments are implemented.
[0110] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use, and processing of relevant data need to comply with the relevant laws, regulations, and standards of relevant countries and regions.
[0111] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods. Among them, any reference to a memory, storage, database, or other medium used in the embodiments provided in this application can include at least one of non-volatile and volatile memories. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, or optical memory, etc. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc.
[0112] The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity of description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as the scope described in this specification.
[0113] The above-described embodiments only represent several implementation manners of this application. The description is relatively specific and detailed, but it should not be construed as a limitation on the scope of the invention patent. It should be noted that for those of ordinary skill in the art, without departing from the concept of this application, several modifications and improvements can still be made, and these all belong to the protection scope of this application. Therefore, the protection scope of this application patent shall be subject to the appended claims.
Claims
1. A method for diagnosing animal diseases, characterized in that: The method comprises: Acquire original multimodal farming data collected from a farming farm; the original multimodal farming data includes environmental data of the farming farm, physiological data and acoustic data of animals in the farming farm, and image data collected from the farming farm; Preprocessing the original multimodal aquaculture data to obtain target multimodal aquaculture data; The target multimodal breeding data is input into a trained disease diagnosis model to extract data features of the target multimodal breeding data through the disease diagnosis model, and diseases of animals in the farm are diagnosed based on the data features to obtain disease diagnosis results.
2. The method according to claim 1, characterized in that The preprocessing of the original multimodal aquaculture data to obtain target multimodal aquaculture data includes: Performing anomaly detection processing on the original multimodal farming data to obtain multimodal farming data after removing the abnormal data; De-noising the multimodal aquaculture data after removing abnormal data to obtain the de-noised multimodal aquaculture data; From the denoised multimodal breeding data, extract the multimodal breeding data that effectively represents the health status of the animals to obtain the target multimodal breeding data.
3. The method according to claim 2, characterized in that The denoised multimodal aquaculture data includes denoised environmental data, denoised physiological data, denoised acoustic data and denoised image data; the target multimodal aquaculture data includes a first temporal feature and a spatial feature; The method of extracting multimodal breeding data that effectively represents the health status of the animal from the denoised multimodal breeding data to obtain target multimodal breeding data includes: Using a preset sliding window, respectively extracting features from the denoised environmental data, the denoised physiological data and the denoised acoustic data to obtain the first time series features; The residual network is used to perform feature extraction on the denoised image data to obtain the spatial features.
4. The method according to claim 1, characterized in that: The disease diagnosis model includes a feature extraction network, a time series modeling network and a diagnosis prediction network; The step of inputting the target multimodal breeding data into a trained disease diagnosis model to extract data features of the target multimodal breeding data through the disease diagnosis model, and performing disease diagnosis on the animals in the farm based on the data features to obtain disease diagnosis results includes: Inputting the target multimodal aquaculture data into a feature extraction network in a trained disease diagnosis model to extract data features of the target multimodal aquaculture data through the feature extraction network; Inputting the data feature into the timing modeling network to perform timing modeling on the data feature through the timing modeling network to obtain a second timing feature; The second time series feature is input into the diagnosis prediction network, so as to perform disease diagnosis on the animals in the farm based on the second time series feature through the diagnosis prediction network to obtain a disease diagnosis result.
5. The method according to claim 1, characterized in that The method further comprises: Obtaining sample multimodal breeding data collected from sample breeding farms, and obtaining reference disease diagnosis results corresponding to the sample multimodal breeding data; Inputting the sample multimodal breeding data into a disease diagnosis model to be trained, so as to perform disease diagnosis prediction on animals in the sample breeding farm based on the sample multimodal breeding data through the disease diagnosis model to be trained, and obtain a predicted disease diagnosis result; Based on the reference disease diagnosis result and the predicted disease diagnosis result, respectively determining a classification loss value, a regression loss value, and a positioning loss value; Performing a weighted summation on the classification loss value, the regression loss value, and the positioning loss value to obtain a target loss value; The disease diagnosis model to be trained is trained based on the target loss value to obtain a trained disease diagnosis model.
6. The method according to any one of claims 1 to 5, characterized in that The environmental data include temperature, humidity, ammonia concentration, light intensity and air quality data; the physiological data include body temperature, heart rate, respiratory data, activity level and food intake; the image data include image data collected by a visible light camera and image data collected by an infrared thermal imager.
7. An animal disease diagnosis device, characterized in that: The device comprises: An acquisition module, used to acquire original multimodal breeding data collected from a breeding farm; the original multimodal breeding data includes environmental data of the breeding farm, physiological data and acoustic data of animals in the breeding farm, and image data collected from the breeding farm; A processing module, used for preprocessing the original multimodal aquaculture data to obtain target multimodal aquaculture data; A diagnosis module is used to input the target multimodal breeding data into a trained disease diagnosis model so as to extract data features of the target multimodal breeding data through the disease diagnosis model, and perform disease diagnosis on animals in the farm based on the data features to obtain disease diagnosis results.
8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.
9. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.
10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.