An animal disease early warning method and system based on behavior and body temperature dual sequence

Abstract

The present invention discloses an animal disease early warning method and system based on behavior and body temperature dual sequences, wherein the animal disease early warning method based on behavior and body temperature dual sequences includes: real-time acquisition of animal body temperature index sequence and motion index sequence; inputting the animal motion index sequence into a behavior recognition network to identify the animal behavior recognition sequence; using a dual sequence decision network to fuse the body temperature index sequence and the behavior recognition sequence, and according to the fused body temperature index sequence and behavior recognition sequence, identifying the animal disease early warning information. The technical solution of the present invention can solve the problem in the prior art that it is difficult to determine whether an animal is sick only by detecting the animal's behavior, and it is difficult to timely warn of animal diseases.

Description

Technical Field

[0001] The present invention relates to the field of neural network technology, and in particular to an animal disease early warning method and system based on behavior and body temperature dual sequences. Background Art

[0002] With the improvement of living standards, more and more people keep pets. In the process of pet raising, pets will inevitably get sick, so it is very important to warn of pet diseases in advance.

[0003] For early warning of pet diseases, the industry usually monitors various indicators of pets (such as body temperature, blood oxygen, blood pressure, etc.), and performs anomaly detection by processing and analyzing sequence data in a rule-based and model-based manner. Existing methods of monitoring pet indicators usually use relatively single data and cannot effectively combine the pet's daily behavior and abnormal behavior when the disease occurs, such as abnormal eating, abnormal licking, etc. In order to solve the above problems, the existing technology provides a solution for early warning of pet diseases by identifying pet behavior. For the identification of pet behavior, the industry usually uses computer vision-based methods to monitor the actions of pets, that is, deploying surveillance cameras within the pet's activity range, and using image processing methods to detect the pet's behavior in real time.

[0004] However, the above-mentioned animal behavior detection method based on computer vision has a limited detection range, and abnormal behaviors of pets usually occur in hidden corners. For example, when abnormal licking behavior occurs, animals usually hide themselves for the purpose of self-protection. Therefore, it is difficult to determine whether an animal is sick only by detecting its behavior, and it is difficult to provide timely and accurate warnings of animal diseases. Summary of the invention

[0005] The present invention provides an animal disease early warning method and system based on behavior and body temperature dual sequences, aiming to solve the problem in the prior art that it is difficult to determine whether an animal is sick only by detecting the animal's behavior, and it is difficult to provide early warning of animal diseases in a timely and accurate manner.

[0006] To achieve the above object, according to a first aspect of the present invention, the present invention proposes an animal disease early warning method based on behavior and body temperature dual sequences, comprising:

[0007] The animal's body temperature index sequence and movement index sequence are collected in real time;

[0008] Inputting the animal's motion index sequence into a behavior recognition network to obtain the animal's behavior recognition sequence;

[0009] A dual sequence decision network is used to fuse the body temperature index sequence and the behavior recognition sequence. Based on the fused body temperature index sequence and behavior recognition sequence, the animal's disease warning information is identified.

[0010] Preferably, in the above-mentioned animal disease early warning method, the step of acquiring the animal's body temperature index sequence and movement index sequence in real time comprises:

[0011] Set the unique identifier of the animal;

[0012] Use a body temperature sensor to collect the animal's body temperature index sequence in real time;

[0013] Use multiple motion sensors to collect multiple motion information of animals in real time, and fuse multiple motion information to obtain motion index sequences;

[0014] Establish corresponding relationships between the unique identifier and the temperature index sequence and the exercise index sequence respectively.

[0015] Preferably, in the above-mentioned animal disease early warning method, the step of inputting the animal's motion index sequence into the behavior recognition network and identifying the animal's behavior recognition sequence comprises:

[0016] Use the sliding window of the behavior recognition network to segment and extract features from the motion index sequence to obtain the animal's motion representation features;

[0017] The sequence model of the behavior recognition network is used to process the motion representation features to obtain the fusion sequence of the animal's motion features;

[0018] The behavior decision model of the behavior recognition network is used to identify the behavior type of the motion feature fusion sequence to obtain the animal's behavior recognition sequence.

[0019] Preferably, in the above-mentioned animal disease early warning method, the step of using the sequence model of the behavior recognition network to process the motion representation features includes:

[0020] Use a sequence model to perform sequence modeling on the motion representation features to obtain the motion feature quantity at each moment in a predetermined period of time;

[0021] The sequence model is used to fuse the motion feature quantities at all moments in a predetermined period of time to obtain a motion feature fusion sequence.

[0022] Preferably, in the above-mentioned animal disease early warning method, the step of using a dual sequence decision network to fuse the body temperature index sequence and the behavior recognition sequence, and identifying the animal disease early warning information according to the fused body temperature index sequence and behavior recognition sequence includes:

[0023] The first sequence network of the dual sequence decision network is used to perform sequence modeling on the temperature index sequence to obtain the temperature feature value;

[0024] The second sequence network of the dual sequence decision network is used to perform sequence modeling on the behavior recognition sequence to obtain the behavior feature quantity;

[0025] The attention network of the dual sequence decision network is used to fuse the body temperature feature quantity and the behavior feature quantity to obtain the body temperature feature fusion sequence and the behavior feature fusion sequence respectively;

[0026] The temperature feature fusion sequence and the behavior feature fusion sequence are input together into the decision network model of the dual sequence decision network for disease classification to obtain animal disease warning information.

[0027] Preferably, in the above-mentioned animal disease early warning method, the steps of using the attention network of the dual sequence decision network to fuse the body temperature feature quantity and the body temperature feature quantity respectively to obtain the body temperature feature fusion sequence and the behavior feature fusion sequence respectively include:

[0028] Control the attention network according to the formula: YT t =∑(α i *HA i )+HT t The temperature feature and behavior feature are fused to obtain the temperature feature fusion sequence; among them, YT t is the temperature feature fusion sequence, α i is the attention network score of body temperature, HA i is the behavior characteristic at time i, HT t are all body temperature characteristics in period t; and,

[0029] Control the attention network according to the formula: YA t =∑(β i *HT i )+HA t The temperature feature and the behavior feature are fused to obtain the behavior feature fusion sequence; among them, YA t is the behavior feature fusion sequence, β i is the attention network score of the behavior, HT i is the body temperature characteristic at time i, HA t is the characteristic quantity of all behaviors in period t.

[0030] According to the second aspect of the present invention, the present invention also provides an animal disease early warning system based on behavior and body temperature dual sequence, comprising:

[0031] Information collection module, behavior recognition network and dual sequence decision network; among them,

[0032] An information collection module is used to collect the animal's body temperature index sequence and movement index sequence in real time;

[0033] A behavior recognition network is used to input a movement index sequence of an animal and identify and obtain a behavior recognition sequence of the animal;

[0034] The dual sequence decision network is used to fuse the body temperature index sequence and the behavior recognition sequence, and to identify and obtain the animal's disease warning information based on the fused body temperature index sequence and behavior recognition sequence.

[0035] Preferably, in the above-mentioned animal disease early warning system, the information collection module includes:

[0036] The identification setting submodule is used to set the unique identification of the animal;

[0037] The body temperature index sequence acquisition submodule is used to use the body temperature sensor to collect the animal's body temperature index sequence in real time;

[0038] The motion information collection submodule is used to use a variety of motion sensors to collect a variety of motion information of animals in real time, and fuse the multiple motion information to obtain a motion index sequence;

[0039] The relationship establishment submodule is used to establish the corresponding relationship between the unique identifier and the body temperature index sequence and the exercise index sequence respectively.

[0040] Preferably, in the above-mentioned animal disease early warning system, the behavior recognition network includes:

[0041] The sliding window is used to segment and extract features from the motion index sequence to obtain the motion representation features of the animal;

[0042] A sequence model is used to process motion representation features and obtain a fusion sequence of animal motion features;

[0043] The behavior decision model is used to identify the behavior type of the motion feature fusion sequence and obtain the animal's behavior recognition sequence.

[0044] Preferably, in the above-mentioned animal disease early warning system, the dual sequence decision network includes:

[0045] The first sequence network is used to perform sequence modeling on the temperature index sequence to obtain the temperature feature value;

[0046] The second sequence network is used to perform sequence modeling on the behavior recognition sequence to obtain the behavior feature quantity;

[0047] The attention network is used to fuse the body temperature feature quantity and the behavior feature quantity to obtain the body temperature feature fusion sequence and the behavior feature fusion sequence respectively;

[0048] The decision network model is used to classify diseases according to the input temperature feature fusion sequence and behavior feature fusion sequence to obtain animal disease warning information.

[0049] In summary, the animal disease early warning method and system based on behavior and body temperature dual sequence provided by the present invention, by real-time collection of animal body temperature index sequence and motion index sequence, so as to adopt a dual sequence decision network, integrate the physiological and behavioral characteristics of animals to warn of animal diseases. Specifically, after collecting the animal's body temperature index sequence and motion index sequence, the animal's motion index sequence is input into the behavior recognition network, so as to identify the animal's behavior recognition sequence, which reflects the animal's behavioral characteristics; then the dual sequence decision network is used to fuse the body temperature index sequence and the behavior recognition sequence, so that the animal's behavioral characteristics and physiological characteristics can be combined to determine whether the animal has a disease, and the animal's disease warning information can be identified, so that the animal's disease warning information can be timely and accurately warned. In summary, through the above method, the dual sequence decision network can be used, and the animal's body temperature index sequence and behavior recognition sequence can be used at the same time to complement each other and double-verify, play the role of the dual sequence signal, reduce the error rate of disease warning, and solve the problem of difficulty in timely and accurate warning of animal diseases in the prior art. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the structures shown in these drawings without paying creative work.

[0051] Figure 1 It is a flow chart of an animal disease early warning method based on behavior and body temperature dual sequence provided by an embodiment of the present invention;

[0052] Figure 2 yes Figure 1 A schematic flow chart of a sequence acquisition method provided by the illustrated embodiment;

[0053] Figure 3 yes Figure 1 A flow chart of a behavior recognition method provided by the illustrated embodiment;

[0054] Figure 4 yes Figure 3 A schematic flow chart of a method for acquiring a motion feature fusion sequence provided by the illustrated embodiment;

[0055] Figure 5 yes Figure 1 A schematic flow chart of a dual sequence decision method provided by the illustrated embodiment;

[0056] Figure 6 yes Figure 5A schematic flow chart of a dual sequence fusion method provided in the illustrated embodiment;

[0057] Figure 7 It is a structural schematic diagram of an animal disease early warning system based on behavior and body temperature dual sequences provided by an embodiment of the present invention;

[0058] Figure 8 yes Figure 7 A schematic diagram of the structure of an information collection module provided in the illustrated embodiment;

[0059] Fig. 9 It is a structural diagram of a dual-sequence decision network provided by an embodiment of the present invention.

[0060] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0061] It should be understood that the specific embodiments described herein are only used to explain the present invention, and are not used to limit the present invention.

[0062] The main technical problems solved by the embodiments of the present invention are:

[0063] The prior art provides a solution for early warning of pet diseases by identifying pet behavior, using a computer vision-based method, that is, deploying surveillance cameras within the pet's activity range and using image processing methods to detect the pet's behavior in real time. However, the computer vision-based animal behavior detection method has a limited detection range, and abnormal pet behaviors usually occur in hidden corners. For example, when abnormal licking occurs, animals usually hide themselves to protect themselves. Therefore, it is difficult to determine whether an animal is sick by only detecting its behavior, and it is difficult to provide timely and accurate early warning of animal diseases.

[0064] In order to solve the above problems, the present invention provides an animal disease early warning solution based on behavior and body temperature dual sequences, which collects the animal's body temperature index sequence and motion index sequence in real time, and then uses a behavior recognition network to identify the motion index sequence to obtain the animal's behavior recognition sequence, thereby obtaining the animal's behavioral characteristics; then a dual-sequence decision network is used to fuse the animal's body temperature index sequence and behavior recognition sequence, so that the two sequences complement each other and are doubly verified, and the role of dual-sequence signals is brought into play, which can improve the prediction probability of animal diseases and reduce the error rate of disease early warnings, thereby solving the problem that the existing technology is difficult to provide timely and accurate early warnings of animal diseases.

[0065] To achieve this, see Figure 1 , Figure 1 is a flow chart of an animal disease early warning method based on behavior and body temperature dual sequence provided by an embodiment of the present invention, such as Figure 1 As shown, the animal disease early warning method based on behavior and body temperature dual sequence includes:

[0066] S110: The animal's body temperature index sequence and motion index sequence are acquired in real time. The animal's body temperature index sequence is a basic health attribute of the animal, which reflects the animal's health status. When the animal is sick or infected, the body temperature usually rises. By acquiring the animal's body temperature index sequence, it is possible to accurately and quickly predict whether the animal is sick. The motion index sequence includes a variety of animal motion characteristics, such as motion speed, motion acceleration and direction. By acquiring the animal's motion index sequence, the animal's behavior can be identified.

[0067] Among them, as a preferred embodiment, Figure 2 As shown, the steps of acquiring the animal's body temperature index sequence and movement index sequence in real time include:

[0068] S111: Setting a unique identification of the animal. In the embodiment of the present application, an implantable chip may be provided and implanted into the animal, and an animal ID is established in the chip as a unique identification of the animal.

[0069] S112: Using a body temperature sensor to collect the animal's body temperature index sequence in real time. A body temperature sensor or a body temperature sensing function may be integrated into the above-mentioned implantable chip to collect the animal's body temperature index sequence in real time, which is body temperature data over a period of time.

[0070] S113: Use multiple motion sensors to collect multiple motion information of animals in real time, and fuse multiple motion information to obtain a motion index sequence. In the embodiment of the present application, an acceleration sensor, a gyroscope, and a magnetic field sensor can be set in the above-mentioned implantable chip, or the above-mentioned functions can be integrated to respectively collect information such as acceleration, angular velocity, and direction of the animal, so as to determine the animal's motion direction, speed, and acceleration; in addition, the above-mentioned multiple motion information is fused to obtain a motion index sequence, which is a multi-dimensional vector, and different dimensions can correspond to the above-mentioned speed, acceleration, direction, and other information.

[0071] S114: Establishing a correspondence between the unique identifier and the temperature index sequence and the motion index sequence. By establishing a correspondence between the unique identifier and the temperature index sequence and the motion index sequence, it is possible to measure different animals in a targeted manner and accurately obtain the body temperature and motion information of the animals.

[0072] The technical solution provided by the embodiment of the present application is to set an implantable chip, which integrates a health information collection module and a motion information collection module; the health information collection module includes a body temperature sensor, which collects and records the animal's body temperature index data in real time as a record of the animal's basic health attributes. In addition, the motion information collection module includes an acceleration sensor, a gyroscope and a magnetic field sensor; wherein the acceleration sensor collects the animal's acceleration in real time, the gyroscope collects the animal's angular velocity in real time, and the magnetic field sensor collects the magnetic field signal in real time to determine the animal's movement direction, and the auxiliary model recognizes the animal's behavior.

[0073] After collecting the animal's body temperature index sequence and movement index sequence, Figure 1 The animal disease early warning method shown also includes the following steps:

[0074] S120: Input the animal's motion index sequence into the behavior recognition network to identify the animal's behavior recognition sequence. The behavior recognition network is a neural network model. By using a machine learning algorithm to identify the above motion index sequence, the animal's behavior category can be identified through an N-classification model, that is, the animal's behavior recognition sequence is output. Figure 7 As shown, the behavior recognition network includes a sliding window, a sequence model and a behavior decision model, wherein the sequence model is a recurrent neural network that can process various forms of sequences, and the behavior decision model belongs to a fully connected neural network (MLP) that can perform N classification operations based on the output of the sequence model.

[0075] Specifically, as a preferred embodiment, Figure 3 As shown, the step S120: inputting the animal's motion index sequence into the behavior recognition network to identify the animal's behavior recognition sequence includes:

[0076] S121: Use the sliding window of the behavior recognition network to slice and extract features from the motion index sequence to obtain the motion representation features of the animal. The sliding window is part of the sequence model and is used to slice the motion index sequence and extract the motion representation features contained in the motion index sequence, so as to analyze the animal's behavior using the motion representation features. The sequence model of the behavior recognition network slices the motion index sequence into time using the sliding window method, and the obtained time window is denoted as T t ; In this time window T t The motion representation feature extracted from the inner layer is denoted as F t The motion representation features include the maximum, mean and variance in the time domain, and the distribution histogram features in the frequency domain. By extracting the above features, the behavior recognition network can accurately judge the behavior of animals.

[0077] S122: Use the sequence model of the behavior recognition network to process the motion representation features to obtain a fusion sequence of the animal's motion features. The sequence model includes a long short-term memory model LSTM, and the above-mentioned motion representation features are input into the long short-term memory model LSTM for sequence modeling to obtain the motion feature quantity at each moment, and the motion feature quantity is fused to obtain a motion feature fusion sequence, which can accurately reflect the animal's motion characteristics, and the subsequent behavior recognition network can accurately identify the animal's behavior based on the motion characteristics to obtain an accurate animal behavior recognition sequence.

[0078] Among them, Figure 4 As shown, the steps of processing motion representation features using the sequence model of the behavior recognition network include:

[0079] S1221: Use a sequence model to perform sequence modeling on the motion representation features to obtain the motion feature quantity at each moment in a predetermined time period.

[0080] S1222: Use the sequence model to fuse the motion feature quantities at all times within the predetermined time period to obtain a motion feature fusion sequence.

[0081] The sequence model is designed based on the long short-term memory model LSTM and the attention mechanism. It can use LSTM to perform sequence modeling on the motion representation features to obtain the motion feature quantity at each moment within a predetermined time period; then the motion feature quantities of all moments within the predetermined time period are input into the Attention model of the sequence model for fusion to obtain a motion feature fusion sequence. Because each fused feature in the motion feature fusion sequence is obtained by fusion of the motion feature quantities of all previous moments within the predetermined time period, the motion feature fusion sequence can accurately detect the motion characteristics of animals.

[0082] The data fused by the Attention model are all the motion features at the previous t moments. Specifically, the motion representation feature F t As the input of the sequence model LSTM, sequence modeling is performed. The output of the sequence model at each moment, that is, the motion representation feature is H t , the motion features of the previous t moments are fused through the Attention mechanism as the input Y of the behavior decision model of the behavior recognition network at the t moment t , that is, the motion feature fusion sequence is Y t ; Among them, the calculation method of the motion feature fusion sequence is as follows:

[0083] Y t =∑(γ i *H i )+H i , where γ iis the attention score of the motion feature, γ i =H t *w*H i (where i=0, 1, 2, …, t-1), w is a trainable parameter.

[0084] S123: Use the behavior decision model of the behavior recognition network to identify the behavior type of the motion feature fusion sequence to obtain the animal's behavior recognition sequence. The behavior decision model includes a fully connected network MLP, which can output N classifications to identify multiple behavior types of animals, and the output is the above-mentioned behavior recognition sequence.

[0085] The technical solution provided in the embodiment of the present application can accurately obtain the animal's motion representation features by using the sliding window of the behavior recognition network to segment and extract features from the motion index sequence, and then use the sequence model to fuse the motion representation features to obtain a motion feature fusion sequence. Because each fusion feature in the motion feature fusion sequence is obtained by fusion of the motion feature quantities of all previous moments within a predetermined period of time, the motion feature fusion sequence can accurately reflect the animal's motion changes, and the behavior decision model of the behavior recognition network can be used to realize the recognition of various types of animal behaviors based on the above motion feature fusion sequence, thereby obtaining the above behavior recognition sequence.

[0086] In addition, after the animal's behavior recognition sequence is obtained, the behavior recognition sequence and the temperature index sequence need to be input into the dual sequence decision network for fusion. Specifically, Figure 1 The animal disease early warning method shown also includes the following steps:

[0087] S130: Using a dual sequence decision network to fuse the body temperature index sequence and the behavior recognition sequence, and identifying the animal's disease warning information based on the fused body temperature index sequence and behavior recognition sequence.

[0088] Combination Fig. 9 It can be seen from the double sequence decision network shown that the double sequence decision network includes two sequence networks (i.e., the first sequence network 131 and the second sequence network 132) and two attention networks 133 and a common decision network model 134. The body temperature index sequence and the behavior recognition sequence are modeled by the two sequence networks to obtain body temperature feature quantities and behavior feature quantities, and then the feature quantities of the body temperature and behavior are fused and then input into the decision network model of the double sequence decision network for disease classification, so as to detect animal diseases in combination with behavior and body temperature.

[0089] Specifically, as a preferred embodiment, Figure 5As shown, the dual sequence decision network is used to fuse the body temperature index sequence and the behavior recognition sequence. The steps of identifying the disease warning information of the animal according to the fused body temperature index sequence and the behavior recognition sequence include:

[0090] S131: Use the first sequence network of the dual sequence decision network to perform sequence modeling on the temperature index sequence to obtain the temperature feature. The first sequence network is designed based on the long short-term memory model LSTM, which can perform sequence modeling on the above temperature index sequence, so as to accurately extract the temperature feature, which can reflect the temperature characteristics of the animal. The temperature index sequence of the animal is recorded as T i (i=0,1,2,…,t).

[0091] S132: Use the second sequence network of the dual sequence decision network to perform sequence modeling on the behavior recognition sequence to obtain the behavior feature quantity. The second sequence network is designed based on the long short-term memory model LSTM, which can perform sequence modeling on the above behavior recognition sequence, thereby accurately extracting the behavior feature quantity, which can reflect the behavior characteristics of the animal. The behavior recognition sequence is the behavior recognition sequence identified by the above behavior recognition network, denoted as A i (i=0,1,2,…,t).

[0092] S133: Use the attention network of the dual-sequence decision network to fuse the body temperature feature quantity and the behavior feature quantity, and obtain the body temperature feature fusion sequence and the behavior feature fusion sequence respectively. The body temperature feature fusion sequence in the embodiment of the present application is fused with the behavior feature quantity, and the behavior feature fusion sequence is fused with the body temperature feature quantity, so that the two sequences can complement each other and double-verify. Using the above-mentioned attention network Attention mechanism to fuse the above-mentioned body temperature feature quantity and behavior feature quantity can further play the role of the dual sequence signal and reduce the error rate of disease warning.

[0093] S134: Input the body temperature feature fusion sequence and the behavior feature fusion sequence together into the decision network model of the dual sequence decision network for disease classification to obtain disease warning information of the animal.

[0094] The decision network model uses a multi-layer perceptron MLP, which is a fully connected network that can process the above-mentioned temperature feature fusion sequence and behavior feature fusion sequence together, and then perform binary or multi-classification to determine whether the animal is sick and accurately obtain disease warning information. Because the decision is made using the temperature feature fusion sequence and the behavior feature fusion sequence, it can be double-verified through the animal's physiological and behavioral characteristics, reducing the error rate of disease warnings.

[0095] Among them, as a preferred embodiment, Figure 6As shown, the above step S133: using the attention network of the dual sequence decision network to fuse the body temperature feature quantity and the body temperature feature quantity respectively, and the steps of obtaining the body temperature feature fusion sequence and the behavior feature fusion sequence respectively include:

[0096] S1331: Control the attention network according to the formula YT t =∑(α i *HA i )+HT t , fuse the body temperature feature and behavior feature to obtain the body temperature feature fusion sequence; among them, YT t is the temperature feature fusion sequence, α i is the attention network score of body temperature, HA i is the body temperature characteristic at time i, HT t are all body temperature characteristics in period t; and,

[0097] S1332: Control the attention network according to formula YA t =∑(β i *HT i )+HA t , fuse the temperature feature and the behavior feature to obtain the behavior feature fusion sequence; among them, YA t is the behavior feature fusion sequence, β i is the attention network score of the behavior, HT i is the body temperature characteristic at time i, HA t is the characteristic quantity of all behaviors in period t.

[0098] In the present embodiment, the animal's body temperature index sequence is recorded as T i (i=0,1,2,…,t), the animal’s behavioral index sequence is recorded as A i (i=0,1,2,…,t), and use the LSTM mechanism of the sequence model to model the temperature index sequence and behavior index sequence respectively, and the outputs are the temperature feature HT i and behavioral characteristics HA i The dual-sequence Attention mechanism can jointly assist in decision-making behavioral warnings by integrating body temperature and behavior sequences, thereby improving the accuracy of animal disease warnings.

[0099] Body temperature feature fusion sequence: YT t =∑(α i *HA i )+HT t ;

[0100] Behavioral fusion sequence: YA t =∑(β i *HT i )+HAt ;

[0101] Among them, α i is the attention score of the body temperature fusion feature, the α i =HT t *WT*HA i β i Same as Attention score, β i =HA t *WA*HT i The above method can quantify the animal's body temperature and behavior characteristics, and fuse the body temperature characteristics into the sequence YT t and behavioral feature fusion sequence YA t The two sequences are used as the input of the fully connected network of MLP to provide early warning of animal diseases. WT and WA are trainable parameters of body temperature and behavioral characteristics, respectively.

[0102] In summary, the animal disease early warning method provided by the above embodiment of the present invention collects the animal's body temperature index sequence and motion index sequence in real time, thereby adopting a dual sequence decision network to fuse the animal's physiological and behavioral characteristics to warn of animal diseases. Specifically, after collecting the animal's body temperature index sequence and motion index sequence, the animal's motion index sequence is input into the behavior recognition network to identify the animal's behavior recognition sequence, which reflects the animal's behavioral characteristics; then the dual sequence decision network is used to fuse the body temperature index sequence and the behavior recognition sequence, so that the animal's behavioral characteristics and physiological characteristics can be combined to determine whether the animal has a disease, and the animal's disease warning information can be identified, so that the animal's disease warning information can be timely and accurately warned. In summary, through the above method, the dual sequence decision network can be used, and the animal's body temperature index sequence and behavior recognition sequence can be used at the same time to complement each other and double-verify, play the role of the dual sequence signal, reduce the error rate of disease warning, and solve the problem of difficulty in timely and accurate warning of animal diseases in the prior art.

[0103] Based on the same concept of the above method embodiment, the embodiment of the present invention also proposes an animal disease early warning system based on a dual sequence of behavior and body temperature, which is used to implement the above method of the present invention. Since the principles and methods of solving the problems in the system embodiment are similar, it has at least all the beneficial effects brought by the technical solutions of the above embodiments, which will not be repeated here one by one.

[0104] See also Figure 7 , Figure 7 Schematic diagram of the structure of an animal disease early warning system based on behavior and body temperature dual sequence provided by an embodiment of the present invention. Figure 7 As shown, the animal disease early warning system based on behavior and body temperature dual sequence includes:

[0105] Information collection module 110, behavior recognition network 120 and dual sequence decision network 130; wherein,

[0106] The information collection module 110 is used to collect the animal's body temperature index sequence and movement index sequence in real time;

[0107] The behavior recognition network 120 is used to input the animal's motion index sequence and identify the animal's behavior recognition sequence;

[0108] The dual sequence decision network 130 is used to fuse the body temperature index sequence and the behavior recognition sequence, and identify and obtain the animal's disease warning information based on the fused body temperature index sequence and behavior recognition sequence.

[0109] In summary, the animal disease early warning system provided by the above embodiment of the present invention collects the animal's body temperature index sequence and motion index sequence in real time through the information acquisition module 110, and adopts the dual sequence decision network 130 to fuse the animal's physiological and behavioral characteristics to warn of animal diseases. Specifically, after the information acquisition module 110 collects the animal's body temperature index sequence and motion index sequence, the animal's motion index sequence is input into the behavior recognition network 120, so as to identify the animal's behavior recognition sequence, which reflects the animal's behavioral characteristics; then the dual sequence decision network 130 is used to fuse the body temperature index sequence and the behavior recognition sequence, so that the animal's behavioral characteristics and physiological characteristics can be combined to determine whether the animal has a disease, and the animal's disease early warning information can be identified, so that the animal's disease early warning information can be timely and accurately warned. In summary, the dual sequence decision network 130 can be used in the above manner, and the animal's body temperature index sequence and behavior recognition sequence can be used at the same time to complement each other and double-verify, play the role of the dual sequence signal, reduce the error rate of disease early warning, and solve the problem that it is difficult to warn animal diseases in a timely and accurate manner in the prior art.

[0110] As a preferred embodiment, Figure 8 As shown, the information collection module 110 in the above-mentioned animal disease early warning system includes:

[0111] The identification setting submodule 111 is used to set a unique identification of the animal;

[0112] The body temperature index sequence collection submodule 112 is used to collect the body temperature index sequence of the animal in real time using a body temperature sensor;

[0113] The motion information collection submodule 113 is used to use a variety of motion sensors to collect a variety of motion information of animals in real time, and fuse the multiple motion information to obtain a motion index sequence;

[0114] The relationship establishing submodule 114 is used to establish corresponding relationships between the unique identifier and the body temperature index sequence and the exercise index sequence.

[0115] In addition, as a preferred embodiment, Figure 7 As shown, in the above-mentioned animal disease early warning system, the behavior recognition network 120 includes:

[0116] The sliding window 121 is used to segment and extract features from the motion index sequence to obtain the motion representation features of the animal;

[0117] A sequence model 122, used for processing motion representation features to obtain a fusion sequence of motion features of the animal;

[0118] The behavior decision model 123 is used to identify the behavior type of the motion feature fusion sequence to obtain the animal's behavior identification sequence.

[0119] In addition, as a preferred embodiment, Figure 7 As shown, the above-mentioned dual sequence decision network 130 includes:

[0120] The first sequence network 131 is used to perform sequence modeling on the temperature index sequence to obtain a temperature feature value;

[0121] The second sequence network 132 is used to perform sequence modeling on the behavior recognition sequence to obtain the behavior feature quantity;

[0122] The attention network 133 is used to fuse the body temperature feature quantity and the behavior feature quantity to obtain a body temperature feature fusion sequence and a behavior feature fusion sequence respectively; Figure 7 As shown, the attention network 133 includes an attention network 1 corresponding to the first sequence network 131 and an attention network 2 corresponding to the second sequence network 132.

[0123] The decision network model 134 is used to classify diseases according to the input temperature feature fusion sequence and behavior feature fusion sequence to obtain animal disease warning information.

[0124] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0125] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0126] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0127] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0128] It should be noted that in the claims, any reference signs placed between brackets shall not be construed as limiting the claims. The word "comprising" does not exclude the presence of components or steps not listed in the claim. The word "a" or "an" preceding a component does not exclude the presence of a plurality of such components. The invention may be implemented by means of hardware comprising several different components and by means of a suitably programmed computer. In a unit claim enumerating several means, several of these means may be embodied by the same item of hardware. The use of the words first, second, and third etc. does not indicate any order. These words may be interpreted as names.

[0129] Although the preferred embodiments of the present invention have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0130] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. An animal disease early warning method based on behavior and body temperature dual sequence, characterized in that: include: The animal's body temperature index sequence and movement index sequence are collected in real time; Inputting the animal's motion index sequence into a behavior recognition network to obtain the animal's behavior recognition sequence; Using a dual sequence decision network to fuse the body temperature index sequence and the behavior recognition sequence, and identifying disease warning information of the animal according to the fused body temperature index sequence and behavior recognition sequence; The step of fusing the body temperature index sequence and the behavior recognition sequence using a dual sequence decision network, and identifying and obtaining the disease warning information of the animal according to the fused body temperature index sequence and behavior recognition sequence, comprises: Use the first sequence network of the dual sequence decision network to perform sequence modeling on the body temperature index sequence to obtain a body temperature feature value; Using the second sequence network of the dual sequence decision network to perform sequence modeling on the behavior recognition sequence to obtain a behavior feature quantity; The attention network of the dual sequence decision network is used to fuse the body temperature feature quantity and the behavior feature quantity to obtain a body temperature feature fusion sequence and a behavior feature fusion sequence respectively; Inputting the body temperature feature fusion sequence and the behavior feature fusion sequence into the decision network model of the dual sequence decision network to perform disease classification, thereby obtaining disease warning information of the animal; The step of fusing the body temperature feature quantity and the behavior feature quantity using the attention network of the dual sequence decision network to obtain a body temperature feature fusion sequence and a behavior feature fusion sequence respectively comprises: Controlling the attention network to fuse the body temperature feature and the behavior feature according to the formula: YT t=∑(αi*HA i)+HT t to obtain the body temperature feature fusion sequence; wherein YT t is the body temperature feature fusion sequence, αi is the attention network score of body temperature, HA i is the behavior feature at time i, and HT t is all body temperature features in time period t; and, Control the attention network to fuse the body temperature feature and the behavior feature according to the formula: YAt=∑(βi*HTi)+HAt to obtain the behavior feature fusion sequence; wherein YAt is the behavior feature fusion sequence, βi is the attention network score of the behavior, HTi is the body temperature feature at time i, and HAt is all the behavior feature in time period t; αi=HTt*WT*HAi; βi=HAt*WA*HTi; WT and WA are trainable parameters for body temperature and behavioral characteristics, respectively.

2. The animal disease early warning method according to claim 1, characterized in that: The step of acquiring the animal's body temperature index sequence and movement index sequence in real time includes: Setting a unique identifier for the animal; Using a body temperature sensor to collect a body temperature index sequence of the animal in real time; Using multiple motion sensors to collect multiple motion information of the animal in real time, and fusing the multiple motion information to obtain the motion index sequence; A corresponding relationship between the unique identifier and the body temperature index sequence and the exercise index sequence is established.

3. The animal disease early warning method according to claim 1, characterized in that: The step of inputting the animal's motion index sequence into a behavior recognition network to identify and obtain the animal's behavior recognition sequence comprises: Using the sliding window of the behavior recognition network to segment and extract features from the motion index sequence, to obtain the motion representation features of the animal; Processing the motion representation features using the sequence model of the behavior recognition network to obtain a fusion sequence of motion features of the animal; The behavior decision model of the behavior recognition network is used to perform behavior type recognition on the motion feature fusion sequence to obtain the behavior recognition sequence of the animal.

4. The animal disease early warning method according to claim 3, characterized in that: The step of processing the motion representation features using a sequence model of an action recognition network comprises: Using the sequence model to perform sequence modeling on the motion representation features, to obtain the motion feature quantity at each moment in a predetermined time period; The sequence model is used to fuse the motion feature quantities at all moments within the predetermined time period to obtain the motion feature fusion sequence.

5. An animal disease early warning system based on behavior and body temperature dual sequences, characterized in that: include: Information collection module, behavior recognition network and dual sequence decision network; among them, The information collection module is used to collect the animal's body temperature index sequence and movement index sequence in real time; The behavior recognition network is used to input the movement index sequence of the animal and identify and obtain the behavior recognition sequence of the animal; The dual sequence decision network is used to fuse the body temperature index sequence and the behavior recognition sequence, and identify and obtain disease warning information of the animal according to the fused body temperature index sequence and behavior recognition sequence; The dual sequence decision network comprises: A first sequence network is used to perform sequence modeling on the body temperature index sequence to obtain a body temperature feature value; A second sequence network is used to perform sequence modeling on the behavior recognition sequence to obtain a behavior feature quantity; An attention network is used to fuse the body temperature feature quantity and the behavior feature quantity to obtain a body temperature feature fusion sequence and a behavior feature fusion sequence respectively; the attention network is controlled to fuse the body temperature feature quantity and the behavior feature quantity according to the formula: YT t=∑(αi*HA i)+HT t to obtain the body temperature feature fusion sequence; wherein YT t is the body temperature feature fusion sequence, αi is the attention network score of the body temperature, HA i is the behavior feature quantity at time i, and HT t is all body temperature feature quantities in time period t; and, the attention network is controlled to fuse the body temperature feature quantity and the behavior feature quantity according to the formula: YAt=∑(βi*HTi)+HAt to obtain the behavior feature fusion sequence; wherein YAt is the behavior feature fusion sequence, βi is the attention network score of the behavior, HTi is the body temperature feature quantity at time i, and HAt is all behavior feature quantities in time period t; αi=HTt*WT*HAi; βi=HAt*WA*HTi; WT and WA are trainable parameters of body temperature and behavioral characteristics, respectively; The decision network model is used to classify diseases according to the input temperature feature fusion sequence and the behavior feature fusion sequence to obtain disease warning information of the animal.

6. The animal disease early warning system according to claim 5, characterized in that: The information collection module comprises: An identification setting submodule, used to set a unique identification of the animal; A body temperature index sequence acquisition submodule, used to use a body temperature sensor to acquire the body temperature index sequence of the animal in real time; A motion information collection submodule, used to use a variety of motion sensors to collect a variety of motion information of the animal in real time, and fuse the multiple motion information to obtain the motion index sequence; The relationship establishing submodule is used to establish the corresponding relationship between the unique identifier and the body temperature index sequence and the exercise index sequence respectively.

7. The animal disease early warning system according to claim 6, characterized in that: The behavior recognition network includes: A sliding window is used to segment and extract features from the motion index sequence to obtain motion representation features of the animal; A sequence model is used to process the motion representation features to obtain a fusion sequence of motion features of the animal; The behavior decision model is used to identify the behavior type of the motion feature fusion sequence to obtain the behavior identification sequence of the animal.

Images