A grazing machine sensor device and abnormal livestock monitoring method

Through the combination of wireless grid network and graph convolutional neural network, the problem of collecting and monitoring of animal husbandry information in areas with weak cellular network signals is solved, and the stable transmission and accurate monitoring of data in animal husbandry is realized, and the accuracy and communication of animal husbandry is improved. flexibility.

CN118614413BActive Publication Date: 2024-11-08CHINA TELECOM SICHUAN BRANCH
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411103977.6
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-08-13
Publication Date
2024-11-08
Estimated Expiration
2044-08-13

AI Technical Summary

Technical Problem

In pastoral areas where cellular network signals are weak or no signals, existing grazing sensing devices are difficult to achieve stable communication and accurately monitor abnormal livestock, and traditional empirical models and threshold methods are limited in application under different spatial conditions, which can easily lead to misjudgment.

Method used

A sensing device for grazing mother machine is designed, connecting multiple earmarked mother machines and collar hosts through wireless grid networks, and transmitting geographical coordinate information, body temperature and activity data of the herds by satellite or cellular networks, and combining them with the picture volume Accumulate neural networks to analyze sensor data and mine spatial correlation information in the herd to improve monitoring accuracy.

Benefits of technology

The monitoring of herd location and physiological status in areas with poor cellular network quality is realized, which reduces communication costs and maintenance difficulties, improves the accuracy of abnormal animal monitoring, adapts to animal movement, and has self-organization and self-healing capabilities.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118614413B_ABST
    Figure CN118614413B_ABST
Patent Text Reader

Abstract

The invention relates to the technical field of animal husbandry control, and discloses a grazing mother-and-child machine sensing device and an abnormal animal husbandry monitoring method, comprising: an ear tag sub-machine, used for collecting animal body temperature and activity data of individual animals, and transmitting the animal body temperature and activity data of individual animals between ear tag sub-machines and between a collar host and the ear tag sub-machines through a wireless mesh network; a collar host, used for receiving the animal body temperature and activity data of individual animals transmitted by the ear tag sub-machines through the wireless mesh network, and simultaneously collecting geographic coordinate information of a livestock group including geographic coordinate time and geographic coordinate position, and transmitting the geographic coordinate information of the livestock group and the animal body temperature and activity data of individual animals through a satellite or cellular network. The invention solves the problems of livestock group position sensing and livestock physiological state monitoring in areas with poor cellular network quality or without a network, improves communication flexibility and accuracy of abnormal livestock monitoring, and reduces communication cost and maintenance difficulty.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to the technical field of animal husbandry control, and in particular to a grazing mother-and-child machine sensor device and an abnormal animal husbandry monitoring method. Background Art

[0002] Grazing is the main mode of production in my country's pastoral areas. In the vast pastoral areas, herders use natural or artificial pastures to let livestock (such as cattle, sheep, horses, etc.) graze freely, thereby realizing animal husbandry production. Grazing has many advantages, such as making full use of natural resources and reducing production costs; at the same time, livestock can move freely during grazing, which is beneficial to their growth, development and health. However, grazing also faces some challenges, such as low efficiency in livestock search and monitoring, low efficiency in livestock disease prevention and control, and low level of livestock herd management. Therefore, in the process of grazing, modern scientific and technological means need to be adopted to improve grazing efficiency and the efficiency of grassland resource information utilization.

[0003] In order to promote the digitalization of grazing, grazing sensor devices such as electronic collars and electronic ear tags and abnormal livestock monitoring methods based on physiological sensing have appeared on the market. However, since most of my country's pastoral areas are located in remote mountainous areas, plateaus or border areas, these areas: First, due to geographical limitations, infrastructure construction lags behind, cellular network signals are weak or even completely missing, making it difficult for existing grazing products such as electronic collars and electronic ear tags on the market to achieve stable communication; second, the environment is complex and the climatic conditions are harsh. The livestock physiological data obtained by sensors under different spatial conditions may have significant differences. For example, the data obtained by sensors in two livestock herds at an altitude of 3,000 meters and 5,000 meters above sea level show significant differences. However, within the same livestock herd, the difference between sensor data is significantly reduced and presents spatial correlation characteristics, that is, the data trends are consistent. Existing methods for monitoring abnormal livestock based on physiological sensor data mainly utilize empirical models or threshold division methods. This method only considers the physiological data of a single livestock obtained by a single sensor. In this case, the traditional empirical model is a simple threshold setting method. When faced with sensor data obtained under different spatial conditions, it has great application limitations and is prone to misjudgment, so it is impossible to accurately expose abnormal livestock.

[0004] For this reason, some solutions solve the communication problem in areas without cellular networks by setting up fixed satellite network repeaters in pastures for areas with weak or no cellular network signals. Electronic collars and electronic ear tags form a star network with satellite network repeaters, and periodically report status and data to satellite network repeaters. Since livestock move over a wide range and migrate to different grazing areas at different times of the year, it is difficult for fixed satellite network repeaters to follow livestock migration. Therefore, the actual application performance is still unsatisfactory. In view of the fact that traditional empirical models and threshold setting cannot accurately analyze and monitor the current status of abnormal livestock, some methods mainly collect a large amount of individual livestock physiological sensor sample data and monitor abnormal data through support vector machines, feedforward neural networks, recurrent neural networks and other methods. However, due to the high degree of independence between these sample data, the spatial correlation between sensor data caused by the clustering effect of livestock groups has not been considered. Summary of the invention

[0005] In view of the above-mentioned deficiencies in the prior art, the present invention provides a grazing mother-and-child machine sensing device and an abnormal livestock monitoring method, which are used to solve the problems of livestock information collection and abnormal livestock monitoring in pastoral areas with poor cellular network quality.

[0006] In order to achieve the above-mentioned object of the invention, the technical solution adopted by the present invention is:

[0007] A sensor device for a grazing mother-and-child machine, comprising:

[0008] The ear tag sub-machine is used to collect the body temperature and activity data of individual livestock, and transmit the body temperature and activity data of individual livestock through the wireless mesh network to realize data transmission between each ear tag sub-machine and between the collar host and the ear tag sub-machine;

[0009] The collar host is used to receive the body temperature and activity data of individual livestock transmitted by the ear tag sub-machine through the wireless mesh network, and collect the geographic coordinate information of the livestock group including the geographic coordinate time and geographic coordinate position, and transmit the geographic coordinate information of the livestock group and the body temperature and activity data of the individual livestock through satellite or cellular network;

[0010] Among them, the livestock group includes head livestock and non-head livestock, and the ear tag sub-unit is worn on the ear of the non-head livestock to monitor the body temperature and activity data of individual livestock in the non-head livestock. The collar host is worn on the neck of the head livestock to use the geographic coordinate information of the head livestock as the geographic coordinate information of the livestock group, and the ear tag sub-unit is worn on the ear of the head livestock.

[0011] Furthermore, the ear tag sub-machine includes a first self-organizing network module, a temperature measurement and activity monitoring module, and a first solar cell module;

[0012] Temperature measurement and activity monitoring module, used to collect body temperature and activity data of individual livestock;

[0013] The first self-organizing network module is used to participate in the network communication and routing forwarding between the collar host and the ear tag sub-machine, and between the ear tag sub-machines, and to transmit the body temperature and activity data of the individual livestock to the ear tag sub-machines or the collar host using the wireless mesh network;

[0014] The first solar cell module is used to utilize solar energy to supply power to the temperature measurement and activity monitoring module and the first ad hoc network module.

[0015] Furthermore, the collar host includes a second ad hoc network module, a positioning module, a satellite and cellular multi-mode communication module, and a second solar cell module;

[0016] A positioning module is used to collect geographic coordinate information of the livestock herd, including geographic coordinate time and geographic coordinate position;

[0017] The second self-organizing network module is used to initiate and establish a wireless mesh network between the collar host and the ear tag slave, forward and process data, and provide routing strategies;

[0018] Satellite and cellular multi-mode communication module, used to transmit the geographic coordinate information of the herd of livestock, the body temperature and activity data of individual livestock to the server through the cellular network in areas with good cellular network quality, and to transmit the geographic coordinate information of the herd of livestock, the body temperature and activity data of individual livestock to the server through the satellite network in areas with poor cellular network quality or no network;

[0019] The second solar cell module is used to utilize solar energy to supply power to the second ad hoc network module, the positioning module, and the satellite and cellular multi-mode communication module.

[0020] Furthermore, the operation cycle of the ear tag sub-machine and the collar main machine is set to once per hour.

[0021] Furthermore, the operation cycle of the ear tag sub-machine and the collar host is divided into a networking time window, a sleep time window, a sampling time window, and a network communication time window;

[0022] The sleep time window and the networking time window appear alternately multiple times in one operation cycle.

[0023] Furthermore, the operation process of the operation cycle of the ear tag sub-machine and the collar host is:

[0024] When the collar host is turned on, the status of the collar host is initialized first. After the initialization is completed, it enters the networking time window and starts to monitor the broadcast beacon status;

[0025] When the ear tag handset is turned on, the state of the ear tag handset is initialized first, and after the initialization is completed, it enters the networking time window and starts to monitor the broadcast beacon state of other ear tag handsets;

[0026] When the ear tag slave that is not connected to the network is turned on, it will be initialized first. After the initialization is completed, it will send an association request broadcast beacon, and then connect to the network through the ear tag slave or collar host that has connected to the network. After the network connection operation is completed, the networking time window ends;

[0027] When the networking time window ends, the collar host and the ear tag slave form a wireless mesh network, and the collar host broadcasts instructions. After the broadcast instructions end, the collar host and the ear tag slave enter the sleep time window;

[0028] The broadcast instructions include time synchronization instructions and timed wake-up instructions; the time synchronization instructions are used to correct the time of the collar host and the ear tag sub-machine to synchronize them; the timed wake-up instructions are used to wake up the collar host and the ear tag sub-machine that have entered the sleep time window when the wake-up time is reached;

[0029] When entering a new networking time window, the ear tag slave that is not connected to the network sends an association request broadcast beacon, and performs network access operation through the ear tag slave or collar host that has been connected to the network. After the network access operation is completed, it enters the sleep time window;

[0030] When entering the sampling time window, the ear tag sub-machine collects the body temperature and activity data of individual livestock, while the collar host collects the geographic coordinate information of the livestock group;

[0031] When the network communication time window is entered, the body temperature and activity data of individual livestock collected by the ear tag sub-machine are transmitted to the collar host through the wireless mesh network. The collar host summarizes the geographical coordinate information of the collected livestock group and the body temperature and activity data of the individual livestock received, and transmits the summarized data to the server through the cellular network or satellite network;

[0032] Finally, the collar host receives the feedback instructions and control instructions sent by the server, completes the control task including reporting the geographic coordinate information of the livestock herd, obtains the feedback instructions for executing the next cycle, and the collar host broadcasts the instructions for executing the next cycle to the wireless mesh network.

[0033] Furthermore, the specific process of initializing the state of the ear tag sub-unit is as follows:

[0034] Use the ear tag slave to send an association request broadcast beacon and request to join the wireless mesh network of the collar host;

[0035] When the collar host receives the network access response from the ear tag slave, it feeds back the response to the ear tag slave and assigns a network address at the same time;

[0036] After the ear tag slave receives the feedback response from the collar host, the initialization state is completed.

[0037] A method for monitoring abnormal livestock by a grazing machine sensor device comprises the following steps:

[0038] S1. Construct a sample set of livestock sensor data, and input the sample set of livestock sensor data into a graph convolutional neural network model for identifying abnormal livestock for training, so as to obtain a trained graph convolutional neural network model for identifying abnormal livestock;

[0039] S2, transmitting the body temperature and activity data of the individual livestock collected by the ear tag sub-machine to the collar host, and the collar host transmits the received body temperature and activity data of the individual livestock to the server;

[0040] S3, normalizing the body temperature and activity data of the livestock individuals received by the server, and classifying the normalized body temperature and activity data of the livestock individuals by age and gender, obtaining the classified body temperature and activity data of the livestock individuals and constructing an adjacency matrix;

[0041] S4. Input the adjacency matrix constructed in step S3 into the graph convolutional neural network model for identifying abnormal livestock obtained in step S1 to perform abnormal livestock detection to obtain normal or abnormal livestock detection results.

[0042] Furthermore, step S1 specifically includes:

[0043] S11, constructing a livestock sensor data sample set;

[0044] Among them, a single livestock sensor sample data in the livestock sensor data sample set includes a business data field, a physiological sensor data field, and a label field;

[0045] The business data fields include the livestock number, the herd to which the livestock belongs, and the livestock age;

[0046] Physiological sensor data fields include livestock body temperature and activity data;

[0047] The label field includes normal animal husbandry or abnormal animal husbandry, where normal animal husbandry is represented by 0 and abnormal animal husbandry is represented by 1;

[0048] S12, normalizing the livestock body temperature and activity data in the physiological sensor data field to obtain normalized livestock physiological sensor sample data;

[0049] S13, dividing the normalized livestock physiological sensor sample data obtained in step S12 into age groups and genders to obtain livestock physiological sensor sample data with age and gender;

[0050] S14, extracting data with the same livestock number from the livestock physiological sensor sample data with age and gender obtained in step S13, and constructing an adjacency matrix;

[0051] Among them, each node in the adjacency matrix contains two features: body temperature and activity;

[0052] S15, dividing the adjacency matrix constructed in step S14 into a training set and a test set according to a set ratio;

[0053] S16, inputting the training set into the graph convolutional neural network model for identifying abnormal livestock for training, using the graph convolutional layer of the graph convolutional neural network model for identifying abnormal livestock to extract features, inputting the extracted features into the fully connected output layer of the graph convolutional neural network model for identifying abnormal livestock, using the sigmoid activation function to fix the feature value between 0 and 1, and using the binary cross entropy loss function to measure the loss between the predicted value output by the fully connected output layer and the label value, to obtain the trained graph convolutional neural network model for identifying abnormal livestock;

[0054] S17. Input the test set into the trained graph convolutional neural network model for identifying abnormal livestock obtained in step S16 for performance testing to obtain a trained graph convolutional neural network model for identifying abnormal livestock.

[0055] Furthermore, step S13 specifically includes:

[0056] S131, dividing the livestock physiological sensor sample data into three age group samples, namely [0, 1) age, [1, 3) age and [3, 3+] age;

[0057] S132. Divide the livestock individuals into genders in the three age group samples divided in step S131, and divide them into male samples with age and female samples with age, namely, [0, 1) age male samples, [1, 3) age male samples and [3, 3+] age male samples, [0, 1) age female samples, [1, 3) age female samples and [3, 3+] age female samples.

[0058] The present invention has the following beneficial effects:

[0059] 1. The grazing mother-and-child machine sensor device proposed in the present invention connects multiple ear tag sub-machines to a collar host, making full use of hardware resources, realizing resource sharing and load balancing, and realizing the herd location sensing and livestock physiological status monitoring in areas with poor cellular network quality or no network, solving the defect that the network signal quality in pastoral areas is poor and it is impossible to deploy Internet of Things nodes;

[0060] 2. A wireless mesh network is constructed between the collar host and the ear tag sub-machine. Compared with the existing fixed satellite network repeater method, the network topology structure of the wireless mesh network can adapt to the dynamic networking process of livestock during movement, which not only improves the flexibility of communication, but also reduces the communication cost and maintenance difficulty; and the network communication link of the wireless mesh network presents a mesh distribution feature, has the characteristics of self-organization and self-healing ability, strong robustness and stable data transmission; at the same time, on the basis of constructing a wireless mesh network between the collar host and the ear tag sub-machine, combined with the satellite communication network, it has the advantages of wide coverage and is not affected by terrain and geographical environment, which expands the communication range and can cover a wider area of ​​the pastoral area;

[0061] 3. The abnormal livestock monitoring method of the grazing mother-and-child machine sensor device proposed in the present invention analyzes the sensor network from the livestock herd scale, and uses the graph convolutional neural network to mine the deeper spatial correlation information between the sensor data in the same livestock herd, avoiding model misjudgment caused by significant differences in sensor values ​​in different spaces, thereby improving the accuracy of abnormal livestock monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] Figure 1 This is a schematic diagram of the principle of the sensor device of the grazing mother-and-child machine proposed by the present invention;

[0063] Figure 2 It is a schematic diagram of the adjacency matrix graph structure formed by the livestock clustering effect in the embodiment. DETAILED DESCRIPTION

[0064] The specific implementation modes of the present invention are described below so that those skilled in the art can understand the present invention. However, it should be clear that the present invention is not limited to the scope of the specific implementation modes. For those of ordinary skill in the art, as long as various changes are within the spirit and scope of the present invention as defined and determined by the attached claims, these changes are obvious, and all inventions and creations utilizing the concept of the present invention are protected.

[0065] like Figure 1 As shown, a grazing mother-and-child machine sensing device comprises:

[0066] The ear tag sub-machine is used to collect the body temperature and activity data of individual livestock, and transmit the body temperature and activity data of individual livestock between each ear tag sub-machine and between the collar host and the ear tag sub-machine through a wireless mesh network.

[0067] The collar host is used to receive the body temperature and activity data of individual livestock transmitted by the ear tag sub-unit through the wireless grid network, and at the same time collect the geographic coordinate information of the livestock group including the geographic coordinate time and geographic coordinate position, and transmit the geographic coordinate information of the livestock group and the body temperature and activity data of individual livestock through satellite or cellular networks.

[0068] Among them, the livestock group includes head livestock and non-head livestock, and the ear tag sub-unit is worn on the ear of the non-head livestock to monitor the body temperature and activity data of individual livestock in the non-head livestock. The collar host is worn on the neck of the head livestock to use the geographic coordinate information of the head livestock as the geographic coordinate information of the livestock group, and the ear tag sub-unit is worn on the ear of the head livestock.

[0069] In this embodiment, Figure 1 The herd can be considered as a whole, specifically: (1) The herd is divided into head herds and non-head herds according to their status. The head herd is the leader of the herd. Usually, there is only one leader in a herd, who is an older or stronger individual in the herd. The collar host is only worn by the head herd, and all the herds in the herd wear ear tag sub-units. The collar host and ear tag sub-units form a Mesh self-organizing network, i.e., a wireless mesh network, and establish communication links between the collar host and the ear tag sub-units, and between the ear tag sub-units. The sensor data collected by the ear tag sub-units, i.e., the body temperature and activity data of the individual herdsmen, are transmitted to the collar host through the communication link and summarized with the geographic coordinate information of the herd collected by the collar host, and then transmitted to the server by the collar host via satellite or cellular network. In addition, according to the wireless mesh network formed between the collar host and the ear tag sub-machine, the collar host can be regarded as a network coordinator and the ear tag sub-machine can be regarded as a network node, thereby realizing the process of dynamic networking of livestock during movement, improving the flexibility of communication, and reducing communication costs and maintenance difficulties. In summary, the present invention connects multiple ear tag sub-machines to a collar host, fully utilizes hardware resources, realizes resource sharing and load balancing, and simultaneously realizes the position sensing of livestock herds and the monitoring of livestock physiological status in areas with poor cellular network quality or no network, solving the defect that the quality of network signals in pastoral areas is poor and it is impossible to deploy Internet of Things nodes; (2) Analyze the sensor network from the scale of livestock herds, and use graph convolutional neural networks in subsequent steps to mine deeper spatial correlation information between sensor data in the same livestock herd, avoiding model misjudgment caused by significant differences in sensor values ​​in different spaces, thereby improving the accuracy of abnormal livestock monitoring.

[0070] Specifically, the ear tag sub-machine includes a first self-organizing network module, a temperature measurement and activity monitoring module, and a first solar cell module; the temperature measurement and activity monitoring module is used to collect the body temperature and activity data of individual livestock; the first self-organizing network module is used to participate in network communication and routing forwarding between the collar host and the ear tag sub-machine, and between each ear tag sub-machine, and to use the wireless mesh network to transmit the body temperature and activity data of individual livestock to the ear tag sub-machines or the collar host; the first solar cell module is used to use solar energy to power the temperature measurement and activity monitoring module and the first self-organizing network module.

[0071] In this embodiment, since photovoltaic resources are sufficient in the vast grazing areas, a solar cell module is provided to utilize solar energy to power the collar host and the ear tag sub-machine, thereby balancing the power consumption loss of the collar host and the ear tag sub-machine.

[0072] Specifically, the collar host includes a second self-organizing network module, a positioning module, a satellite and cellular multi-mode communication module, and a second solar cell module; the positioning module is used to collect geographic coordinate information of the livestock herd, including geographic coordinate time and geographic coordinate position; the second self-organizing network module is used to initiate and establish a wireless grid network between the collar host and the ear tag sub-machine, forward and process data, and provide routing strategies; the satellite and cellular multi-mode communication module is used to transmit the geographic coordinate information of the livestock herd, the body temperature and activity data of individual livestock to the server through the cellular network in areas with good cellular network quality, and transmit the geographic coordinate information of the livestock herd, the body temperature and activity data of individual livestock to the server through the satellite network in areas with poor cellular network quality or no network; the second solar cell module is used to use solar energy to power the second self-organizing network module, the positioning module, and the satellite and cellular multi-mode communication module.

[0073] Specifically, the operation cycle of the ear tag slave machine and the collar host machine is set to once per hour.

[0074] In this embodiment, the operation cycle can also be set to other time intervals according to the actual situation of the livestock herd.

[0075] Specifically, the operation cycle of the ear tag sub-machine and the collar host is divided into a networking time window, a sleep time window, a sampling time window and a network communication time window; among which, the sleep time window and the networking time window appear alternately multiple times in one operation cycle.

[0076] Specifically, the operation process of the operation cycle of the ear tag sub-machine and the collar host is:

[0077] When the collar host is turned on, the status of the collar host is initialized first. After the initialization is completed, it enters the networking time window and starts to monitor the broadcast beacon status.

[0078] When the ear tag handset is turned on, the state of the ear tag handset is initialized first, and after the initialization is completed, it enters the networking time window and starts to monitor the broadcast beacon state of other ear tag handsets.

[0079] Specifically, the specific process of initializing the state of the ear tag sub-machine is as follows:

[0080] The ear tag slave sends an association request broadcast beacon and requests to join the wireless mesh network of the collar host. When the collar host receives the network access response from the ear tag slave, it feeds back a response to the ear tag slave and assigns a network address. After the ear tag slave receives the feedback response from the collar host, the initialization state is completed.

[0081] When an ear tag slave that is not connected to the network is turned on, it is initialized first. After the initialization is completed, an association request broadcast beacon is sent, and the network access operation is performed through the ear tag slave or collar host that has been connected to the network. After the network access operation is completed, the networking time window ends.

[0082] When the networking time window ends, the collar host and the ear tag slave form a wireless mesh network, and the collar host broadcasts instructions. After the broadcast instructions end, the collar host and the ear tag slave enter the sleep time window.

[0083] Among them, the broadcast instructions include time synchronization instructions and timed wake-up instructions; the time synchronization instructions are used to correct the time of the collar host and the ear tag sub-machine to synchronize them; the timed wake-up instructions are used to wake up the collar host and ear tag sub-machine that have entered the sleep time window when the wake-up time is reached.

[0084] When entering a new networking time window, the ear tag slave that is not connected to the network sends an association request broadcast beacon, and performs network access operation through the ear tag slave or collar host that has connected to the network. After the network access operation is completed, it enters the sleep time window.

[0085] When entering the sampling time window, the ear tag sub-machine collects the body temperature and activity data of individual livestock, while the collar host collects the geographic coordinate information of the livestock group.

[0086] In this embodiment, when entering the sampling time window, the ear tag sub-machine and the collar host machine can also collect voltage and solar cell charging status data.

[0087] When the network communication time window is entered, the body temperature and activity data of individual livestock collected by the ear tag sub-unit are transmitted to the collar host through the wireless mesh network. The collar host summarizes the collected geographic coordinate information of the livestock group and the received body temperature and activity data of individual livestock, and transmits the summarized data to the server through the cellular network or satellite network.

[0088] Finally, the collar host receives the feedback instructions and control instructions sent by the server, completes the control task including reporting the geographic coordinate information of the livestock herd, obtains the feedback instructions for executing the next cycle, and the collar host broadcasts the instructions for executing the next cycle to the wireless mesh network.

[0089] In the present embodiment, a wireless mesh network is constructed between the collar host and the ear tag sub-machine. Compared with the existing fixed satellite network repeater method, this network topology can adapt to the dynamic networking process of livestock during movement, which not only improves the flexibility of communication, but also reduces the communication cost and maintenance difficulty; and the network communication link presents a mesh distribution feature, has the characteristics of self-organization and self-healing ability, strong robustness and stable data transmission; at the same time, on the basis of constructing a wireless mesh network between the collar host and the ear tag sub-machine, combined with the satellite communication network, it has the advantages of wide coverage and is not affected by terrain and geographical environment, further expanding the communication range and can cover a wider area of ​​​​the pastoral area.

[0090] like Figure 1 As shown, a method for monitoring abnormal livestock by a grazing machine sensor device includes the following steps S1-S4:

[0091] S1. Construct a sample set of livestock sensor data, and input the sample set of livestock sensor data into a graph convolutional neural network model for identifying abnormal livestock for training, so as to obtain a trained graph convolutional neural network model for identifying abnormal livestock.

[0092] In this embodiment, a livestock sensor data sample set is constructed through pre-training, so that the livestock sensor sample set data can be used to train a convolutional neural network model for identifying abnormal livestock graphs.

[0093] Specifically, step S1 includes S11-S17:

[0094] S11. Construct a livestock sensor data sample set; wherein, a single livestock sensor sample data in the livestock sensor data sample set includes a business data field, a physiological sensor data field, and a label field; the business data field includes the livestock number, the herd to which the livestock belongs, and the livestock age; the physiological sensor data field includes the livestock body temperature and activity data; the label field includes normal livestock or abnormal livestock, normal livestock is represented by 0, and abnormal livestock is represented by 1.

[0095] S12, normalizing the livestock body temperature and activity data in the physiological sensing data field to obtain normalized livestock physiological sensing sample data.

[0096] In this embodiment, the purpose of normalization is to improve the model convergence speed and accuracy and prevent gradient explosion or disappearance when the data is used for model training.

[0097] S13. Divide the normalized livestock physiological sensor sample data obtained in step S12 into age groups and genders to obtain livestock physiological sensor sample data with age and gender.

[0098] In this embodiment, since there are differences between the physiological sensor sample data of livestock of different age groups and different genders, it is necessary to divide the livestock physiological sensor sample data into age groups and genders when constructing the sample data.

[0099] Specifically, step S13 specifically includes S131-S1332:

[0100] S131. Divide the livestock physiological sensor sample data into three age group samples, namely, [0, 1) age, [1, 3) age and [3, 3+] age.

[0101] S132. Divide the livestock individuals into genders in the three age group samples divided in step S131, and divide them into male samples with age and female samples with age, namely, [0, 1) age male samples, [1, 3) age male samples and [3, 3+] age male samples, [0, 1) age female samples, [1, 3) age female samples and [3, 3+] age female samples.

[0102] S14. Extract data with the same livestock number from the livestock physiological sensor sample data with age and gender obtained in step S13, and construct an adjacency matrix; wherein each node in the adjacency matrix contains two features: body temperature and activity level.

[0103] In this embodiment, the physiological sensor sample data of livestock belonging to the same livestock group is extracted to construct an adjacency matrix, and each node of the adjacency matrix contains two feature vectors of body temperature and activity. The same livestock group is queried according to the livestock number.

[0104] Assume that there are 10 livestock individuals aged 0 to 1 in a herd. All of these livestock individuals are equipped with ear tags. The graph structure of the sensor network formed by the ear tags is as follows: Figure 2 As shown, Figure 2 10-19 are nodes of the graph structure, and the number of nodes is determined by the number of livestock individuals in the herd. In addition, these nodes are divided into two categories according to the label field, namely normal nodes and abnormal nodes. The category of the node corresponds to the livestock status. Normal livestock means normal nodes, and abnormal livestock means abnormal nodes. In the subsequent model training, the body temperature and activity of the node are input into the model as features. The output label of the trained model is 0 or 1, 0 for normal livestock, and 1 for abnormal livestock.

[0105] S15, dividing the adjacency matrix constructed in step S14 into a training set and a test set according to a set ratio.

[0106] S16. Input the training set into the graph convolutional neural network model for identifying abnormal livestock for training, use the graph convolution layer of the graph convolutional neural network model for identifying abnormal livestock to extract features, input the extracted features into the fully connected output layer of the graph convolutional neural network model for identifying abnormal livestock, use the sigmoid activation function to fix the feature value between 0 and 1, and use the binary cross entropy loss function to measure the loss between the predicted value and the label value output by the fully connected output layer, so as to obtain the trained graph convolutional neural network model for identifying abnormal livestock.

[0107] S17. Input the test set into the trained graph convolutional neural network model for identifying abnormal livestock obtained in step S16 for performance testing to obtain a trained graph convolutional neural network model for identifying abnormal livestock.

[0108] In this embodiment, in steps S15-S17, the data in the adjacency matrix are randomly divided into a training set and a test set in a ratio of 8:2, the training set is input into the graph convolutional neural network model for identifying abnormal livestock for model training, and the test set is used to detect the accuracy of the trained model, and finally the trained graph convolutional neural network model for identifying abnormal livestock is output.

[0109] S2. The body temperature and activity data of the individual livestock collected by the ear tag sub-machine are transmitted to the collar host, and the collar host transmits the received body temperature and activity data of the individual livestock to the server.

[0110] S3. Normalize the body temperature and activity data of the individual livestock received by the server, and classify the normalized body temperature and activity data of the individual livestock by age and gender to obtain the classified body temperature and activity data of the individual livestock and construct an adjacency matrix.

[0111] In this embodiment, the body temperature and activity data of individual livestock collected by the ear tag sub-units in the same livestock group at the same time and received by the server are normalized, and the age and gender classification is performed using the method in step S13, so as to obtain the classified body temperature and activity data of individual livestock and construct an adjacency matrix so that the adjacency matrix data can be input into the model for abnormal livestock status identification.

[0112] S4. Input the adjacency matrix constructed in step S3 into the graph convolutional neural network model for identifying abnormal livestock obtained in step S1 to perform abnormal livestock detection to obtain normal or abnormal livestock detection results.

[0113] In summary, the grazing mother-and-child machine sensor device and abnormal livestock monitoring method proposed in the present invention first connect multiple ear tag sub-machines to a collar host, fully utilize hardware resources, realize resource sharing and load balancing, and simultaneously realize livestock herd position sensing and livestock physiological state monitoring in areas with poor cellular network quality or no network, solving the defects of poor network signal quality in pastoral areas and inability to deploy Internet of Things nodes; secondly, the sensor network is analyzed from the livestock herd scale, and the graph convolutional neural network is used to mine deeper spatial correlation information between sensor data in the same livestock herd, avoiding model misjudgment caused by significant differences in sensor values ​​in different spaces, thereby improving the accuracy of abnormal livestock monitoring.

[0114] The present invention uses specific embodiments to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core idea. At the same time, for those skilled in the art, according to the idea of ​​the present invention, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as a limitation on the present invention.

[0115] Those skilled in the art will appreciate that the embodiments described herein are intended to help readers understand the principles of the present invention, and should be understood that the protection scope of the present invention is not limited to such specific statements and embodiments. Those skilled in the art can make various other specific variations and combinations that do not deviate from the essence of the present invention based on the technical revelations disclosed by the present invention, and these variations and combinations are still within the protection scope of the present invention.

Claims

1. A method for monitoring abnormal livestock using a grazing machine sensor device, the grazing machine sensor device comprising: The ear tag sub-machine is used to collect the body temperature and activity data of individual livestock, and transmit the body temperature and activity data of individual livestock through the wireless mesh network to realize data transmission between each ear tag sub-machine and between the collar host and the ear tag sub-machine; The collar host is used to receive the body temperature and activity data of individual livestock transmitted by the ear tag sub-machine through the wireless mesh network, and collect the geographic coordinate information of the livestock group including the geographic coordinate time and geographic coordinate position, and transmit the geographic coordinate information of the livestock group and the body temperature and activity data of the individual livestock through satellite or cellular network; The herd includes a head animal and non-head animals, and the ear tag sub-machine is worn on the ear of the non-head animal to monitor the body temperature and activity data of the individual animals in the non-head animal. The collar host is worn on the neck of the head animal to use the geographic coordinate information of the head animal as the geographic coordinate information of the herd, and the ear tag sub-machine is worn on the ear of the head animal. The ear tag sub-machine includes a first self-organizing network module, a temperature measurement and activity monitoring module, and a first solar cell module; Temperature measurement and activity monitoring module, used to collect body temperature and activity data of individual livestock; The first self-organizing network module is used to participate in the network communication and routing forwarding between the collar host and the ear tag sub-machine, and between the ear tag sub-machines, and to transmit the body temperature and activity data of the individual livestock to the ear tag sub-machines or the collar host using the wireless mesh network; A first solar cell module, used to use solar energy to power the temperature measurement and activity monitoring module and the first ad hoc network module; The collar host includes a second self-organizing network module, a positioning module, a satellite and cellular multi-mode communication module, and a second solar cell module; A positioning module is used to collect geographic coordinate information of the livestock herd, including geographic coordinate time and geographic coordinate position; The second self-organizing network module is used to initiate and establish a wireless mesh network between the collar host and the ear tag slave, forward and process data, and provide routing strategies; Satellite and cellular multi-mode communication module, used to transmit the geographic coordinate information of the herd of livestock, the body temperature and activity data of individual livestock to the server through the cellular network in areas with good cellular network quality, and to transmit the geographic coordinate information of the herd of livestock, the body temperature and activity data of individual livestock to the server through the satellite network in areas with poor cellular network quality or no network; A second solar cell module is used to utilize solar energy to power the second ad hoc network module, the positioning module, and the satellite and cellular multi-mode communication module; The operation cycle of the ear tag sub-machine and the collar main machine is set to once per hour; The operation cycle of the ear tag slave and collar host is divided into a networking time window, a sleep time window, a sampling time window, and a network communication time window; Among them, the sleep time window and the networking time window appear alternately multiple times in one operation cycle; The operation process of the ear tag sub-machine and the collar host machine is as follows: When the collar host is turned on, the status of the collar host is initialized first. After the initialization is completed, it enters the networking time window and starts to monitor the broadcast beacon status; When the ear tag handset is turned on, the state of the ear tag handset is initialized first, and after the initialization is completed, it enters the networking time window and starts to monitor the broadcast beacon state of other ear tag handsets; When the ear tag slave that is not connected to the network is turned on, it will be initialized first. After the initialization is completed, it will send an association request broadcast beacon, and then connect to the network through the ear tag slave or collar host that has connected to the network. After the network connection operation is completed, the networking time window ends; When the networking time window ends, the collar host and the ear tag slave form a wireless mesh network, and the collar host broadcasts instructions. After the broadcast instructions end, the collar host and the ear tag slave enter the sleep time window; The broadcast instructions include time synchronization instructions and timed wake-up instructions; the time synchronization instructions are used to correct the time of the collar host and the ear tag sub-machine to synchronize them; the timed wake-up instructions are used to wake up the collar host and the ear tag sub-machine that have entered the sleep time window when the wake-up time is reached; When entering a new networking time window, the ear tag slave that is not connected to the network sends an association request broadcast beacon, and performs network access operation through the ear tag slave or collar host that has been connected to the network. After the network access operation is completed, it enters the sleep time window; When entering the sampling time window, the ear tag sub-machine collects the body temperature and activity data of individual livestock, while the collar host collects the geographic coordinate information of the livestock group; When the network communication time window is entered, the body temperature and activity data of individual livestock collected by the ear tag sub-machine are transmitted to the collar host through the wireless mesh network. The collar host summarizes the geographical coordinate information of the collected livestock group and the body temperature and activity data of the individual livestock received, and transmits the summarized data to the server through the cellular network or satellite network; Finally, the collar host receives the feedback instructions and control instructions sent by the server, completes the control task including reporting the geographic coordinate information of the herd, obtains the feedback instructions for executing the next cycle, and broadcasts the execution instructions for the next cycle to the wireless mesh network; The specific process of initializing the state of the ear tag sub-machine is as follows: Use the ear tag slave to send an association request broadcast beacon and request to join the wireless mesh network of the collar host; When the collar host receives the network access response from the ear tag slave, it feeds back the response to the ear tag slave and assigns a network address at the same time; After the ear tag slave receives the feedback response from the collar host, the initialization state is completed, which is characterized by including the following steps: S1. Construct a sample set of livestock sensor data, and input the sample set of livestock sensor data into a graph convolutional neural network model for identifying abnormal livestock for training, so as to obtain a trained graph convolutional neural network model for identifying abnormal livestock; Wherein, step S1 specifically includes: S11, constructing a livestock sensor data sample set; Among them, a single livestock sensor sample data in the livestock sensor data sample set includes a business data field, a physiological sensor data field, and a label field; The business data fields include the livestock number, the herd to which the livestock belongs, and the livestock age; Physiological sensor data fields include livestock body temperature and activity data; The label field includes normal animal husbandry or abnormal animal husbandry, where normal animal husbandry is represented by 0 and abnormal animal husbandry is represented by 1; S12, normalizing the livestock body temperature and activity data in the physiological sensor data field to obtain normalized livestock physiological sensor sample data; S13, dividing the normalized livestock physiological sensor sample data obtained in step S12 into age groups and genders to obtain livestock physiological sensor sample data with age and gender; S14, extracting data with the same livestock number from the livestock physiological sensor sample data with age and gender obtained in step S13, and constructing an adjacency matrix; Among them, each node in the adjacency matrix contains two features: body temperature and activity; S15, dividing the adjacency matrix constructed in step S14 into a training set and a test set according to a set ratio; S16, inputting the training set into the graph convolutional neural network model for identifying abnormal livestock for training, using the graph convolutional layer of the graph convolutional neural network model for identifying abnormal livestock to extract features, inputting the extracted features into the fully connected output layer of the graph convolutional neural network model for identifying abnormal livestock, using the sigmoid activation function to fix the feature value between 0 and 1, and using the binary cross entropy loss function to measure the loss between the predicted value output by the fully connected output layer and the label value, to obtain the trained graph convolutional neural network model for identifying abnormal livestock; S17, inputting the test set into the trained graph convolutional neural network model for identifying abnormal livestock obtained in step S16 for performance testing, to obtain a trained graph convolutional neural network model for identifying abnormal livestock; S2, transmitting the body temperature and activity data of the individual livestock collected by the ear tag sub-machine to the collar host, and the collar host transmits the received body temperature and activity data of the individual livestock to the server; S3, normalizing the body temperature and activity data of the livestock individuals received by the server, and classifying the normalized body temperature and activity data of the livestock individuals by age and gender, obtaining the classified body temperature and activity data of the livestock individuals and constructing an adjacency matrix; S4. Input the adjacency matrix constructed in step S3 into the graph convolutional neural network model for identifying abnormal livestock obtained in step S1 to perform abnormal livestock detection to obtain normal or abnormal livestock detection results.

2. The abnormal livestock monitoring method of the grazing mother-and-child machine sensor device according to claim 1 is characterized in that: Step S13 specifically includes: S131, dividing the livestock physiological sensor sample data into three age group samples, namely, [0, 1) age, [1, 3) age and [3, 3+] age; S132. Divide the livestock individuals into genders in the three age group samples divided in step S131, and divide them into male samples with age and female samples with age, namely, [0, 1) age male samples, [1, 3) age male samples and [3, 3+] age male samples, [0, 1) age female samples, [1, 3) age female samples and [3, 3+] age female samples.