Condition monitoring method

Abstract

The application provides a state monitoring method, and relates to the technical field of livestock data analysis. The method is applied to a monitoring device, and the method comprises the following steps: performing preprocessing on original data collected by a monitoring device arranged on a target animal to obtain a plurality of detection samples; performing feature extraction based on each detection sample to obtain multi-dimensional feature data; processing the feature data through a preset recognition model to determine state data of the target animal; and regularly uploading the state data to a cloud server for processing. The method is applied to the cloud server, and the method comprises the following steps: acquiring a plurality of state data of the target animal at different times sent by the monitoring device; determining whether the target animal has a state abnormal condition based on the plurality of state data; and if it is determined that the target animal has a state abnormal condition, generating a prompt information.

Description

Technical Field

[0001] This application relates to the field of livestock data analysis technology, and more specifically, to a condition monitoring method. Background Technology

[0002] In large-scale farming, the movement status of farmed animals, such as cattle, directly affects their health and thus farming efficiency. Currently, monitoring the movement status of cattle relies heavily on manual observation or simple sensor data collection, lacking precise algorithmic support. This makes it difficult to accurately identify specific movement states such as lying down, ruminating, eating, walking, getting up, and mounting, leading to misjudgments (e.g., misidentifying ruminating as lying down, or walking as getting up). Current movement detection algorithms are designed for single behaviors, lacking versatility and failing to adapt to the diverse movement states of cattle. Furthermore, they are insufficient in processing noisy data collected by sensors, resulting in low recognition accuracy. The energy consumption required for data transmission during processing is also high, failing to meet the needs of precise monitoring of animal movement status in large-scale farming. Summary of the Invention

[0003] In view of this, the purpose of this application is to provide a state monitoring method to improve the problems of low accuracy and high energy consumption in the existing technology for monitoring the state of animals.

[0004] To address the aforementioned problems, in a first aspect, embodiments of this application provide a state monitoring method, which is applied to a monitoring device, and the method includes: The raw data collected from the monitoring equipment installed on the target animal is preprocessed to obtain multiple test samples; Feature extraction is performed on each of the detected samples to obtain multi-dimensional feature data; The feature data is processed by a preset recognition model to determine the state data of the target animal; The status data is periodically uploaded to a cloud server for processing.

[0005] In the above implementation process, monitoring devices installed on the target animal collect raw data from the animal in real time and preprocess the raw data to obtain multiple valid detection samples at different times. Feature extraction is performed on each detection sample to extract multi-dimensional feature data reflecting the target animal's state. Based on this feature data, a preset recognition model is used to identify the target animal's corresponding state data. The monitoring device can directly process the raw data, obtaining the corresponding short-term activity state data of the target animal through feature extraction and model recognition, effectively improving the accuracy of the state data. The processing requires low energy consumption, and there is no need for direct transmission of the raw data; instead, the processed state data is periodically uploaded to a cloud server for further processing, further reducing the energy consumption required for data transmission. This achieves high-precision and low-power real-time state monitoring, meeting the needs of precise monitoring of animal movement status in large-scale farming.

[0006] Optionally, the training method of the recognition model includes: Based on the historical data of the target animal, the state characteristics of the target animal are determined; Based on the aforementioned state characteristics, the parameters of the initial learning model are optimized to obtain an optimized learning model; Based on the historical data and the corresponding historical state labels, the optimization learning model is trained to obtain the recognition model; wherein, the recognition model includes: a clustering model or a decision tree model.

[0007] In the above implementation process, considering the differences in the states of different animal groups, the state characteristics of the target animal can be determined first based on historical data of similar animals. Based on these state characteristics, the parameters of various initial learning models are optimized to obtain an optimized learning model. Then, based on the historical data of the historical animals and the corresponding historical state labels, the optimized learning model is trained to obtain multiple types of recognition models that can output corresponding motion state labels, achieving real-time state recognition. This allows for model optimization and training based on the actual motion state characteristics of the target animal, effectively improving the accuracy and effectiveness of the recognition model in identifying states.

[0008] Optionally, the method further includes: Determine the individual information of the target animal; Obtain environmental information about the area where the target animal is located; The parameters of the recognition model are adjusted based on the individual information and the environmental information.

[0009] In the above implementation process, considering the individual differences between each target animal, individual information of each target animal and environmental information of the area where the target animal is located can be determined. Based on the individual information and environmental information, the parameters of the recognition model can be personalized and adjusted in real time. This allows for targeted and automatic adjustments to the recognition model based on the individual situation of each target animal and the environmental conditions, effectively improving the targeted recognition accuracy of the model.

[0010] Optionally, when the recognition model is the clustering model, adjusting the parameters of the recognition model based on the individual information and the environmental information includes: Based on the individual information and the environmental information, the cluster centers of the clustering model are moved.

[0011] In the above implementation process, when the identification model is a clustering model, the cluster center of the clustering model can be moved based on individual information and environmental information by utilizing the adaptive adjustment function of the clustering algorithm, so as to reduce the misjudgment rate of the identification model in identifying the target animal.

[0012] Optionally, the step of processing the feature data using a preset recognition model to determine the state data of the target animal includes: The feature data is processed by the recognition model to obtain the state label of the target animal; wherein, the state label includes: lying down, ruminating, eating, walking, getting up or mounting. The state tags are time-fused according to the preset period corresponding to the original data to obtain the state data of the target animal.

[0013] In the above implementation process, feature data can be used as input data for the recognition model, which then outputs corresponding state labels. Furthermore, considering that the original data is collected based on a preset period, to correlate the state with time, the state labels can undergo time fusion processing according to the preset period to obtain state data that can characterize the distribution of motion states over time. By incorporating the influence of time into the processing of state labels, the state data can characterize short-term behavioral state changes within the preset period, thus improving the effectiveness of the state data.

[0014] Optionally, the raw data collected from the monitoring device installed on the target animal is preprocessed to obtain multiple test samples, including: Acquire the raw data collected periodically within a preset period by the acquisition and collection device installed on the target animal; wherein, the raw data includes: multi-axis acceleration data and multi-axis angular velocity data; The raw data is preprocessed to obtain standard data; The time is divided based on the preset period to obtain multiple time windows; The standard data is segmented based on multiple time windows to obtain multiple detection samples.

[0015] In the above implementation process, considering the redundancy of the raw data, to obtain multiple effective detection samples, all raw data within a pre-defined period of regular collection can be acquired first, and then preprocessed to obtain standardized data. Since the pre-defined period is a time interval, to classify data under different time conditions, the time can be divided based on the pre-defined period to obtain multiple time windows. The standard data can then be segmented based on these multiple time windows to obtain multiple effective detection samples under different time windows. This approach enables data preprocessing, time segmentation, and data segmentation to improve the effectiveness and temporal characteristics of the detection samples.

[0016] Optionally, the preprocessing of the raw data to obtain standard data includes: The original data is denoised using a filtering algorithm to obtain denoised data. The denoised data is standardized to obtain the standard data.

[0017] In the above implementation process, the original data can be denoised using a filtering algorithm to remove noise and errors, resulting in denoised data. This denoised data is then standardized to eliminate the influence of dimensions, yielding standard data that is easier to analyze and process later. This approach enables denoising and standardization of the original data, improving the effectiveness and format consistency of the standard data.

[0018] Optionally, the method further includes: Based on historical data of historical animals, the state characteristics of the target animal are determined; Based on the aforementioned state characteristics, the preset period is determined.

[0019] In the above implementation process, considering the differences in the state of different animal groups, which will affect the monitoring cycle, the state characteristics of the target animal can be determined first based on historical data of historical animals similar to the target animal. Based on the state characteristics, a preset monitoring cycle for the target animal can be determined. Data collection and state recognition processing of the target animal can then be performed based on the preset cycle to achieve the effect of identifying the short-term behavior of the target animal. Based on the preset cycle, the state data is periodically uploaded to the cloud server for processing, which effectively reduces the energy consumption required to directly upload the raw data and extends the battery life of the monitoring equipment.

[0020] Optionally, the feature extraction based on each of the detected samples to obtain multi-dimensional feature data includes: Temporal feature extraction is performed on each of the detected samples to obtain temporal features; Perform a Fourier transform on each of the detected samples to obtain frequency domain features; The time-domain features and the frequency-domain features are fused to obtain multi-dimensional fused features; Based on the state characteristics of the target animal, the fused features are filtered to obtain multi-dimensional feature data.

[0021] In the above implementation process, when processing each detection sample, time-domain and frequency-domain features can be extracted separately, and then fused to obtain multi-dimensional fused features. Considering the differences in the state of different animal groups, the fused features can be filtered based on the state characteristics of the target animal to reduce the adverse effects of insensitive features on state recognition, thus obtaining multi-dimensional feature data that contributes significantly to state recognition. This capability to perform feature extraction, fusion processing, and filtering improves the effectiveness and conciseness of the feature data.

[0022] Secondly, embodiments of this application also provide a status monitoring method, which is applied to a cloud server and includes: Acquire multiple status data of the target animal at different times, sent by the monitoring equipment; Based on multiple states of data, determine whether the target animal exhibits any abnormal state. If the target animal is determined to be in an abnormal state, a prompt message will be generated.

[0023] In the above implementation process, considering the large energy consumption required for long-term behavior analysis and evaluation based on multiple short-term state data, the cloud server can evaluate and analyze the long-term behavior of the target animal based on the received multiple state data to determine whether there are any abnormal conditions in the target animal during the breeding process. When it is determined that there are abnormal conditions in the target animal, corresponding prompt information is generated in real time to prompt the staff to check, adjust and handle the abnormality, reduce the adverse effects of the abnormal condition on the target animal, and improve the economic benefits of breeding.

[0024] Thirdly, embodiments of this application also provide an electronic device, which includes a memory and a processor. The memory stores program instructions, and when the processor executes the program instructions, it performs the steps in the method described in any one of the first or second aspects above.

[0025] Fourthly, embodiments of this application also provide a computer program product, the computer program product including a computer program / instruction, which, when executed by a processor, implements the steps in the method described in any one of the first or second aspects above.

[0026] In summary, the embodiments of this application provide a state monitoring method that enables monitoring equipment to directly process raw data and obtain the corresponding short-term activity state data of the target animal through feature extraction and model recognition. This effectively improves the accuracy of the state data, requires less energy for processing, and eliminates the need for direct transmission of raw data. Instead, the processed state data is periodically uploaded to a cloud server for further processing, further reducing the energy consumption required for data transmission. This achieves high-precision and low-power real-time state monitoring, meeting the needs of precise monitoring of animal movement status in large-scale farming. Attached Figure Description

[0027] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments of this application will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0028] Figure 1 A block diagram illustrating an electronic device provided in an embodiment of this application; Figure 2 A flowchart illustrating the first state monitoring method provided in this application embodiment; Figure 3 A flowchart illustrating the second state monitoring method provided in this application embodiment; Figure 4 A flowchart illustrating the third state monitoring method provided in this application embodiment; Figure 5 A detailed flowchart of step S400 provided for an embodiment of this application; Figure 6 A detailed flowchart of step S200 provided for an embodiment of this application; Figure 7 A detailed flowchart of step S220 provided for an embodiment of this application; Figure 8 A flowchart illustrating the fourth state monitoring method provided in this application embodiment; Figure 9 A detailed flowchart of step S300 provided for an embodiment of this application; Figure 10This is a flowchart illustrating the fifth state monitoring method provided in the embodiments of this application.

[0029] Icons: 100 - Electronic device; 111 - Memory; 112 - Memory controller; 113 - Processor; 114 - Peripheral interface; 115 - Input / output unit; 116 - Display unit. Detailed Implementation

[0030] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of the embodiments of this application.

[0031] The movement status of farmed animals directly affects the health of cattle, thus impacting farming efficiency. Currently, monitoring the movement status of cattle relies heavily on manual observation or simple sensor data collection, lacking precise algorithmic support. This makes it difficult to accurately identify specific movement states such as lying down, ruminating, eating, walking, getting up, and mounting, leading to misjudgments (e.g., misidentifying ruminating as lying down, or walking as getting up). Current movement detection algorithms are designed for single behaviors, lacking versatility and failing to adapt to the diverse movement states of cattle. Furthermore, they are insufficient in processing noisy data collected by sensors, resulting in low recognition accuracy. The energy consumption required for data transmission during processing is also high, failing to meet the needs of precise monitoring of animal movement status in large-scale farming.

[0032] To address the aforementioned issues, this application provides a status monitoring method that enables monitoring equipment to directly process raw data. Through feature extraction and model recognition, it obtains the status data of the target animal's short-term activities, effectively improving the accuracy of the status data. The method requires less energy during processing and eliminates the need for direct transmission of raw data. Instead, the processed status data is periodically uploaded to a cloud server for further processing, further reducing the energy consumption required for data transmission. This achieves high-precision and low-power real-time status monitoring, meeting the needs of precise monitoring of animal movement status in large-scale farming.

[0033] This application provides a status monitoring method applied to an electronic device. The electronic device may include a monitoring device installed on a target animal and a cloud server that is communicatively connected to the monitoring device. The method is capable of data collection and processing to monitor the short-term and long-term behaviors of the target animal in real time.

[0034] Optionally, please refer to Figure 1 , Figure 1This is a block diagram illustrating an electronic device according to an embodiment of this application. The electronic device 100 may include a memory 111, a memory controller 112, a processor 113, a peripheral interface 114, an input / output unit 115, and a display unit 116. Those skilled in the art will understand that... Figure 1 The structure shown is for illustrative purposes only and does not limit the structure of the electronic device 100. For example, the electronic device 100 may also include components that are more... Figure 1 The more or fewer components shown, or having the same Figure 1 The different configurations shown.

[0035] The aforementioned memory 111, memory controller 112, processor 113, peripheral interface 114, input / output unit 115, and display unit 116 are electrically connected directly or indirectly to each other to achieve data transmission or interaction. For example, these components can be electrically connected to each other through one or more communication buses or signal lines. The aforementioned processor 113 is used to execute executable modules stored in the memory.

[0036] The memory 111 can be, but is not limited to, Random Access Memory (RAM), Read Only Memory (ROM), Programmable Read-Only Memory (PROM), Erasable Programmable Read-Only Memory (EPROM), Electrically Erasable Programmable Read-Only Memory (EEPROM), etc. The memory 111 stores programs. After receiving execution instructions, the processor 113 executes the programs. The methods executed by the electronic device 100 as defined in any embodiment of this application can be applied to the processor 113, or implemented by the processor 113.

[0037] The aforementioned processor 113 may be an integrated circuit chip with signal processing capabilities. The processor 113 may be a general-purpose processor, including a Central Processing Unit (CPU), a Network Processor (NP), etc.; it may also be a digital signal processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field-Programmable Gate Array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this application. The general-purpose processor may be a microprocessor or any conventional processor.

[0038] The peripheral interface 114 described above couples various input / output devices to the processor 113 and the memory 111. In some embodiments, the peripheral interface 114, the processor 113, and the memory controller 112 can be implemented on a single chip. In other instances, they can be implemented on separate chips.

[0039] The input / output unit 115 described above is used to provide user input data. The input / output unit 115 can be, but is not limited to, a mouse and keyboard, etc.

[0040] The aforementioned display unit 116 provides an interactive interface (e.g., a user interface) between the electronic device 100 and the user, or displays image data for the user's reference. In this embodiment, the display unit can be a liquid crystal display (LCD) or a touch display. If it is a touch display, it can be a capacitive touchscreen or a resistive touchscreen that supports single-point and multi-point touch operations. Supporting single-point and multi-point touch operations means that the touch display can sense touch operations generated simultaneously from one or more locations on the touch display and pass the sensed touch operations to the processor for calculation and processing. In this embodiment, the display unit 116 can display various information such as obtained status data, abnormal status conditions, and prompt information.

[0041] The electronic device in this embodiment can be used to execute various steps in the various state monitoring methods provided in the embodiments of this application. The implementation process of the state monitoring method is described in detail below through several embodiments.

[0042] Please see Figure 2 , Figure 2 This is a flowchart illustrating a first state monitoring method provided in an embodiment of this application, applied to a monitoring device. The method may include steps S200-S500.

[0043] Step S200: Preprocess the raw data collected from the monitoring equipment set on the target animal to obtain multiple test samples.

[0044] The monitoring equipment installed on the target animal collects raw data of the target animal in real time and preprocesses the raw data to obtain valid test samples at multiple different times.

[0045] Optionally, the monitoring device installed on the target animal can be a monitoring collar or similar device worn by the target animal. The monitoring collar integrates multiple modules such as a food intake monitoring unit, an activity monitoring unit, a data transmission unit, and a positioning unit, which can collect the target animal's raw data in real time.

[0046] Optionally, the target animal can be a variety of domesticated animals, such as cattle or sheep.

[0047] Step S300: Feature extraction is performed based on each detection sample to obtain multi-dimensional feature data.

[0048] In this process, feature extraction is performed on each test sample to extract multi-dimensional feature data that can reflect the state of the target animal.

[0049] It should be noted that the collection time for each test sample is different. Therefore, multiple test samples can reflect the characteristics of multiple consecutive time periods.

[0050] Step S400: The feature data is processed by a preset recognition model to determine the state data of the target animal.

[0051] Among them, it can identify and process feature data based on a preset recognition model to determine the state data corresponding to the target animal.

[0052] It should be noted that status data can include the status and the corresponding time information. For example, at a certain moment or during a certain time period, the target animal is in a feeding state.

[0053] Step S500: The status data is periodically uploaded to the cloud server for processing.

[0054] Among them, the monitoring equipment directly processes the raw data and obtains the corresponding short-term activity status data of the target animal through feature extraction and model recognition, which effectively improves the accuracy of the status data. The energy consumption required for processing is small, and there is no need to directly transmit the raw data. Instead, the processed status data is periodically uploaded to the cloud server for processing, which further reduces the energy consumption required for data transmission.

[0055] exist Figure 2 The embodiments shown can achieve high-precision and low-power real-time status monitoring, meeting the needs of precise monitoring of animal movement status in large-scale farming.

[0056] Optionally, please refer to Figure 3 , Figure 3 This is a flowchart illustrating a second state monitoring method provided in an embodiment of this application. The method may further include steps S611-S613.

[0057] Step S611: Determine the state characteristics of the target animal based on historical data of historical animals.

[0058] Considering the differences in the state of different groups of animals, the state characteristics of the target animal can be determined first based on historical data of historical animals similar to the target animal.

[0059] Optionally, state characteristics may include the specificity of the target animal for different movement states. For example, when the target animal is a cow, state characteristics may include the cow's tendency to ruminate for a period of time after eating (such as 2 hours), the cow's walking range, the cow's diurnal rhythm, and other characteristics.

[0060] Step S612: Based on the state characteristics, optimize the parameters of the initial learning model to obtain the optimized learning model.

[0061] Among these, the parameters of the initial learning model can be optimized based on the state characteristics to obtain an optimized learning model.

[0062] Optionally, the initial learning model may include various machine learning models such as decision trees and clustering models (e.g., K-clustering models). Some parameters in the model can be adjusted according to the state characteristics of the target animal to make the optimized learning model more suitable for state recognition of the target animal.

[0063] Step S613: Based on historical data and corresponding historical state labels, train the optimization learning model to obtain the recognition model.

[0064] Among them, an optimization learning model can be trained based on historical data of historical animals and historical state labels corresponding to the historical data to obtain a recognition model that can output corresponding motion state labels and realize real-time state recognition.

[0065] Optionally, since the recognition model is obtained by optimizing and training an initial learning model, the recognition model can also include various machine learning models such as decision trees and clustering models (such as K-clustering models).

[0066] Optionally, various motion state samples of different historical animals of different breeds, weights, and ages can be obtained from historical databases as historical data. The historical data is divided into training set, validation set, and test set in a 7:2:1 ratio. The training set is used to train the optimization learning model, and the validation set is used to adjust the parameters of the optimization learning model. The model outputs corresponding motion state labels, and the output motion state labels are compared with the corresponding historical state labels to determine the recognition accuracy. Iterative training is performed until the optimization learning model converges. The convergence condition can be set to test set accuracy ≥ 80%.

[0067] Alternatively, different historical data from different ranches can be selected to validate the model.

[0068] exist Figure 3 In the illustrated embodiment, the model can be optimized and trained based on the characteristics of the target animal's actual movement state, effectively improving the accuracy and effectiveness of the recognition model in recognizing the state.

[0069] Optionally, please refer to Figure 4 , Figure 4 This is a flowchart illustrating a third state monitoring method provided in an embodiment of this application. The method may further include steps S614-S616.

[0070] Step S614: Determine the individual information of the target animal.

[0071] Step S615: Obtain environmental information about the area where the target animal is located.

[0072] In this regard, taking into account the individual differences between each target animal, the individual information of each target animal and the environmental information of the area where the target animal is located can be determined.

[0073] Optionally, corresponding environmental information can be obtained from the positioning unit in the monitoring equipment by associating it with the environmental monitoring equipment. The environmental information may include temperature, humidity, and data such as temperature change trends and humidity change trends.

[0074] Optionally, the database can be queried based on the target animal's identity information (such as number, barcode, etc.) to obtain the target animal's individual information, which may include the target animal's breed, sex, age, and other information.

[0075] Step S616: Adjust the parameters of the recognition model based on individual information and environmental information.

[0076] Among these features, the parameters of the recognition model can be personalized and adjusted in real time based on individual and environmental information.

[0077] Optionally, when the recognition model is a clustering model, adjusting the parameters of the recognition model based on individual and environmental information can include moving the cluster centers of the clustering model based on individual and environmental information. When adjusting parameters, the adaptive adjustment function of the clustering algorithm can be used to move the cluster centers of the clustering model based on individual and environmental information, thereby reducing the misclassification rate of the recognition model when identifying the target animal.

[0078] For example, a dynamic threshold adjustment mechanism can be set up to automatically adjust the recognition threshold of the clustering model based on the individual and environmental information of the target animal. The specific adjustment logic may include: adaptively adjusting the position of each cluster center using the K-clustering algorithm. When each new data set is classified into the range of that cluster center, the cluster center is moved a small step toward that data, with a step size of 0.001.

[0079] It should be noted that the decision tree model uses the sum-x_avg features for analysis, while the clustering model uses x_std, y_std, and z_std, resulting in two classified examples. Alternatively, the three largest components after the Fourier transform of the original data can be used, or r=sqrt( The mean, standard deviation, kurtosis, and skewness of the data are used to study and derive new classification methods. This application can process the data using either an acceleration correlation algorithm or a triaxial angular velocity algorithm.

[0080] Optionally, when the recognition model is a decision tree model, through observation, it is found that the activity level has the closest relationship with sum=x_std+y_std+z_std, and the relationship with x_avg is the second closest. Therefore, on the x_avg-sum plane, the parameter space can be divided according to the parameters given by the decision tree, and each block of the parameter space is marked with an activity level. By merging the parameter spaces with the same activity level, the parameter space and threshold corresponding to each activity level (target animal behavior) can be obtained. Accordingly, the following thresholds can be roughly obtained: 0-1 (here 0 and 1 are similar to the activity levels 1 and 2 in the following table, which are abbreviations, and can be roughly determined as lying without ruminating) Lying without ruminating: sum<75; 2-4 (here 2, 3, 4 are similar to the activity levels 3, 4, 5 in the following table, which are abbreviations, and can be roughly determined as ruminating, that is, the following three behaviors) Lying ruminating, standing without ruminating, standing ruminating: 75<sum<145, or 145<sum<230, and |x_avg|<470; 5 (here 5 is similar to the activity level 6 in the following table, which is an abbreviation, and can be roughly determined as feeding) Feeding: sum>145 and |x_avg|>470 or sum>230 and 215<|x_avg|<470; 6-9 (here 6, 7, 8, 9 are similar to the activity levels 7, 8, 9, 10 in the following table, which are abbreviations, and can be roughly determined as the three behaviors of walking, lying down and standing up, and mounting. In some cases, the three behaviors are difficult to distinguish, so they can be classified into the same category) Walking: sum>230 and |x_avg|<215. The specific value can be selected based on the accuracy of historical situations. After obtaining these thresholds, for any point (sum, x_avg), the corresponding activity level (target animal behavior) can be determined. 1440 pieces of data are generated every 2 hours (one every 5 seconds), and these 1440 pieces of data are filled into the corresponding categories according to the intervals to obtain a histogram distribution, producing an effect similar to the following: Lying without ruminating: 538 pieces, corresponding to 44.8min; Lying ruminating, standing without ruminating, standing ruminating: 398 pieces, corresponding to 33.2min; Feeding: 256 pieces, corresponding to 21.3min; Walking: 248 pieces, corresponding to 20.7min. By determining the time proportion of the target animal in each activity level interval within these 2 hours, that is, within the preset period, the activity pattern map of the target animal within these 2 hours can be obtained. For the decision tree algorithm, the following data can be referred to when the wearing directions of the monitoring devices are consistent (the above data and the following data are essentially similar, except that the above data assumes that the workers wearing the devices cannot strictly follow the direction, and there is a situation where the devices are worn reversely, while the following data assumes that all devices can be worn strictly and consistently.The embodiments of the present application show different parameters under two hypothetical wearing modes, for different practical application scenarios): 012 (here 0, 1, 2 are similar to the activity levels 1, 2, 3 in the following table, as abbreviations, which can be roughly determined as lying down): sum<50 or 50<sum<315 and x_avg>0; 34 (here 3, 4 are similar to the activity levels 4, 5 in the following table, as abbreviations, which can be roughly determined as ruminating): 50<sum<185 and x_avg<0 or 185<sum<315 and -470<x_avg<0; 5 (here 5 is similar to the activity level 6 in the following table, as an abbreviation, which can be roughly determined as eating): sum>185 and x_avg<-470; 6789 (here 6, 7, 8, 9 are similar to the activity levels 7, 8, 9, 10 in the following table, as abbreviations, which can be roughly determined as the three behaviors of walking, lying down and standing up, and mounting. For the decision tree model, these three behaviors are difficult to distinguish, so they are classified into the same category): sum>315 and x_avg>-470. The above classification method can be obtained through simulation tests, with an accuracy rate of 0.8554, a variance of 0.0171, and a bias of 0.0018.

[0081] Optionally, for the k-means clustering method, through analysis it can be known that the activity level has the closest relationship with x_std, y_std, z_std, so the clustering method can be used for classification. The simplest way is to average all data by activity level, and data similar to the following table can be obtained:

[0082] It is evident that clusters 1-4 can be considered as one class, 5-8 as another, and 9 and 10 as two separate classes. Therefore, the following cluster centers can be roughly selected: 1-2 lying down (0, 0, 0); 3-4 ruminating (40, 40, 40); 5-6 eating (95, 95, 95); 7-8 walking (150, 150, 150); 9 getting up and lying down (240, 240, 240); 10 mounting (320, 320, 320). These parameters are best practice parameters selected based on actual historical data. Thus, for any (x_std, y_std, z_std), its distance to the above four centers can be calculated, and it can then be classified into the class with the closest distance. For example, the distance from (128, 44, 19) to (10, 10, 10) is 123.13, the distance to (45, 45, 45) is 86.98, the distance to (85, 85, 85) is 88.8, and the distance to (135, 135, 135) is 147.6. Therefore, it is classified as rumination (45, 45, 45). Using this method, we can also obtain the number of data points for each behavior, and then, similar to decision trees, obtain the percentage of time the target animal spends in each activity interval within those 2 hours. This gives us a map of the target animal's activity over those 2 hours. It should be noted that specific parameters can be adjusted when uploading data every 2 hours. For example, for the k-means clustering algorithm, we can average all points in each category and then slightly shift the old center value towards the new center value. For example, if the old center values ​​are denoted as c1:(10, 10, 10), c2:(45, 45, 45), c3:(85, 85, 85), and c4:(135, 135, 135), after one hour, there are 538 points classified as c1 with an average of m1:(15, 13, 22); 398 points classified as c2 with an average of m2:(53, 40, 55); 256 points classified as c3 with an average of m3:(90, 79, 101); and 248 points classified as c4 with an average of m4:(155, 145, 162). Then, the new center point can be defined as Ci' = ci + 0.0001 × (mi - ci), where 0.0001 is called the learning rate. Therefore, the new center points are c1'=(10.05, 10.03, 10.12), c2'=(45.08, 44.95, 45.1), c3'=(85.05, 84.94, 85.16), and c4'=(135.2, 135.1, 135.27). Thus, the recognition model has different algorithm parameters for each target animal, reflecting the individual characteristics of each animal, such as different sizes and temperaments (active or quiet), thereby adaptively determining accurate state data for each target animal.

[0083] existFigure 4 In the illustrated embodiment, the recognition model can be automatically adjusted to suit the individual circumstances and environmental conditions of each target animal, effectively improving the targeted recognition accuracy of the recognition model.

[0084] Optionally, please refer to Figure 5 , Figure 5 This is a detailed flowchart of step S400 provided in an embodiment of the present application. Step S400 may include steps S410-S420.

[0085] Step S410: The feature data is processed by the recognition model to obtain the status label of the target animal.

[0086] In this approach, feature data can be used as input data for the recognition model, and the recognition model can output the corresponding status label.

[0087] Optionally, the status tags may include tags related to the target animal's movement status, such as lying down, ruminating, eating, walking, getting up or mounting.

[0088] For example, the recognition of different state labels can include: 1. Lying down state recognition: At this time, the target animal is very calm, the sensor fluctuations in the monitoring device are very small, and the monitoring device is in a dormant state; Judgment condition: the sum of the standard deviations of the three-axis accelerations is <75, and the recognition accuracy is 80.14%. 2. Rumination state recognition: At this time, the target animal is trembling slightly, and the head is usually in a normal position; Judgment condition: the sum of the standard deviations of the three-axis accelerations is [75, 145], or the sum of the standard deviations of the three-axis accelerations is between [145, 230], and the absolute value of the x-axis mean is less than 470, and the recognition accuracy is 84.37%. 3. Feeding state recognition: At this time, the target animal lowers its head, resulting in a larger gravitational component on the x-axis and more vigorous movement; Judgment condition: the sum of the standard deviations of the three-axis accelerations is greater than 145 and the absolute value of the x-axis mean is greater than 470, and the recognition accuracy is 71.53%. 4. Walking state recognition: The target animal's head is in a normal position, but the range of motion is large. Judgment criteria: The sum of the standard deviations of the three axes of acceleration is greater than 230 and the average value of the x-axis is less than 470. The recognition accuracy is 50.88%. 5. Standing / Lying state recognition: The target animal's movement is relatively vigorous, and the sum of the standard deviations of the three axes is approximately 750. Judgment criteria: The distances of the standard deviations of the x, y, and z axes of acceleration to (250, 250, 250) are less than the distances to (150, 150, 150) and (320, 320, 320). The recognition accuracy is 62.31%. 6. Climbing / Mounting state recognition: The target animal's movement is abnormally vigorous, and the standard deviations of the three axes fluctuate wildly. Judgment criteria: The distances of the standard deviations of the x, y, and z axes of acceleration to (320, 320, 320) are less than the distances to (250, 250, 250). The recognition accuracy is 90.53%.

[0089] Step S420: Perform time fusion processing on the status labels according to the preset period corresponding to the original data to obtain the status data of the target animal.

[0090] In order to associate the state with time, the state labels can be processed by time fusion according to the preset period to obtain state data that can characterize the distribution of motion state over time.

[0091] Optionally, status tags can be associated with time nodes within a preset period to achieve time fusion processing. The status data can include the association between multiple sets of status tags and time nodes. For example, during the time period from 12:03 on July 7, 2026 to 12:45 on July 7, 2026, the status tag is "lying down". The status data can also include the time percentage of each status tag within the preset period, as well as the change frequency between multiple status tags and other data.

[0092] exist Figure 5 In the illustrated embodiment, the state labels can be processed in conjunction with the influence of time, thereby enabling the state data to characterize the state changes of short-term behavior within a preset period, thus improving the effectiveness of the state data.

[0093] Optionally, please refer to Figure 6 , Figure 6 The following is a detailed flowchart of step S200 provided in an embodiment of this application. Step S200 may include steps S210-S240.

[0094] Step S210: Obtain raw data from the collection and equipment set on the target animal within a preset period of time.

[0095] In order to obtain multiple valid detection samples, considering the redundancy of the raw data, all raw data collected periodically within a preset period can be acquired first.

[0096] Optionally, the raw data may include multi-axis acceleration data and multi-axis angular velocity data. The monitoring device may be equipped with a three-axis accelerometer and a gyroscope. The three-axis accelerometer collects multi-axis acceleration data, such as acceleration data along the x, y, and z axes, and the gyroscope collects multi-axis angular velocity data, such as angular velocity data along the x, y, and z axes. The raw data acquisition frequency can be once every 10 Hz to improve the timeliness of the data.

[0097] Step S220: Preprocess the raw data to obtain standard data.

[0098] This process involves preprocessing the raw data to obtain standardized data.

[0099] Optionally, preprocessing may include various methods such as anomaly screening, smoothing, and standardization.

[0100] Step S230: Divide the time based on a preset period to obtain multiple time windows.

[0101] Since the preset period is a time period, in order to classify data under different time conditions, the time can be divided based on the preset period to obtain multiple time windows.

[0102] Optionally, the size of the time window can be set based on the specific size of the preset period and actual needs. For example, when the preset period is 2 hours, the corresponding time window can be set to 5 seconds to reasonably divide the time period within the preset period.

[0103] Step S240: The standard data is segmented based on multiple time windows to obtain multiple test samples.

[0104] In this way, standard data can be segmented based on multiple time windows to obtain multiple valid detection samples under multiple different time windows.

[0105] Optionally, the standard data stream can be segmented according to a time window, with each segment serving as a detection sample for subsequent feature extraction.

[0106] exist Figure 6 In the illustrated embodiments, data preprocessing, time-division, and data segmentation can be performed to improve the effectiveness and temporal characteristics of the detected samples.

[0107] Optionally, please refer to Figure 7 , Figure 7 The following is a detailed flowchart of step S220 provided in an embodiment of this application. Step S220 may include steps S221-S222.

[0108] Step S221: Denoise the original data based on the filtering algorithm to obtain denoised data.

[0109] One approach is to use filtering algorithms to denoise the original data, thereby removing noise and errors and obtaining denoised data.

[0110] Optionally, various filtering algorithms such as Kalman filtering and Wiener filtering can be used to remove noise from the original data. Noise may include vibrations in the breeding environment, errors of the sensor itself, etc., in order to obtain highly effective denoised data.

[0111] Step S222: Standardize the denoised data to obtain standard data.

[0112] This process can be used to standardize the denoised data to eliminate the influence of dimensions and obtain standard data that is easier to analyze and process later.

[0113] Optionally, the acceleration and angular velocity data in the denoised data can be standardized. The min-max standardization method can be used to map the data to the interval [0, 1] to eliminate the influence of dimensions. The standardization formula can include: xnorm=(x-xmin) / (xmax-xmin), where x is the denoised data, xmin is the minimum value of the data in this dimension, xmax is the maximum value of the data in this dimension, and xnorm is the standardized data.

[0114] exist Figure 7 In the illustrated embodiment, the original data can be denoised and standardized to improve the effectiveness and format uniformity of the standard data.

[0115] Optionally, please refer to Figure 8 , Figure 8 This is a flowchart illustrating the fourth state monitoring method provided in this application embodiment. The method may further include steps S621-S622.

[0116] Step S621: Determine the state characteristics of the target animal based on historical data of historical animals.

[0117] Considering the differences in the state of different animal groups, which can affect the monitoring cycle, the state characteristics of the target animal can be determined first based on historical data of historical animals similar to the target animal.

[0118] It should be noted that the method for determining the state characteristics is similar to that in the aforementioned embodiments, and will not be repeated here.

[0119] Step S622: Determine the preset period based on the state characteristics.

[0120] Based on the state characteristics, the monitoring equipment can determine the preset period for monitoring the target animal, and then collect data and process the state of the target animal based on the preset period. This enables the identification of the short-term behavior of the target animal. Based on the preset period, the state data is periodically uploaded to the cloud server for processing, which effectively reduces the energy consumption required to directly upload the raw data and extends the battery life of the monitoring equipment.

[0121] Optionally, when the state characteristics of the target animal indicate frequent changes in its state, a shorter preset period, such as 15 minutes, can be set. Conversely, when the state characteristics indicate a lower frequency of changes in the target animal's state, a longer preset period, such as 3-4 hours, can be set. This ensures data timeliness while minimizing the frequency of data processing and transmission, thereby extending the monitoring equipment's battery life. Alternatively, based on historical data, a corresponding threshold can be set for the data collected by the sensors. The monitoring equipment will only begin data processing and transmission when the data collected by the sensors exceeds this threshold; otherwise, the monitoring equipment will remain in a dormant state to conserve energy and extend its operating time.

[0122] exist Figure 8 In the illustrated embodiment, a suitable preset period can be determined based on the actual situation of the target animal to improve the timeliness of short-term behavior state recognition.

[0123] Optionally, please refer to Figure 9 , Figure 9 This is a detailed flowchart of step S300 provided in an embodiment of the present application. Step S300 may include steps S310-S340.

[0124] Step S310: Perform time-domain feature extraction processing on each detection sample to obtain time-domain features.

[0125] Step S320: Perform a Fourier transform on each detected sample to obtain frequency domain features.

[0126] When processing each detection sample, the time-domain features and frequency-domain features of the detection sample can be extracted separately.

[0127] Optionally, the time-domain features may include multiple features such as mean, standard deviation, peak value, valley value, kurtosis, and skewness, while the frequency-domain features may include multiple features such as extracted spectral peak value, second peak value, spectral energy, dominant frequency, and frequency band energy ratio.

[0128] Step S330: The time-domain features and frequency-domain features are fused to obtain multi-dimensional fused features.

[0129] Among them, time-domain features and frequency-domain features can be fused to obtain multi-dimensional fused features.

[0130] Optionally, time-domain features and frequency-domain features can be fused to form a multi-dimensional, such as a 36-dimensional feature vector, which can be used as a fused feature for subsequent classification and recognition.

[0131] Step S340: Based on the state characteristics of the target animal, the fused features are filtered to obtain multi-dimensional feature data.

[0132] Considering the differences in the states of different groups of animals, the fusion features can be screened based on the state characteristics of the target animal to reduce the adverse effects of insensitive features on state recognition and obtain multi-dimensional feature data that contributes more to state recognition.

[0133] Optionally, based on the state characteristics of the target animal, multiple feature vectors in the fused features can be screened to initially test the sensitivity and accuracy of each feature vector to state behavior, remove insensitive feature vectors, and retain feature vectors that contribute highly to motion state recognition, so as to reduce the data size of feature data and the computational load of recognition model, and improve the efficiency of recognition model in state recognition based on feature response. The feature dimensions of the screened feature data can include multiple key feature vectors such as mean, standard deviation, peak-valley value, etc.

[0134] exist Figure 9 In the illustrated embodiments, feature extraction, fusion processing, and filtering processes can be performed, improving the effectiveness and conciseness of feature data.

[0135] Please see Figure 10 , Figure 10 This is a flowchart illustrating the fifth state monitoring method provided in this application embodiment, applied to a cloud server. The method may include steps S710-S730.

[0136] Step S710: Acquire multiple status data of the target animal at different times sent by the monitoring device.

[0137] In this process, the cloud server receives status data of the target animal every preset period.

[0138] Step S720: Based on multiple state data, determine whether the target animal has any abnormal state.

[0139] The cloud server can perform long-term behavioral assessment based on multiple state data within a preset time period, such as the 24 hours prior to the current moment, to determine whether the target animal has any abnormal state.

[0140] Optionally, abnormal conditions may include estrus, disease, parturition, etc. Estrus lasts for about 10 hours, while disease can last from a few hours to several days. Although parturition itself may be short-lived, there are signs of labor. Considering the signs of labor and the completion of parturition, the time can range from several hours to more than a day. Therefore, these behaviors cannot be directly obtained from a single state data point (e.g., not from data with large fluctuations or peaks). Instead, they need to be determined by comparing long-term data and analyzing the individual characteristics of the target animal.

[0141] Optionally, multiple models can be built on the cloud server to predict and determine long-term behavior. For example, one model might identify estrus when mounting behavior exceeds the daily average by two standard deviations. Another model might identify estrus when the current lying-down time is less than half the daily average. Each model has a corresponding weight. Summing all these models according to their weights yields a value that can be used to determine the probability of long-term behavior occurring. The correlation of multiple models requires adjustments for different pastures, seasons, and feeding methods (free-range or pen-raising), making the process complex. Therefore, performing this on a cloud server eliminates concerns about power consumption and does not affect the lifespan of the monitoring equipment, flexibly meeting the needs of different customers.

[0142] For example, taking the monitoring device as an example that uploads status data every 2 hours, based on the average power consumption and battery capacity of the existing monitoring collar, the monitoring device can theoretically be used for 10 years, which greatly extends the battery life of the monitoring device.

[0143] Optionally, once the monitoring device has completed uploading a certain status data, the status data can be deleted to reduce the data storage pressure on the monitoring device.

[0144] Step S730: If it is determined that the target animal has an abnormal condition, a prompt message is generated.

[0145] The prompts can include information reminding staff to check the actual condition of the target animal or to handle any abnormalities (such as assisting with birthing), and can be in various forms such as voice, text, sound and light.

[0146] Optionally, in the event of a software malfunction in the monitoring device, the cloud server can also send new initial values ​​for the algorithm parameters to adjust the algorithm within the monitoring device, without the need to replace the monitoring device or re-wear it.

[0147] exist Figure 10 In the illustrated embodiment, staff can be prompted in real time to perform inspections, adjustments, and other abnormal handling, thereby reducing the adverse effects of abnormal conditions on the target animals and improving the economic benefits of breeding.

[0148] This application also provides a computer program product, which includes a computer program / instructions. When the computer program / instructions are executed by a processor, they implement the steps in any of the above-described status monitoring methods.

[0149] In summary, this application provides a state monitoring method that addresses the technical problems of low recognition accuracy, high false positive rate, poor versatility, weak noise resistance, and high data transmission energy consumption in existing algorithms. It achieves accurate identification of the movement states of farmed animals, including lying down, ruminating, eating, walking, getting up and lying down, and mounting, providing reliable data support for subsequent health assessment. By removing noise through data preprocessing and extracting time-domain + frequency-domain fusion features, combined with an improved machine learning model, the method effectively improves the recognition accuracy of various movement states of target animals, achieving an overall accuracy of 80.43%. The targeted denoising algorithm effectively filters noise from vibrations and sensor errors in the farming environment, adapting to complex outdoor farming environments. It is adaptable to target animals of different breeds, weights, and ages, and can be extended to the detection of movement states of other animals by adjusting feature extraction parameters and model parameters. Data segmentation and feature filtering reduce computational load, enabling real-time recognition of movement states with a recognition latency of less than 5 seconds, meeting the real-time monitoring needs of large-scale farming. Based on a dynamic threshold adjustment mechanism, it can automatically adjust according to individual differences and environmental changes, further reducing the false positive rate and improving the algorithm's practicality.

[0150] In the several embodiments provided in this application, it should be understood that the disclosed device can also be implemented in other ways. The device embodiments described above are merely illustrative; for example, the block diagrams in the accompanying drawings illustrate the possible architecture, functions, and operations of the device according to various embodiments of this application. In this regard, each block in the block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order than those marked in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagram, and combinations of block diagrams, can be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions.

[0151] In addition, the functional modules in the various embodiments of this application can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.

[0152] If the aforementioned functions are implemented as software functional modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0153] The above description is merely an embodiment of this application and is not intended to limit the scope of protection of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application. It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.

[0154] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

[0155] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising..." does not exclude the presence of additional identical elements in the process, method, article, or apparatus that includes said element.

Claims

1. A condition monitoring method, characterized in that, The method is applied to a monitoring device, and the method includes: The raw data collected from the monitoring equipment installed on the target animal is preprocessed to obtain multiple test samples; Feature extraction is performed on each of the detected samples to obtain multi-dimensional feature data; The feature data is processed by a preset recognition model to determine the state data of the target animal; The status data is periodically uploaded to a cloud server for processing.

2. The method according to claim 1, characterized in that, The training methods for the recognition model include: Based on historical data of historical animals, the state characteristics of the target animal are determined; Based on the aforementioned state characteristics, the parameters of the initial learning model are optimized to obtain an optimized learning model; Based on the historical data and the corresponding historical state labels, the optimization learning model is trained to obtain the recognition model; wherein, the recognition model includes: a clustering model or a decision tree model.

3. The method according to claim 2, characterized in that, The method further includes: Determine the individual information of the target animal; Obtain environmental information about the area where the target animal is located; The parameters of the recognition model are adjusted based on the individual information and the environmental information.

4. The method according to claim 3, characterized in that, in, When the recognition model is the clustering model, adjusting the parameters of the recognition model based on the individual information and the environmental information includes: Based on the individual information and the environmental information, the cluster centers of the clustering model are moved.

5. The method according to any one of claims 1-4, characterized in that, The step of processing the feature data using a preset recognition model to determine the state data of the target animal includes: The feature data is processed by the recognition model to obtain the state label of the target animal; wherein, the state label includes: lying down, ruminating, eating, walking, getting up or mounting. The state tags are time-fused according to the preset period corresponding to the original data to obtain the state data of the target animal.

6. The method according to any one of claims 1-4, characterized in that, The raw data collected from the monitoring equipment installed on the target animal is preprocessed to obtain multiple test samples, including: Acquire the raw data collected periodically within a preset period by the acquisition and collection device installed on the target animal; wherein, the raw data includes: multi-axis acceleration data and multi-axis angular velocity data; The raw data is preprocessed to obtain standard data; The time is divided based on the preset period to obtain multiple time windows; The standard data is segmented based on multiple time windows to obtain multiple detection samples.

7. The method according to claim 6, characterized in that, The preprocessing of the raw data to obtain standard data includes: The original data is denoised using a filtering algorithm to obtain denoised data. The denoised data is standardized to obtain the standard data.

8. The method according to claim 6, characterized in that, The method further includes: Based on historical data of historical animals, the state characteristics of the target animal are determined; Based on the aforementioned state characteristics, the preset period is determined.

9. The method according to any one of claims 1-4, characterized in that, The feature extraction based on each of the detected samples yields multi-dimensional feature data, including: Temporal feature extraction is performed on each of the detected samples to obtain temporal features; Perform a Fourier transform on each of the detected samples to obtain frequency domain features; The time-domain features and the frequency-domain features are fused to obtain multi-dimensional fused features; Based on the state characteristics of the target animal, the fused features are filtered to obtain multi-dimensional feature data.

10. A condition monitoring method, characterized in that, The method is applied to a cloud server, and the method includes: Acquire multiple status data of the target animal at different times, sent by the monitoring equipment; Based on multiple states of data, determine whether the target animal exhibits any abnormal state. If the target animal is determined to be in an abnormal state, a prompt message will be generated.

Images