Pet state monitoring method and device, electronic equipment and storage medium

Abstract

The embodiment of the invention provides a pet state monitoring method and device, electronic equipment and a storage medium, and relates to the technical field of data processing, and the method comprises the steps: obtaining real-time video data associated with a pet; if the key behavior of the pet is monitored from the real-time video data, determining the pet state of the pet based on the key behavior; acquiring feature data corresponding to the pet; according to the pet state, obtaining weight coefficients corresponding to various feature data; and inputting each feature data and the corresponding weight coefficient into a state monitoring model, outputting a health index for the pet, and executing a processing operation matched with the health index, thereby improving the accuracy and pertinence of health index prediction, and further executing a corresponding processing operation on the pet in time based on the predicted state.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing, in particular to a pet state monitoring method, a pet state monitoring device, an electronic device and a computer readable storage medium. BACKGROUND

[0002] With the improvement of economic level, the proportion of pet-keeping families is increasing. Nowadays, people cannot always accompany pets due to busy life or life, or cannot timely understand the state of pets due to lack of pet-keeping experience, so that when pets suddenly change their living habits due to illness or other conditions, users cannot timely understand, resulting in pets cannot get timely help. SUMMARY

[0003] The embodiments of the present application provide a pet state monitoring method, device, electronic device and computer readable storage medium to solve or partially solve the problem of not being able to timely and accurately predict the state of pets.

[0004] The embodiments of the present application disclose a pet state monitoring method, comprising:

[0005] Obtaining real-time video data associated with a pet;

[0006] If a key behavior of the pet is monitored from the real-time video data, determining a pet state of the pet based on the key behavior;

[0007] Obtaining feature data corresponding to the pet;

[0008] According to the pet state, obtaining a weight coefficient corresponding to each feature data;

[0009] Inputting each feature data and the corresponding weight coefficient into a state monitoring model, outputting a health index for the pet, and performing a processing operation matched with the health index.

[0010] In some feasible implementation manners, the feature data at least includes behavior data, physiological data and environmental data, and the obtaining of the weight coefficient corresponding to each feature data according to the pet state comprises:

[0011] According to the pet state, obtaining a first weight coefficient corresponding to the behavior data, a second weight coefficient corresponding to the physiological data and a third weight coefficient corresponding to the environmental data.

[0012] In some feasible implementation manners, the performing of the processing operation matched with the health index comprises:

[0013] If the health indicator meets a preset health condition, no processing is required.

[0014] If the health indicator does not meet the preset health condition, a response level corresponding to the health indicator is determined, and a prompt operation corresponding to the response level is performed.

[0015] In some possible implementation manners, the response level includes at least a first response level, a second response level, and a third response level, the first response level, the second response level, and the third response level correspond to an impact degree that gradually increases, respectively, and the determination of the response level corresponding to the health indicator includes:

[0016] If the single health indicator does not meet the preset health condition, a first response level is generated.

[0017] If a plurality of health indicators do not meet the preset health condition for a preset time threshold, a second response level is generated.

[0018] If the health indicator indicates that the pet is in a vital sign crisis, a third response level is generated.

[0019] In some possible implementation manners, the performance of the prompt operation corresponding to the response level includes:

[0020] If the first response level is generated, an environment adjustment instruction and diet prompt information are generated, the environment adjustment instruction is used to instruct a smart home device to perform a corresponding temperature and humidity adjustment operation.

[0021] If the second response level is generated, a veterinarian consultation link is sent to a user terminal.

[0022] If the third response level is generated, an emergency prompt message is sent to the user terminal, and a pet hospital is automatically contacted.

[0023] In some possible implementation manners, if the key behavior of the pet is monitored from the real-time video data, the pet state of the pet is determined based on the key behavior, including:

[0024] The pet body corresponding to the pet is separated from the real-time video data.

[0025] The real-time pet behavior of the pet body is acquired.

[0026] If the real-time pet behavior is a key behavior, the resolution of the real-time video data is increased, the process in which the pet performs the key behavior is recorded, and corresponding first video data is obtained.

[0027] frame the first video data to obtain corresponding second video data;

[0028] input the second video data into a convolution network to obtain a pet state of the pet.

[0029] In some possible implementation manners, the frame extraction of the second video data to obtain corresponding second video data includes:

[0030] identify a target pet behavior of the pet in the first video data, the target pet behavior including a static behavior and a dynamic behavior;

[0031] extract a first video frame corresponding to the static behavior from the first video data, and extract a second video frame corresponding to the dynamic behavior from the first video data;

[0032] obtain a first frame rate for the first video frame and a second frame rate for the second video frame, extract a first target video frame from the first video frame according to the first frame rate, and extract a second target video frame from the second video frame according to the second frame rate;

[0033] obtain second video data based on the first target video frame and the second target video frame.

[0034] In some possible implementation manners, the key behavior at least includes one of an abnormal behavior, a health risk behavior, and an emotional behavior.

[0035] The embodiment of the present application further discloses a pet state monitoring device, which comprises:

[0036] a data acquisition module configured to acquire real-time video data associated with a pet;

[0037] a state determination module configured to determine a pet state of the pet based on a key behavior of the pet if the key behavior is monitored from the real-time video data;

[0038] a feature acquisition module configured to acquire feature data corresponding to the pet;

[0039] a weight coefficient acquisition module configured to acquire a weight coefficient corresponding to each of the feature data according to the pet state;

[0040] a health prediction module configured to input each of the feature data and the corresponding weight coefficient into a state monitoring model, output a health index of the pet, and perform a processing operation matched with the health index.

[0041] In some possible implementation manners, the feature data at least includes behavior data, physiological data, and environmental data, and the weight coefficient acquisition module is specifically configured to:

[0042] According to the pet state, a first weight coefficient corresponding to the behavior data, a second weight coefficient corresponding to the physiological data, and a third weight coefficient corresponding to the environmental data are acquired.

[0043] In some possible implementation manners, the health prediction module is specifically configured to:

[0044] If the health index satisfies a preset health condition, no processing is required.

[0045] If the health index does not satisfy the preset health condition, a response level corresponding to the health index is determined, and a prompt operation corresponding to the response level is performed.

[0046] In some possible implementation manners, the response level at least includes a first response level, a second response level, and a third response level, the first response level, the second response level, and the third response level correspond to response levels of the health index, and the health prediction module is specifically configured to:

[0047] If only one of the health indexes does not satisfy the preset health condition, a first response level is generated.

[0048] If a plurality of health indexes do not satisfy the preset health condition for a preset time threshold, a second response level is generated.

[0049] If the health index indicates that the pet is in a life sign critical situation, a third response level is generated.

[0050] In some possible implementation manners, the health prediction module is specifically configured to:

[0051] If the first response level is selected, an environment adjustment instruction and diet prompt information are generated, and the environment adjustment instruction is used to instruct a smart home device to perform a corresponding temperature and humidity adjustment operation.

[0052] If the second response level is selected, a veterinarian consultation link is sent to a user terminal.

[0053] If the third response level is selected, an emergency prompt message is sent to the user terminal, and a pet hospital is automatically contacted.

[0054] In some possible implementation manners, the state determination module is specifically configured to:

[0055] A pet subject corresponding to the pet is separated from the real-time video data.

[0056] acquire a real-time pet behavior of the pet subject;

[0057] if the real-time pet behavior belongs to a key behavior, enhance a resolution of the real-time video data, record a process in which the pet performs the key behavior, and acquire corresponding first video data;

[0058] frame the first video data to obtain corresponding second video data;

[0059] input the second video data into a convolution network to perform prediction, and acquire a pet state of the pet.

[0060] In some possible implementation manners, the state determination module is specifically configured to:

[0061] identify a target pet behavior of the pet in the first video data, the target pet behavior including a static behavior and a dynamic behavior;

[0062] extract a first video frame corresponding to the static behavior from the first video data, and extract a second video frame corresponding to the dynamic behavior from the first video data;

[0063] acquire a first frame rate for the first video frame and a second frame rate for the second video frame, extract a first target video frame from the first video frame according to the first frame rate, and extract a second target video frame from the second video frame according to the second frame rate;

[0064] obtain second video data based on the first target video frame and the second target video frame.

[0065] In some possible implementation manners, the key behavior at least includes one of an abnormal behavior, a health risk behavior, and an emotional behavior.

[0066] An electronic device is also disclosed in the embodiments of the present application, which includes a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory complete communication with each other through the communication bus.

[0067] The memory is configured to store a computer program.

[0068] The processor is configured to execute the program stored on the memory, and implement the method according to the embodiments of the present application.

[0069] A computer readable storage medium is also disclosed in the embodiments of the present application, which stores instructions, and when executed by one or more processors, causes the processor to execute the method according to the embodiments of the present application.

[0070] Embodiments of the present application include the following advantages:

[0071] In embodiments of the present application, by acquiring real-time video data associated with the pet, if it is monitored that the pet has a key behavior from the real-time video data, the pet state of the pet is determined based on the key behavior, then the feature data corresponding to the pet is acquired, and the weight coefficient corresponding to each feature data is acquired according to the pet state, then each feature data and the corresponding weight coefficient are input into the state monitoring model, and the health index of the pet is output, and the processing operation matched with the health index is executed, so that in the process of monitoring the pet state, the behavior of the pet is monitored through the video data, the state corresponding to the key behavior of the pet is identified in the case of monitoring the key behavior of the pet, the state of the pet is preliminarily identified, and in the case of identifying the state of the pet, the health index of the pet is predicted from different dimensions based on multiple feature data, and the corresponding weight coefficient is acquired according to the pet state, so that the prediction process of the health index can be matched with the pet state, the accuracy and pertinence of the health index prediction are improved, and then the pet can be executed corresponding processing operation in time based on the predicted state. BRIEF DESCRIPTION OF DRAWINGS

[0072] Figure 1 is a step flow chart of a pet state monitoring method provided in embodiments of the present application;

[0073] Figure 2 is a schematic diagram of a system architecture provided in embodiments of the present application;

[0074] Figure 3 is a flowchart of health monitoring provided in embodiments of the present application;

[0075] Figure 4 is a structure block diagram of a pet state monitoring device provided in embodiments of the present application. DETAILED DESCRIPTION

[0076] In order to make the above-mentioned purposes, features and advantages of the present application more obvious and easy to understand, the present application will be further described in detail below with reference to the drawings and specific embodiments.

[0077] As an example, in the related pet management system, it often relies on single motion trajectory data to judge the abnormal behavior of the pet, and ignores the influence of other data, resulting in that the prediction result is often one-sided or the false positive rate is high. In addition, in the related art, a static threshold alarm mechanism is adopted, and the dynamic prediction ability of health risk is lacked, for example, when the pet has early signs of disease, the prior art is difficult to capture in time, and the user often cannot perceive until the symptoms worsen. These problems expose the defects of single data dimension and passive response mode.

[0078] To this end, in the present application, by acquiring real-time video data associated with a pet, if it is monitored from the real-time video data that the pet has a key behavior, the pet state of the pet is determined based on the key behavior, then the feature data corresponding to the pet is acquired, and according to the pet state, the weight coefficient corresponding to each of the feature data is acquired, then each of the feature data and the corresponding weight coefficient is input into a state monitoring model, and the health index for the pet is output, and a processing operation matching the health index is performed, so that in the process of monitoring the pet state, the behavior of the pet is monitored through the video data, in the case that the key behavior of the pet is monitored, the corresponding state is identified, the state of the pet is preliminarily identified, and in the case that the state of the pet is identified, the health index of the pet is predicted from different dimensions based on multiple feature data, and at the same time, the corresponding weight coefficient is acquired according to the pet state, so that the prediction process of the health index can be matched with the pet state, the accuracy and pertinence of the health index prediction are improved, and then the corresponding processing operation can be performed on the pet in a timely manner based on the predicted state.

[0079] Referring to Figure 1 , a step flowchart of a pet state monitoring method provided in an embodiment of the present application is shown, which can specifically include the following steps:

[0080] Step 101, acquiring real-time video data associated with a pet;

[0081] In an embodiment of the present application, the state of the pet can be monitored in real time through a corresponding pet management system, which can include intelligent home devices, monitoring devices, and management devices, etc., wherein the intelligent home devices can create a suitable environment for the pet, the monitoring devices can collect relevant indicators of the pet in real time, and the management devices can be responsible for data processing, abnormality prompting, etc.

[0082] In an example, the monitoring device can collect corresponding data, and then transmit the collected data to the management device, and the management device performs corresponding data processing based on the received data, such as predicting the health index of the pet, etc., and then determines what kind of processing operation to perform based on the health index, such as controlling the intelligent home device to create a suitable environment for the pet, or feeding, or sending a corresponding prompt message to the user terminal, so that the user can learn about the state of the pet in a timely manner, etc., which is not limited by the present application.

[0083] In some possible implementation manners, the monitoring device can include a camera with wide-angle shooting and infrared light supplementing function deployed in a home environment, a wearable device for pets (e.g., with a nine-axis motion sensor built-in), and the like. The real-time video data of the pet can be collected through the camera, and the physiological indicators of the pet can be collected through the wearable device. The present application does not make any limitation in this regard.

[0084] In step 102, if the key behavior of the pet is monitored from the real-time video data, the pet state of the pet is determined based on the key behavior.

[0085] After the corresponding real-time video data is collected, the pet management system can monitor the pet behavior of the pet from the real-time video data, and determine whether the pet has a corresponding key behavior, so as to further determine the pet state of the pet based on the key behavior. The key behavior can be a core behavior pattern reflecting the health state or potential risk of the pet. The depth analysis of the video data can determine whether the pet has a corresponding key behavior. The key behavior at least includes one of an abnormal behavior, a health risk behavior, and an emotional behavior. For example, for the abnormal behavior, the excessive scratching and the rapid turning can be included. For the health risk behavior, the vomiting action and the lameness can be included. For the emotional behavior, the anxious barking and the relaxed snoring sound can be included. The present application does not make any limitation in this regard.

[0086] Optionally, for different key behaviors, timely analysis is needed to ensure the health of the pet. For example, for the abnormal behavior (which needs to be immediately focused on), the excessive scratching (which can indicate skin disease or anxiety), the rapid turning (which can be ear infection or nervous abnormality), and the vomiting action (which needs to be combined with the shape of the vomit to determine the cause). For the health risk behavior (which needs to be long-term monitored), the lameness (joint injury or skeletal problem), and the frequent licking of a specific part (pain or inflammation signal). For the emotional behavior (which needs to be associated with environmental factors), the anxious barking (environmental stress or separation anxiety), and the abnormal snoring sound (which can mask the pain signal). Therefore, after the pet has a corresponding key behavior is monitored, the health indicators of the pet can be further analyzed to determine whether the pet needs to be timely handled.

[0087] In some possible implementation manners, the pet subject corresponding to the pet can be separated from the real-time video data first, and then the real-time pet behavior of the pet subject is acquired. If the real-time pet behavior belongs to a key behavior, the resolution of the real-time video data is improved, the process in which the pet performs the key behavior is recorded, the corresponding first video data is obtained, the first video data is frame-extracted, the corresponding second video data is obtained, and finally the second video data is input into a convolution network for prediction to obtain the pet state of the pet. In this way, by classifying the pet subject of the pet from the real-time video data, irrelevant content of the pet behavior in the video data is ignored, the behavior of the pet is focused on, the accuracy of behavior recognition is improved, and in the case where the key behavior of the pet is monitored, the resolution of the video data is improved, the behavior details are recorded, the accuracy of recognition is improved, the video data is further frame-extracted, the computational load is reduced, the feature extraction efficiency is improved, the generalization of model prediction is enhanced, and the analysis accuracy in a complex environment is optimized.

[0088] Optionally, for processing of the video data, the pet subject can be separated by a background modeling technology, and then the behavior of the pet subject is monitored. When it is monitored that the pet has a key behavior, the monitoring device can automatically switch to a high-definition mode to record the corresponding behavior details to obtain the corresponding first video data. In the process of state monitoring, the first video data can be dynamically frame-extracted, so that by using the adaptive frame-extraction algorithm, the frame interval can be adjusted according to the behavior complexity of the pet, and then the computational load in the subsequent data processing process, the feature extraction efficiency, and the model generalization are improved.

[0089] In the frame-extraction process, the target pet behavior of the pet in the first video data can be identified first, the target pet behavior including a static behavior and a motion behavior. Then, the first video frame corresponding to the static behavior is extracted from the first video data, and the second video frame corresponding to the motion behavior is extracted from the first video data. Then, the first frame-extraction rate for the first video frame and the second frame-extraction rate for the second video frame are acquired. The first target video frame is extracted from the first video frame according to the first frame-extraction rate, and the second target video frame is extracted from the second video frame according to the second frame-extraction rate. Finally, the second video data is obtained based on the first target video frame and the second target video frame. In this way, by using the adaptive frame-extraction algorithm, the frame interval can be adjusted according to the behavior complexity of the pet, and then the computational load in the subsequent data processing process, the feature extraction efficiency, and the model generalization are improved.

[0090] For example, first, the pet management system identifies the pet's static behavior (e.g., a cat sleeping for a long time) and motion behavior (e.g., a dog running quickly) from the original video stream (30fps) and extracts the corresponding video frames; then, for static behavior, a low frame rate (e.g., 1fps) is used to reduce redundant data, and for motion behavior, a high frame rate (e.g., 15fps) is used to retain motion details, including extracting 10 frames from a 10-second static segment (300 frames) and 30 frames from a 2-second motion segment (60 frames), and finally using the compressed 40 frames (original 360 frames) for subsequent analysis, thereby reducing the computational load (suitable for edge device deployment) and ensuring that key behavior characteristics (e.g., abnormal posture when in pain or tail whipping frequency when angry) are accurately captured, thereby optimizing the accuracy and real-time performance of pet health monitoring and emotion recognition.

[0091] After the corresponding second video data is extracted, the pet state corresponding to the pet can be determined according to the pet behavior identified from the second video data. Specifically, a semi-supervised learning framework can be used in advance to train a small amount of labeled video and a large amount of unlabeled data to obtain a corresponding convolutional network, while introducing a temporal consistency constraint to ensure that the continuous frame behavior label is logically coherent, for example, the "vomiting" action needs to last more than 3 seconds to trigger a warning, and thus in the prediction process, by inputting the second video data into the corresponding convolutional network for prediction, the pet state corresponding to the current pet behavior of the pet can be obtained.

[0092] In some examples, data collection work is first performed. On the one hand, a certain number of pet videos are collected and the behaviors of the pets in them are labeled, such as different behaviors like "vomiting", "eating", and "sleeping" are explicitly labeled; on the other hand, a large number of unlabeled pet videos are collected, which can enrich the learning samples of the model and enable the model to learn more extensive features. After the collection is completed, data preprocessing is performed, the video is extracted into image frames at a fixed frame rate, for example, 10 frames per second, and then normalization, scaling, and other operations are performed on the extracted image frames to make the size and pixel value range of all images consistent, facilitating subsequent processing, such as scaling all images to 224*224 pixels and normalizing pixel values to the [0, 1] range.

[0093] Then, a suitable convolutional neural network architecture is selected, such as ResNet, etc. After selecting the architecture, the model is adjusted according to the specific pet behavior classification task, such as modifying the last fully connected layer to adapt to the number of classifications. Then, semi-supervised learning algorithms such as Mean Teacher, FixMatch, etc. are used. In this process, the data needs to be divided into labeled data and unlabeled data, and corresponding datasets and data loaders are constructed. Then define the optimizer and loss function, and start the training cycle. In each training cycle, the labeled data is forward propagated to calculate the labeled loss, and the unlabeled data is first forward propagated to obtain the pseudo label, then the unlabeled data is strongly enhanced and forward propagated again to calculate the unlabeled loss, the total loss is obtained by adding the two losses, and finally the back propagation and optimization operation are performed.

[0094] Among them, a time window is defined, and the time is converted into the corresponding frame number according to the frame rate extracted from the previous video frame. For example, set a 3-second time window, if 10 frames are extracted per second, the corresponding frame number is 30 frames. In the prediction process, consecutive frames are checked to ensure that the behavior label is logically consistent. Taking the "vomiting" action as an example, only when the prediction results of the last 30 consecutive frames are all "vomiting", the early warning is triggered.

[0095] Finally, in the actual application process, the second video data to be predicted is preprocessed to extract image frames. These processed image frames are sequentially input into the trained convolutional network for prediction to obtain the prediction result of each frame. Finally, the timing consistency of the prediction result is checked, and for a specific behavior such as "vomiting", the timing window defined earlier is checked, and if the consecutive frame requirement is met, the corresponding early warning is triggered, thereby obtaining the current state of the pet.

[0096] Step 103, obtaining feature data corresponding to the pet;

[0097] After obtaining the pet state corresponding to the pet, the system can further obtain the feature data corresponding to the pet through various monitoring devices. Among them, the feature data can be multi-modal data of the pet, such as video data collected by a camera, physiological data of the pet collected by a wearable device, and environmental data collected by a smart home device, etc. In order to predict the health indicators of the pet based on multi-modal data.

[0098] Step 104, according to the pet state, obtaining the weight coefficient corresponding to each feature data;

[0099] After obtaining multiple types of feature data, weight coefficients corresponding to various feature data can be obtained according to the pet state, so as to predict the health of the pet based on various feature data and corresponding weight coefficients. Among them, when the pet is in different pet states, different weight coefficients can be assigned to different feature data, so as to predict the health indicators of the pet in a targeted manner.

[0100] In some feasible implementations, the weight coefficients corresponding to various feature data are obtained, then the system can obtain the first weight coefficient corresponding to the behavior data, the second weight coefficient corresponding to the physiological data, and the third weight coefficient corresponding to the environmental data according to the pet state, so that through flexible allocation of feature weights, the accuracy of health indicator identification can be improved (based on cross verification between multiple dimensions of features, the misjudgment rate is reduced), the system resource allocation is optimized (the feature data with higher weight coefficients are preferentially processed, the power consumption of the feature data with lower weight coefficients is reduced, and dynamic load balancing is realized), and individual adaptability is enhanced (attention is paid to the features of the pet itself).

[0101] In some examples, assuming that the pet is in an anxious state, the weight of the sound feature can be higher, followed by the motion feature, and the environmental feature can be reduced to the lowest weight, and so on, so as to predict the health indicators of the pet in an anxious state in a targeted manner. For example, for different pet states, the weight coefficients corresponding to various feature data can be as follows:

[0102] For the anxious state, the corresponding weight coefficients can include:

[0103] ① Behavior feature (40%)

[0104] Repetitive pacing (0.5-1.2 Hz frequency)

[0105] Avoidance behavior (hiding duration > 10 minutes)

[0106] ② Sound feature (30%)

[0107] High-frequency barking (> 6000 Hz) or cat's sharp meowing

[0108] Panting sound (> 40 times / minute in non-motion state)

[0109] ③ Physiological data (20%)

[0110] Increased heart rate variability (Heart Rate Variability, HRV) by more than 15%

[0111] Pupil dilation (tracked by eye movement)

[0112] ④ Environmental data (10%)

[0113] Thunderstorm weather (sudden pressure change)

[0114] Stranger visit record

[0115] In addition, dynamic adjustment can also be made: if the continuous anxiety lasts for more than 1 hour, the behavior feature weight is increased to 50%, and the sound feature is reduced to 20%.

[0116] For the state of the pain / discomfort state, the corresponding weight coefficient can include:

[0117] ① Physiological data (45%)

[0118] Local temperature anomaly (infrared thermal imaging temperature difference > 1.5℃)

[0119] Abnormal posture muscle electrical signal (EMG (Electromyography) intensity increased by 2 times)

[0120] ② Behavior characteristics (35%)

[0121] Gait analysis of asymmetric index of lameness

[0122] Frequency of licking specific parts (> 15 times / min)

[0123] ③ Sound characteristics (15%)

[0124] Low-frequency moaning (300-800Hz, sound pressure level <70dB)

[0125] ④ Environmental data (5%)

[0126] Recent fall record (smart home collision detection)

[0127] In addition, dynamic adjustment can also be made: the gait feature weight of large dogs is additionally increased by 5%.

[0128] For the state of the happy / play state, the corresponding weight coefficient can include:

[0129] ① Behavior characteristics (50%)

[0130] Toy throwing action (acceleration > 3m / s 2 )

[0131] Invitation to play posture (forelimb lowered + tail wagging)

[0132] ② Sound characteristics (25%)

[0133] Short barking rhythm (2-4 times / s)

[0134] Cat purring sound (50-150Hz vibration)

[0135] ③ Physiological data (15%)

[0136] Heart rate in the exercise suitable interval

[0137] ④Environmental data (10%)

[0138] Toy appearance signal (RFID tag recognition)

[0139] For the state of depression / depression state, the corresponding weight coefficient can include:

[0140] ①Behavioral characteristics (50%)

[0141] Static duration (> 30 minutes without moving)

[0142] Reject interaction (response delay to the toy > 5 seconds)

[0143] ②Physiological data (30%)

[0144] Reduced basal metabolic rate (20% reduction in activity consumption)

[0145] Sleep cycle disorder (abnormal REM sleep proportion)

[0146] ③Environmental data (15%)

[0147] Insufficient sunlight (< 1 hour / day)

[0148] Owner away duration (> 8 hours / day)

[0149] ④Sound characteristics (5%)

[0150] Call frequency reduced by 50%

[0151] Through the above process, when the pet is in different pet states, different weight coefficients can be assigned to different feature data, so as to predict the health indicators of the pet in a targeted manner.

[0152] Step 105, inputting each of the feature data and the corresponding weight coefficient into a state monitoring model, outputting a health indicator for the pet, and performing a processing operation matched with the health indicator.

[0153] After obtaining the weight coefficients corresponding to various feature data, the various feature data and the corresponding weight coefficients can be input into the state monitoring model to predict the current health indicators of the pet, and then the health indicators are compared with the corresponding health conditions to determine whether the current pet is in an abnormal state, so as to perform a processing operation corresponding to the health indicators, thereby in the process of monitoring the state of the pet, the behavior of the pet is monitored through the video data, and in the case that the key behavior of the pet is monitored, the corresponding state of the pet is identified, the state of the pet is preliminarily identified, and in the case that the state of the pet is identified, the health indicators of the pet are predicted from different dimensions based on various feature data, and the corresponding weight coefficients are obtained according to the state of the pet, so that the prediction process of the health indicators can be matched with the state of the pet, the accuracy and pertinence of the health indicator prediction are improved, and then the corresponding processing operation can be performed on the pet in a timely manner based on the predicted state.

[0154] It should be noted that before the various feature data and the corresponding weight coefficients are input into the state monitoring model, a data alignment operation can be performed first, and different feature data are aligned in the time dimension by establishing a unified time stamp. Alternatively, when the key behavior of the pet is monitored, the time stamp of each device is calibrated based on the time point corresponding to the key behavior, and linear interpolation or Kalman filter compensation is performed on the non-synchronous data segment to realize data alignment.

[0155] In some examples, the system can perform standardization preprocessing on various feature data, convert physiological indicators such as heart rate and body temperature into standard deviation values through the Z-score method, and align data of different sampling frequencies to a 10Hz time axis to ensure consistency in the time dimension. Then, the system can load a preset feature weight configuration, for example, behavior features account for 40% in the anxiety state, sound features account for 30%, etc., and generate a fusion feature vector by weighted summation. When entering the health indicator prediction stage, the system selects a model architecture with different complexities according to the deployment scenario, and the model outputs the prediction probability distribution of various health states after receiving the weighted feature vector, and finally selects the state with the highest probability as the current determination result. The system can have a corresponding health benchmark library built-in, which stores the normal range of various physiological indicators of pets of different breeds, different ages, and different categories, and can also use a dynamic benchmark adjustment algorithm to continuously update individualized health standards, so that after predicting the health indicators of the pet, the health indicators of the pet can be compared with the normal range of physiological indicators to determine whether the pet is in an abnormal state.

[0156] In some possible implementation manners, after the health indicator corresponding to the pet is predicted, the health indicator can be analyzed. If the health indicator meets a preset health condition, no processing is needed. If the health indicator does not meet the preset health condition, a response level corresponding to the health indicator is determined, and a prompt operation corresponding to the response level is performed, so that in the case where it is monitored that the pet is abnormal, a hierarchical response can be performed according to the response level corresponding to the health risk, the adaptability of the response scenario is improved, and effective processing operations can be provided for the pet.

[0157] The response level at least includes a first response level, a second response level, and a third response level. The influence degrees corresponding to the first response level, the second response level, and the third response level are increased in stages, respectively. If a single health indicator does not meet a preset health condition, the first response level is generated. If multiple health indicators do not meet the preset health condition for a preset time threshold, the second response level is generated. If the health indicator indicates that the pet is in a vital sign crisis, the third response level is generated.

[0158] Correspondingly, if the first response level is generated, an environment adjustment instruction and diet prompt information are generated. The environment adjustment instruction is used to instruct the smart home device to perform a corresponding temperature and humidity adjustment operation. If the second response level is generated, a veterinarian consultation link is sent to the user terminal. If the third response level is generated, an emergency prompt message is sent to the user terminal, and a pet hospital is automatically contacted, so that in the case where it is monitored that the pet is abnormal, a hierarchical response can be performed according to the response level corresponding to the health risk, the adaptability of the response scenario is improved, and effective processing operations can be provided for the pet.

[0159] For example, for the first response level (mild abnormality), the trigger condition can be:

[0160] A single health indicator deviates from a normal range (for example, the body temperature is slightly high, and the water consumption is reduced)

[0161] The duration is short (<30 minutes)

[0162] The corresponding response measure can be:

[0163] The environment adjustment instruction can be:

[0164] If it is detected that the body temperature of the pet is high (>39.2℃), the temperature of the smart air conditioner is automatically reduced by 1-2℃

[0165] If the environmental humidity is too low (<40%), the humidifier is started

[0166] The diet prompt information can be:

[0167] If the water intake decreases by 20%, the app will send a push notification reminding the owner to check the water source or provide wet food.

[0168] For the first response level, the applicable scenario can be: the cat's activity decreases due to hot weather, and the body temperature is slightly high (39.3°C) → automatically adjust the room temperature and remind to drink water.

[0169] For the second response level (moderate abnormality), the triggering conditions can be:

[0170] Multiple health indicators are abnormal (such as decreased appetite + reduced activity)

[0171] Lasts more than a preset duration (such as > 2 hours)

[0172] The corresponding response measures can be:

[0173] Push veterinary consultation link:

[0174] Send to the owner's mobile phone: "Your pet has not eaten for more than 2 hours and the activity has decreased by 50%, please consult an online veterinarian immediately

click to contact

[0175] Record abnormal trends:

[0176] Generate a health report, mark the change curve of abnormal indicators (such as eating frequency in the past 24 hours).

[0177] For the second response level, the applicable scenario can be: the old dog has not eaten for two consecutive meals, and the gait is slow → trigger a second-level alarm and recommend professional consultation.

[0178] For the third response level (emergency danger), the triggering conditions can be:

[0179] Critical vital signs (such as heart rate < 40 times / min or respiratory arrest)

[0180] Sudden severe symptoms (such as convulsions, massive vomiting)

[0181] The corresponding response measures can be:

[0182] Emergency prompt message:

[0183] Send a red alert to the owner: "Warning! Your pet's heart rate has dropped to 35 times / min, which may indicate heart failure, please seek medical treatment immediately!"

[0184] Automatically contact the pet hospital:

[0185] Link the intelligent system to dial the phone number of the pet hospital signed, send the pet GPS (Global Positioning System) positioning and health data summary.

[0186] Start emergency measures:

[0187] If equipped with first aid equipment (such as pet oxygen chamber), automatically start standby oxygen supply.

[0188] For the third response level, its applicable scene can be: the dog suddenly falls to the ground, the pupil dilates, and the heart rate monitoring shows ventricular fibrillation → the system directly contacts the nearest hospital and pushes the first aid guide.

[0189] It should be noted that the embodiments of the present application include but are not limited to the above examples, and it can be understood that those skilled in the art can set it according to actual needs, and the present application does not limit it.

[0190] In the embodiments of the present application, by acquiring real-time video data associated with the pet, if it is monitored from the real-time video data that the pet has a key behavior, the pet state of the pet is determined based on the key behavior, then the feature data corresponding to the pet is acquired, and according to the pet state, the weight coefficient corresponding to each feature data is acquired, then each feature data and the corresponding weight coefficient are input into the state monitoring model, the health index for the pet is output, and the processing operation matched with the health index is executed, so that in the process of monitoring the pet state, the behavior of the pet is monitored through the video data, in the case that the key behavior of the pet is monitored, the corresponding state is identified, the state of the pet is preliminarily identified, and in the case that the state of the pet is identified, the health index of the pet is predicted from different dimensions based on multiple feature data, and the corresponding weight coefficient is acquired according to the pet state, so that the prediction process of the health index can be matched with the pet state, the accuracy and pertinence of the health index prediction are improved, and then the pet can be executed corresponding processing operation in time based on the predicted state.

[0191] In order for those skilled in the art to better understand the technical solutions in the embodiments of the present application, some examples are exemplarily illustrated as follows:

[0192] As an example, referring to Figure 2 , a schematic diagram of a system architecture provided in the embodiments of the present application is shown, wherein:

[0193] For the home edge node layer, it can be deployed in the intelligent devices (such as pet cameras, wearable collars) in the user's home, real-time collect multi-modal data (video, physiological signal, environmental parameter), and complete preliminary cleaning and feature extraction locally, ensure that the original data does not leave the local, meet the privacy protection requirements.

[0194] For local data lake, it can be used to store the time series data collected by edge devices (retain 7-30 days rolling window), support fast backtracking analysis. Data is partitioned by behavior, physiology, and environment, and columnar storage is used to improve query efficiency.

[0195] For federated learning cycle, each edge node uploads model parameters (non-original data) through an encrypted tunnel to participate in global model training. Secure aggregation engine is used for homomorphic encryption and multi-party computation of parameters to prevent individual node data from being traced back.

[0196] For local model training cabin, based on incremental learning technology, new data is used to continuously optimize edge lightweight model (such as TinyML architecture), and daily training time is controlled within 15 minutes, and power consumption is less than 2W.

[0197] For global model factory, parameter updates from tens of thousands of edge nodes are aggregated to generate a new generation of global model through FedAvg algorithm, and customized model branches are supported according to varieties, age, etc.

[0198] For intelligent distribution system, the model processed by differential privacy technology is distributed through CDN (Content Delivery Network) network in layers: emergency patch: real-time push (<5 minute delay); regular update: batch execution at night (bandwidth occupancy <1 Mbps / node).

[0199] For self-improvement mechanism, online A / B testing can be realized: 5% of edge nodes run double model comparison to evaluate the effect of new model; and feedback loop: false positive cases marked by users automatically trigger reinforcement learning of corresponding modules

[0200] For secure transmission layer, two-way authentication can be realized: edge nodes and hub use two-way mTLS (Mutual Transport Layer Security) authentication; and encrypted tunnel: parameter transmission uses AES-256-GCM encryption, and key is rotated every 24 hours.

[0201] Correspondingly, based on the above system architecture, the multi-modal perception network is constructed: through dynamic frame rate adjustment and background interference filtering algorithm, accurate behavior capture in complex scene is realized, combined with motion artifact elimination technology, the reliability of physiological signal collection is improved, and the limitation of single data dimension is broken through; cross-modal correlation analysis model: establish the dynamic correlation graph of "behavior-physiology-environment", realize early risk identification through progressive early warning mechanism, upgrade health intervention from passive response to active prevention; adaptive decision system: realize group intelligence optimization under privacy protection by using federated learning framework, combine with user feedback to dynamically correct model parameters, and form the continuously evolving health management ability.

[0202] Specifically, referring to Figure 3 , a flowchart of health monitoring provided in the embodiment of the application is shown, and the process of state monitoring of the pet can include:

[0203] 1. Start stage:

[0204] The system starts and prepares for data collection and analysis.

[0205] 2. Data collection layer:

[0206] Respiratory monitoring: collect the respiratory data of the pet through the sensor.

[0207] Pet wearable device deployment: wear the device to collect data such as motion and activity.

[0208] Visual monitoring device deployment: install visual collection devices such as cameras.

[0209] Camera: used for video data collection.

[0210] Body temperature monitoring: collect the body temperature data of the pet.

[0211] Physiological data collection: obtain physiological indicators such as heart rate.

[0212] Video stream collection: continuously record the activity video of the pet.

[0213] 3. Data processing layer:

[0214] Signal optimization: denoising and enhancement of the collected original signal.

[0215] Preprocessing: data cleaning and formatting processing.

[0216] Pet behavior labeling: label and classify the behavior in the video.

[0217] Machine video data: computer vision processing of video data.

[0218] 4. Model training and analysis layer:

[0219] Model training: Train behavior recognition model using labeled data.

[0220] Multi-source data computation: Integrate data from different sensors.

[0221] Adaptive weights: Dynamically adjust weights of different data sources.

[0222] 5、Output and application layer:

[0223] Recording: Store analysis results and data.

[0224] Hierarchical response: Take different levels of response measures according to analysis results.

[0225] Continuous detection and prediction: Real-time monitoring and prediction of pet status.

[0226] Continuous optimization: System continuously improves algorithm.

[0227] 6、End:

[0228] Process is completed or loops back to start phase to continue monitoring.

[0229] Through the above process, the behavior characteristics, physiological indicators and voiceprint data of the pet can be cross-analyzed to construct a precise health evaluation system, and personalized care suggestions are generated for the pet; at the same time, the intelligent device linkage channel is opened, and when an abnormality is detected, the environmental parameters are automatically adjusted and a warning is pushed, forming a closed-loop management; in addition, the system continuously learns the behavior baseline of different breeds of pets, and adaptively optimizes the model threshold, significantly improving the monitoring efficiency and user experience.

[0230] It should be noted that, for the method embodiments, in order to simply describe, they are all described as a series of action combinations, but those skilled in the art should know that the embodiments of the present application are not limited to the action sequence described, because according to the embodiments of the present application, certain steps can be performed in other sequences or simultaneously. Secondly, those skilled in the art should know that the embodiments described in the specification all belong to preferred embodiments, and the actions involved are not necessarily necessary for the embodiments of the present application.

[0231] Referring to Figure 4 , a structural block diagram of a pet status monitoring device provided in an embodiment of the present application is shown, which can specifically include the following modules:

[0232] The data acquisition module 401 is configured to acquire real-time video data associated with a pet.

[0233] The state determination module 402 is configured to determine a pet status of the pet based on a key behavior of the pet if the key behavior of the pet is monitored from the real-time video data.

[0234] The feature acquisition module 403 is configured to acquire feature data corresponding to the pet;

[0235] The weight coefficient acquisition module 404 is configured to acquire weight coefficients corresponding to various feature data according to the pet state.

[0236] The health prediction module 405 is configured to input the feature data and the corresponding weight coefficients into a state monitoring model, output a health index of the pet, and perform a processing operation matched with the health index.

[0237] In some possible implementation manners, the feature data at least includes behavior data, physiological data, and environmental data, and the weight coefficient acquisition module 404 is specifically configured to:

[0238] According to the pet state, acquire a first weight coefficient corresponding to the behavior data, a second weight coefficient corresponding to the physiological data, and a third weight coefficient corresponding to the environmental data.

[0239] In some possible implementation manners, the health prediction module 405 is specifically configured to:

[0240] If the health index satisfies a preset health condition, no processing is required.

[0241] If the health index does not satisfy the preset health condition, a response level corresponding to the health index is determined, and a prompt operation corresponding to the response level is performed.

[0242] In some possible implementation manners, the response level at least includes a first response level, a second response level, and a third response level, the first response level, the second response level, and the third response level correspond to an impact degree that gradually increases, respectively, and the health prediction module 405 is specifically configured to:

[0243] If a single health index does not satisfy the preset health condition, a first response level is generated.

[0244] If multiple health indexes continuously do not satisfy the preset health condition for a preset time threshold, a second response level is generated.

[0245] If the health index represents that the pet is in a vital sign critical situation, a third response level is generated.

[0246] In some possible implementation manners, the health prediction module 405 is specifically configured to:

[0247] If the first response level is generated, an environmental adjustment instruction and diet prompt information are generated, and the environmental adjustment instruction is used to instruct a smart home device to perform a corresponding temperature and humidity adjustment operation.

[0248] If the second response level is selected, a veterinarian consultation link is sent to the user terminal;

[0249] If the third response level is selected, an emergency prompt message is sent to the user terminal, and a pet hospital is contacted automatically.

[0250] In some possible implementation manners, the state determination module 402 is specifically configured to:

[0251] Separate the pet subject corresponding to the pet from the real-time video data;

[0252] Obtain real-time pet behavior of the pet subject;

[0253] If the real-time pet behavior belongs to a key behavior, the resolution of the real-time video data is increased, the process of the pet performing the key behavior is recorded, and corresponding first video data is obtained;

[0254] Frame extraction is performed on the first video data, and corresponding second video data is obtained;

[0255] The second video data is input into a convolutional network for prediction, and pet state of the pet is obtained.

[0256] In some possible implementation manners, the state determination module 402 is specifically configured to:

[0257] Identify a target pet behavior of the pet in the first video data, the target pet behavior including a static behavior and a motion behavior;

[0258] Extract a first video frame corresponding to the static behavior from the first video data, and extract a second video frame corresponding to the motion behavior from the first video data;

[0259] Obtain a first frame extraction rate for the first video frame and a second frame extraction rate for the second video frame, extract a first target video frame from the first video frame according to the first frame extraction rate, and extract a second target video frame from the second video frame according to the second frame extraction rate;

[0260] Obtain second video data based on the first target video frame and the second target video frame.

[0261] In some possible implementation manners, the key behavior at least includes one of an abnormal behavior, a health risk behavior, and an emotional behavior.

[0262] For the device embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the related parts refer to the part of the method embodiment.

[0263] In addition, the embodiment of the present application also provides an electronic device, comprising a processor, a memory, and a computer program stored in the memory and executable on the processor, wherein the computer program is executed by the processor to implement each process of the embodiment of the pet state monitoring method and achieve the same technical effects. To avoid repetition, details are not described herein.

[0264] The embodiment of the present application also provides a computer readable storage medium, wherein the computer readable storage medium stores a computer program, and the computer program is executed by a processor to implement each process of the embodiment of the pet state monitoring method and achieve the same technical effects. To avoid repetition, details are not described herein. The computer readable storage medium includes a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0265] Each embodiment in the specification is described in a progressive manner, and each embodiment focuses on the difference from other embodiments. The same or similar parts between each embodiment are referred to each other.

[0266] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, device or computer program product. Therefore, the embodiments of the present application can be in the form of a complete hardware embodiment, a complete software embodiment or an embodiment combining software and hardware aspects. Moreover, the embodiments of the present application can be in the form of a computer program product implemented on one or more computer usable storage media (including but not limited to disk storage, CD-ROM, optical storage, EEPROM, Flash and eMMC, etc.) containing computer usable program code.

[0267] The embodiments of the present application are described with reference to flowcharts and / or block diagrams of the method, terminal device (system) and computer program product according to the embodiments of the present application. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, and the combination of the flows and / or blocks in the flowcharts and / or block diagrams can be implemented by computer program instructions. These computer program instructions can be provided to a general-purpose computer, a special-purpose computer, an embedded processor or other programmable data processing terminal device to produce a machine, so that the instructions executed by the computer or other programmable data processing terminal device produce a machine that implements the functions specified in the flowcharts and / or block diagrams. Figure 1 The device that implements the functions specified in one flow or multiple flows and / or blocks Figure 1 The device that implements the functions specified in one flow or multiple flows and / or blocks

[0268] These computer program instructions can also be stored in a computer- readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instructions which implement the flow Figure 1 one or more flows and / or blocks Figure 1 one or more blocks or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instructions which implement the flow

[0269] These computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the flow Figure 1 one or more flows and / or blocks Figure 1 one or more blocks or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instructions which implement the flow

[0270] While preferred embodiments of the application have been described, those skilled in the art will recognize that additional modifications and changes can be made thereto without departing from the scope of the application. It is therefore intended to cover in the appended claims all such changes and modifications that fall within the scope of the application.

[0271] Finally, it should be noted that the terms "first", "second", and the like, herein do not denote any order, quantity, combination, or importance, but rather are used to distinguish one element from another, and do not imply singular or plural. Also, the terms "comprises", "comprising", or other variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but can include other elements not expressly listed or inherent to such process, method, article, or apparatus. An element proceeded by "comprises... a" does not, without more constraints, exclude the existence of additional identical elements in the process, method, article, or apparatus that comprises the element.

[0272] The above describes in detail the pet state monitoring method and pet state monitoring device provided by the present application, and the principles and implementation manners of the present application are described by using specific examples. The above description of the embodiments is only used to help understand the method of the present application and its core idea. Meanwhile, for those skilled in the art, the specific implementation manners and application ranges can be changed according to the idea of the present application. In summary, the content of the present description should not be understood as a limitation of the present application.

Claims

1. A method for monitoring the condition of a pet, characterized in that, include: Acquire real-time video data associated with pets; If key behaviors are detected in the pet from the real-time video data, the pet's status is determined based on the key behaviors. Obtain the feature data corresponding to the pet; Based on the pet's state, obtain the weight coefficients corresponding to various feature data; The feature data and corresponding weight coefficients are input into the status monitoring model, which outputs health indicators for the pet and performs processing operations that match the health indicators.

2. The method according to claim 1, characterized in that, The feature data includes at least behavioral data, physiological data, and environmental data. The step of obtaining weight coefficients corresponding to various feature data based on the pet's state includes: Based on the pet's status, obtain the first weighting coefficient corresponding to the behavioral data, the second weighting coefficient corresponding to the physiological data, and the third weighting coefficient corresponding to the environmental data.

3. The method according to claim 1, characterized in that, The process of performing the operation matching the health indicator includes: If the health indicators meet the preset health conditions, no processing is required; If the health indicator does not meet the preset health conditions, then the response level corresponding to the health indicator is determined, and the prompt operation corresponding to the response level is executed.

4. The method according to claim 3, characterized in that, The response levels include at least a first response level, a second response level, and a third response level, with the degree of influence corresponding to the first response level, the second response level, and the third response level increasing progressively. Determining the response level corresponding to the health indicator includes: If any of the health indicators does not meet the preset health conditions, a first response level is generated; If multiple health indicators fail to meet the preset health conditions for an extended period of time, a second response level is generated. If the health indicators indicate that the pet is in a critical condition, a third response level is generated.

5. The method according to claim 4, characterized in that, The execution of the prompting operation corresponding to the response level includes: If it is the first response level, an environmental adjustment instruction and dietary tips are generated. The environmental adjustment instruction is used to instruct smart home devices to perform corresponding temperature and humidity adjustment operations. If it is the second response level, a veterinary consultation link is sent to the user terminal; If the response level is the third level, an emergency alert message will be sent to the user terminal, and the pet hospital will be contacted automatically.

6. The method according to claim 1, characterized in that, If key behaviors are detected in the pet from the real-time video data, the pet's status is determined based on these key behaviors, including: The pet subject corresponding to the pet is separated from the real-time video data; Obtain the real-time pet behavior of the pet subject; If the real-time pet behavior is a key behavior, then the resolution of the real-time video data is increased, the process of the pet performing the key behavior is recorded, and the corresponding first video data is obtained. Frames are extracted from the first video data to obtain the corresponding second video data; The second video data is input into a convolutional network for prediction to obtain the pet's pet status.

7. The method according to claim 6, characterized in that, The step of extracting frames from the second video data to obtain the corresponding second video data includes: Identify the target pet behavior of the pet in the first video data, the target pet behavior including stationary behavior and moving behavior; Extract the first video frame corresponding to the static behavior from the first video data, and extract the second video frame corresponding to the motion behavior from the first video data; Obtain a first frame extraction rate for the first video frame and a second frame extraction rate for the second video frame, and extract a first target video frame from the first video frame according to the first frame extraction rate, and extract a second target video frame from the second video frame according to the second frame extraction rate. Second video data is obtained based on the first target video frame and the second target video frame.

8. The method according to claim 6 or 7, characterized in that, The key behaviors include at least one of the following: abnormal behaviors, health risk behaviors, and emotional behaviors.

9. A pet status monitoring device, characterized in that, include: The data acquisition module is used to acquire real-time video data associated with pets; The status determination module is used to determine the pet's status based on the key behaviors detected from the real-time video data. The feature acquisition module is used to acquire feature data corresponding to the pet; The weight coefficient acquisition module is used to acquire the weight coefficients corresponding to various feature data based on the pet's state. The health prediction module is used to input the aforementioned feature data and corresponding weight coefficients into the status monitoring model, output health indicators for the pet, and perform processing operations that match the health indicators.

10. An electronic device, characterized in that, It includes a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus; The memory is used to store computer programs; When the processor executes a program stored in the memory, it implements the method as described in any one of claims 1-8.

11. A computer-readable storage medium having instructions stored thereon that, when executed by one or more processors, cause the processors to perform the method as described in any one of claims 1-8.

Images