JP6635995B2 - Behavior identification device, behavior identification method, and program - Google Patents

Description

  The present invention relates to a behavior specifying device, a behavior specifying method, and a program.

  The milking period of a dairy cow is after the dairy cow has been mated (artificial insemination) to produce a calf (after calving), and this period is called the lactation period. It is known that the lactation period is about 305 days. In general, milking is stopped about 60 days before the scheduled delivery date in preparation for the next delivery. This 60-day period is called the dry period. Dairy cows repeat the calving, lactation and dry periods every year.

  Dairy farmers who raise dairy cows need to breed (artificial insemination) dairy cows every year during the dry period in order to continuously milk from dairy cows. In mating dairy cows (artificial insemination), it is necessary to discover the estrus of dairy cows. In order to detect the estrus of dairy cows, external signs of dairy cows are confirmed, and actions that are frequently observed during estrus are confirmed. In addition, management of the estrus time using a breeding record using a ledger is also performed.

Ryo Abe, `` Basic Seminar on Agriculture, Basics of Livestock Raising '', New Edition, Agricultural and Fishing Village Cultural Association, April 2008, p.109, p.122-124.

  Here, for example, on a large-scale dairy farm where the number of breeding animals is several thousand or more, it is a heavy burden for dairy farmers to confirm external signs and behavior of each dairy cow. . On the other hand, for example, if the behavior of the dairy cow being bred is specified, and the confirmation of the behavior often observed at the time of estrus can be easily performed, the burden on the dairy farmer can be reduced.

  Further, by specifying the behavior of the dairy cow, it is possible not only to detect estrus but also to detect an abnormal behavior due to illness or injury, which can contribute to health management of the dairy cow.

  Embodiments of the present invention have been made in view of the above points, and have as its object to specify the behavior of livestock.

  In order to solve the above problem, an embodiment of the present invention is a behavior identification device that identifies the behavior of livestock, the acceleration data measured by an acceleration sensor attached to the livestock, the pressure data measured by a pressure sensor Storage means for storing the data, at least one acceleration data and one or more atmospheric pressure data during a predetermined time from the storage means, and the one or more acceleration data obtained by the obtaining means An index value calculating means for calculating a predetermined index value from the, the index value calculated by the index value calculating means, and a specific model for specifying the behavior of the livestock created in advance, the livestock, A first specifying means for specifying the action of the first, and, if the action specified by the first specifying means is a predetermined action, the one or more atmospheric pressure data obtained by the obtaining means. Based on, having a second specifying means for specifying the behavior of the livestock.

  The behavior of livestock can be specified.

It is a figure showing an example of the whole composition of the action identification system concerning this embodiment. It is a figure explaining an example of action specification of a cow. FIG. 4 is a diagram illustrating an example of measurement data stored in a measurement data storage unit. It is a figure showing an example of an action specific model. It is a figure showing an example of functional composition of an action identification processing part concerning this embodiment. It is a flowchart which shows an example of the action identification processing by action intensity analysis. It is a flowchart which shows an example of the action identification processing by atmospheric pressure analysis. It is a figure explaining an example of action specification by atmospheric pressure analysis. It is a figure explaining an example in the case of specifying an action by barometric pressure analysis in a free burn type barn.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings. Hereinafter, a case where the behavior of a cow is specified as an example of livestock will be described. However, livestock is not limited to cattle.

<Overall configuration>
First, the overall configuration of the behavior identification system 1 according to the present embodiment will be described with reference to FIG. FIG. 1 is a diagram illustrating an example of an overall configuration of a behavior identification system 1 according to the present embodiment.

  As shown in FIG. 1, a behavior identification system 1 according to the present embodiment includes a behavior identification device 10 for identifying a behavior of a cow, one or more tags 20 attached to the cow, and a reference for measuring a barometric pressure serving as a reference. A pressure sensor 30 is included. In addition, it is preferable that the tag 20 is fixedly attached to the neck of the cow.

  The tag 20 is a device attached to a cow. One tag 20 is attached to one cow. The tag 20 includes an acceleration sensor that measures the acceleration (three-axis acceleration of the X axis, the Y axis, and the Z axis) of the cow wearing the tag 20, and a barometric pressure sensor that measures the barometric pressure.

  The tag 20 transmits measurement data including the acceleration sensor value measured by the acceleration sensor and the barometric pressure sensor value measured by the barometric sensor to the behavior identification device 10 at predetermined time intervals (for example, every two seconds). The measurement data transmitted to the behavior identification device 10 is accumulated (stored) in a measurement data storage unit 200 described later.

  The reference atmospheric pressure sensor 30 is installed at a predetermined position in the barn (for example, on the ground in the barn) and measures the reference barometric pressure. The reference atmospheric pressure sensor 30 transmits the reference atmospheric pressure data including the reference atmospheric pressure sensor value indicating the measured atmospheric pressure to the behavior specifying device 10 at every predetermined time (for example, every 2 seconds). The reference atmospheric pressure data transmitted to the action specifying device 10 is stored (stored) in a reference atmospheric pressure data storage unit 400 described later.

  The behavior identification device 10 is one or more computers that identify the behavior of a cow. The behavior specifying device 10 includes a behavior specifying processing unit 100, a measurement data storage unit 200, a behavior specification model 300, a reference atmospheric pressure data storage unit 400, and a behavior data storage unit 500.

  The behavior identification processing unit 100 identifies the behavior of the cow by the operation intensity analysis based on the measurement data stored in the measurement data storage unit 200 and the behavior identification model 300. In addition, when a predetermined action is specified by the action intensity analysis, the action specifying processing unit 100 performs the measurement data stored in the measurement data storage unit 200 and the reference pressure data stored in the reference pressure data storage unit 400. Based on the behavior data stored in the behavior data storage unit 500, more accurate cow behavior is specified by barometric pressure analysis. The behavior identification processing unit 100 is realized by processing that causes one or more programs installed in the behavior identification device 10 to be executed by a CPU (Central Processing Unit) or the like.

  The measurement data storage unit 200 stores the measurement data received from the tag 20. The measurement data storage unit 200 can be realized by using an auxiliary storage device such as a hard disk drive (HDD) or a solid state drive (SSD). The measurement data storage unit 200 stores a plurality of measurement data every predetermined time (for example, every two seconds).

  The action specifying model 300 is a model for specifying an action from an acceleration sensor value included in the measurement data. The action specifying model 300 is created in advance by a machine learning method such as an SVM (Support Vector Machine) according to the type of livestock and the type of action to be specified. The behavior identification model 300 is stored in, for example, an auxiliary storage device such as an HDD or an SSD.

  The reference pressure data storage unit 400 stores the reference pressure data received from the reference pressure sensor 30. The reference atmospheric pressure data storage unit 400 can be realized using an auxiliary storage device such as an HDD or an SSD. The reference atmospheric pressure data storage unit 400 stores a plurality of reference atmospheric pressure data every predetermined time (for example, every two seconds).

  The behavior data storage unit 500 stores behavior data indicating the behavior specified by the behavior specification processing unit 100. The behavior data storage unit 500 can be realized using an auxiliary storage device such as an HDD or an SSD. The behavior data storage unit 500 stores behavior data indicating the behavior identified by the behavior identification processing unit 100 at every predetermined time (for example, every 10 minutes).

  The configuration of the behavior identification system 1 shown in FIG. 1 is an example, and another configuration may be used. For example, the behavior identification device 10 may be configured by a plurality of computers. Further, for example, a device (such as a cloud server) connected to the behavior identification device 10 via a network may have a part of the functions of the behavior identification processing unit 100. Further, for example, instead of the tag 20, an acceleration sensor and a barometric pressure sensor may be separately mounted on the cow.

<Identification of cow behavior>
Here, the behavior of the cow identified by the behavior identification system 1 according to the present embodiment and an outline of the identification method will be described with reference to FIG. FIG. 2 is a diagram illustrating an example of identifying the behavior of a cow.

  First, it is assumed that the behaviors of the cow identified by the behavior identification system 1 according to the present embodiment include, in ascending order of movement, four “standing”, “lying down”, “rubbing”, and “activity”. Therefore, in the present embodiment, it is specified which one of the four actions of “standing”, “lying down”, “rubbing” and “activity” is performed by the cow. It should be noted that the magnitudes of the “standing” and “lying” operations may be reversed. In other words, “lying down”, “standing”, “rubbing”, and “activity” may be used in the order of smaller motion.

  "Standing" refers to a state where a cow is standing. "Lying down" refers to a state in which a cow is lying down. Note that “stand up” and “lying down” are states in which the cow is not performing any motion (in other words, a state in which the cow is standing and standing still or a state in which the cow is lying still).

  "Rumming" refers to a state in which a cow is performing rumination (returning food once swallowed to the mouth and chewing it again). "Activity" refers to a state in which a cow is performing a predetermined operation, such as a state in which a cow is walking, a state in which a cow is eating food, a state in which a cow is drinking water, and the like. It is.

  Whether the cow is performing any of the “standing”, “lying”, “ruminating”, and “activity” is specified by the motion intensity analysis. In the motion intensity analysis, based on the acceleration sensor value included in the measurement data and the behavior identification model 300, it is determined whether the behavior of the cow is “standing”, “lying down”, “ruminating”, or “activity”. Identify. More specifically, in the motion intensity analysis, the acceleration sensor value during a predetermined time (for example, 10 minutes) is converted from a predetermined index value (the maximum value of the swing width of the L2 norm of the acceleration sensor value during the predetermined time and The L2 norm of the acceleration sensor value is calculated. Note that the maximum value of the swing width of the L2 norm is the difference between the average value of the L2 norm of the acceleration sensor value during a predetermined time and the value of the L2 norm of each acceleration sensor value during the predetermined time. The absolute value is the maximum value.

  Then, in the behavior identification model 300 classified into four regions of “standing”, “lying”, “rubbing”, and “activity”, a point indicating the calculated index value is included in any of these four regions. Identify actions.

  In addition, when “standing”, “lying” or “rubbing” is specified by the motion intensity analysis, more accurate behavior is specified by the atmospheric pressure analysis. This is because the behavior intensity analysis may not be able to accurately identify the cow's posture during the behavior identified using the behavior identification model 300.

  In the barometric pressure analysis, a barometric pressure difference minimum value indicating a minimum value of a difference between a barometric pressure sensor value included in each measurement data and a reference barometric pressure sensor value included in each reference barometric pressure data during a predetermined time (for example, 5 minutes) is calculated. The behavior of the cow is specified based on a predetermined threshold value.

  More specifically, for example, when the action 10 minutes ago (that is, the action immediately before) is “lying down”, the first pressure difference minimum value from 5 minutes ago to the present time is 10 minutes. It is determined whether or not the second air pressure difference minimum value between the minutes before and five minutes before has decreased by a predetermined threshold or more. Then, when the first pressure difference minimum value decreases by a predetermined threshold value or more from the second pressure difference minimum value, it is specified as “standing”. On the other hand, if the first air pressure difference minimum value has not decreased from the second air pressure difference minimum value by a predetermined threshold or more, it is specified that the action is the same as the immediately preceding action (that is, “lying down”).

  Further, for example, when the action 10 minutes ago (ie, the action immediately before) is other than “lying down” (ie, “upright” or “rubbing”), the first pressure difference minimum value Is determined to have increased from the second minimum pressure difference by a predetermined threshold or more. Then, when the first minimum pressure difference is greater than the second minimum pressure difference by a predetermined threshold or more, it is specified as “lying down”. On the other hand, if the first pressure difference minimum value does not increase by more than the predetermined threshold value from the second pressure difference minimum value, it is specified that the action is the same as the immediately preceding action (ie, “stand up” or “ruminate”). I do.

  As described above, the behavior identification system 1 according to the present embodiment uses the acceleration sensor value received from the tag 20 attached to the cow to determine four behaviors of “standing”, “lying”, “ruminating”, and “activity”. To identify. In addition, the behavior identification system 1 according to the present embodiment, when the behavior of the cow is specified to be "standing", "lying down", or "ruminating", more information from the barometric pressure sensor value and the reference barometric pressure sensor value Identify exact behavior.

  Thereby, it is possible to specify whether the behavior of the cow is “standing”, “lying”, “ruminating”, or “activity”. At this time, “standing”, “lying” or “ruminating” can be specified with high accuracy by the atmospheric pressure analysis. Moreover, according to the barometric pressure analysis according to the present embodiment, even in a barn having a hill or the like such as a free barn barn, for example, “stand”, “lying”, or “ruminating” is specified with high accuracy. be able to.

  Therefore, by displaying the specified behavior of the cow on, for example, the display device (display or the like) of the behavior identification device 10, it is possible to easily confirm the behavior that is often seen during estrus. In addition, it leads to early detection of abnormal behavior due to illness, injury, and the like, and health management of cattle can be easily performed.

<Measurement data stored in measurement data storage unit 200>
Here, the measurement data stored in the measurement data storage unit 200 according to the present embodiment will be described with reference to FIG. FIG. 3 is a diagram illustrating an example of the measurement data stored in the measurement data storage unit 200 according to the present embodiment. Note that when the measurement data is received from the tag 20, the behavior identification device 10 may cause the behavior identification processing unit 100 to store (accumulate) the received measurement data in the measurement data storage unit 200.

  As shown in FIG. 3, the measurement data storage unit 200 stores one or more measurement data for each tag ID for identifying a tag. Since one tag 20 is attached to one cow, the tag ID may be information for identifying a cow (individual cow identification information).

  The measurement data includes a date and time, an acceleration sensor value, and a barometric pressure sensor value. The date and time is, for example, the date and time when the tag 20 transmitted the measurement data. Note that the date and time may be the date and time when the behavior identification device 10 receives the measurement data.

  The acceleration sensor value is a value of an acceleration measured by an acceleration sensor included in the tag 20. The acceleration sensor value includes an X component indicating an acceleration component in the X-axis direction, a Y component indicating an acceleration component in the Y-axis direction, and a Z component indicating an acceleration component in the Z-axis direction. For example, the measurement data at the date and time “t1” includes an X component “X1”, a Y component “Y1”, and a Z component “Z1” of the acceleration sensor value. The barometric pressure sensor value is a barometric pressure value measured by a barometric pressure sensor included in the tag 20.

  As described above, the measurement data stored in the measurement data storage unit 200 according to the present embodiment accumulates (stores) the measurement data including the date and time, the acceleration sensor value, and the barometric pressure sensor value for each tag ID. ) Has been.

<Behavior identification model 300>
Here, an action specifying model 300 for specifying an action from the acceleration sensor value included in the measurement data will be described with reference to FIG. FIG. 4 is a diagram illustrating an example of the behavior identification model 300.

  As shown in FIG. 4, the behavior identification model 300 has a horizontal axis representing the maximum value of the amplitude of the L2 norm of the acceleration sensor value during a predetermined time (for example, 10 minutes), and a vertical axis representing the standard deviation of the L2 norm. In this case, the relationship is shown as a relationship graph showing the relationship between the maximum value and the standard deviation of the swing width and the action.

  For example, when the maximum value and the standard deviation of the swing width of the L2 norm of the acceleration sensor value for a certain 10 minutes are included in the area D1, the behavior of the cow for the 10 minutes is specified as “standing”. In addition, for example, when the maximum value and the standard deviation of the amplitude of the L2 norm of the acceleration sensor value for a certain 10 minutes are included in the area D2, the behavior of the cow during the 10 minutes is specified as “lying down”.

  Similarly, for example, when the maximum value and the standard deviation of the amplitude of the L2 norm of the acceleration sensor value for a certain 10-minute period are included in the area D3, the behavior of the cow for the 10-minute period is specified as "ruminating". The area D3 specified as “ruminating” further includes a case where the cow performs a ruminating operation in a state where the cow is standing (in this case, referred to as “ruminating during standing”), and a case where the cow is lying in a state where the cow is lying. An operation (in this case, referred to as “lying in bed”) is performed.

  Similarly, for example, when the maximum value and the standard deviation of the amplitude of the L2 norm of the acceleration sensor value for a certain 10 minutes are included in the area D4, the behavior of the cow during the 10 minutes is specified as “activity”. .

  Note that, in the example illustrated in FIG. 4, a case where the regions D <b> 1 to D <b> 4 of the behavior identification model 300 do not overlap each other is shown. May exist.

  As described above, the behavior identification model 300 is a model in which a region representing the relationship between the maximum value and the standard deviation of the L2 norm of the acceleration sensor value during a predetermined time (for example, 10 minutes) and the behavior of the cow is defined. is there. Such an action specifying model 300 is created in advance by a machine learning method such as SVM. The SVM is an example, and may be created by various machine learning methods such as a neural network. Further, for example, even when the cow performs the same operation, the specific acceleration sensor value differs depending on the type of the acceleration sensor and the like. However, the positional relationship between the areas D1 to D4 indicating the respective actions is substantially fixed.

  Hereinafter, the maximum value of the swing width of the L2 norm is also simply referred to as “the maximum value of the L2 norm” or “the maximum value”.

<Functional configuration of action identification processing unit 100>
Next, a functional configuration of the behavior identification processing unit 100 according to the present embodiment will be described with reference to FIG. FIG. 5 is a diagram illustrating an example of a functional configuration of the behavior identification processing unit 100 according to the present embodiment.

  As illustrated in FIG. 5, the behavior identification processing unit 100 includes an acquisition unit 101, a preprocessing unit 102, an index value calculation unit 103, a minimum difference value calculation unit 104, and an activity identification unit 105.

  The acquisition unit 101 acquires measurement data from the measurement data storage unit 200. At this time, the acquisition unit 101 acquires measurement data for a predetermined time (for example, 10 minutes) from the measurement data storage unit 200 for each tag ID.

  Further, the acquiring unit 101 acquires the reference atmospheric pressure data from the reference atmospheric pressure data storage unit 400. At this time, the acquiring unit 101 acquires the reference atmospheric pressure data during the predetermined time.

  Further, the acquiring unit 101 acquires the latest behavior data (that is, behavior data indicating the behavior specified immediately before) from the behavior data stored in the behavior data storage unit 500.

  The preprocessing unit 102 performs preprocessing on the measurement data acquired by the acquisition unit 101. The pre-processing includes, for example, a process of complementing loss (resampling) of measurement data, a process of removing noise, and the like. The preprocessing unit 102 calculates the L2 norm of the acceleration sensor value included in the measurement data after the loss complement, and performs a noise removal process using the calculated L2 norm.

  The index value calculation unit 103 calculates an index value from the L2 norm of the acceleration sensor value included in the measurement data after the preprocessing performed by the preprocessing unit 102. For example, the index value calculation unit 103 calculates the maximum value and the standard deviation from the L2 norm of the acceleration sensor value included in the measurement data after the preprocessing as the index value.

  The minimum difference calculating unit 104 calculates the minimum pressure difference from the measurement data acquired by the acquiring unit 101 and the reference atmospheric pressure data. That is, the minimum difference calculating unit 104 calculates, for example, the first air pressure difference minimum value in the latter 5 minutes and the first air pressure difference in the first 5 minutes from the measurement data for 10 minutes and the reference atmospheric pressure data acquired by the acquiring unit 101. A second pressure difference minimum value is calculated.

  The minimum value of the pressure difference is the minimum value of the difference between the pressure sensor value included in each measurement data during a predetermined time and each reference pressure sensor value included in each reference pressure data during the predetermined time. That is.

  The action specifying unit 105 specifies four actions (“stand up”, “lying down”, “ruminating”, and “activity”) from the maximum value and the standard deviation calculated by the index value calculating unit 103 and the action specifying model 300. (Specify the action by the action intensity analysis).

  Then, the action specifying unit 105 causes the action data storage unit 500 to store action data indicating the specified action. The behavior specifying unit 105 may display information indicating the specified behavior on, for example, a display device such as a display, or another device (for example, a PC, a smartphone, or a tablet) connected to the behavior specifying device 10. Terminal).

  The action specifying unit 105 may specify the above four actions based on the maximum value and the standard deviation and a processing result of a program or the like that generates data equivalent to the action specifying model 300.

  Further, the behavior specifying unit 105 further specifies the behavior indicated by the behavior data acquired by the acquisition unit 101 (that is, the behavior specified immediately before) when “standing”, “lying”, or “ruminating” is specified above. In accordance with the determined action), a more accurate action is specified from the first pressure difference minimum value and the second pressure difference minimum value calculated by the difference minimum value calculation unit 104 (action specification by pressure analysis). .

  That is, the behavior specifying unit 105 determines that the user is standing up when the previous behavior is “lying down” and the first minimum pressure difference is smaller than the second minimum pressure difference by a predetermined threshold or more. Identify. On the other hand, when the previous action is “lying down” and the first pressure difference minimum value does not decrease from the second pressure difference minimum value by a predetermined threshold or more, Identify the previous action as the same.

  In addition, the behavior specifying unit 105 determines that the previous behavior is other than “lying down” and that the first pressure difference minimum value has increased from the second pressure difference minimum value by a predetermined threshold or more, and that And specify. On the other hand, the behavior specifying unit 105 determines that the immediately preceding behavior is other than “lying down”, and that the first pressure difference minimum value has not increased from the second pressure difference minimum value by a predetermined threshold or more. Identify that it is the same as the previous action.

  The predetermined threshold is, for example, a value preset by the user. Such a threshold value is determined, for example, by an empirical rule regarding behavior identification of the cow.

<Action identification processing>
Next, an action specifying process according to the present embodiment will be described.

行動 Behavior identification processing by motion intensity analysis ≫
First, the action specifying process based on the action strength analysis will be described with reference to FIG. FIG. 6 is a flowchart illustrating an example of the action specifying process based on the action intensity analysis. The action specifying process shown in FIG. 6 is repeatedly executed, for example, every 10 minutes. However, the action specifying process shown in FIG. 6 may be executed, for example, at a preset date and time, or may be executed at a predetermined time interval.

  First, the acquisition unit 101 acquires measurement data for a predetermined time (for example, 10 minutes) for each tag ID from the measurement data storage unit 200 (step S11). As described above, the acquisition unit 101 acquires measurement data in a predetermined time unit (for example, 10 minutes) for each cow (tag ID) from the measurement data storage unit 200. In addition, such a predetermined time is not limited to 10 minutes, and for example, the user of the behavior identification device 10 can set an arbitrary time.

In the following, measurement data 1 for 10 minutes of the tag ID "Tag1", measured data 2, ..., when the measured data _{N 1} is obtained by the obtaining unit 101 continues the description.

Next, the preprocessing unit 102 performs preprocessing on the measurement data acquired by the acquisition unit 101 (Step S12). That is, the pre-processing unit 102 performs a missing data complementing (resampling) process, a noise removal process, and the like of the measurement data. The loss complementing process is a process of complementing a loss when data cannot be acquired in order to increase the accuracy of collected data. Further, the noise removal processing is processing for removing data indicating instantaneous motion of the cow (for example, an operation of instantly shaking the body or an operation of instantly shaking the body). Thus, measurement data 1 after the pretreatment, measurement data 2, ..., measurement data N ₂ is obtained.

Next, the index value calculation unit 103 calculates an index value (the maximum value and the standard deviation of the L2 norm) from the L2 norm of the acceleration sensor value included in each measurement data after the preprocessing by the preprocessing unit 102 (step S13). ). That is, before the L2 norm of the acceleration sensor value included in the measurement data after the processing a1, a2, when the · · · aN _2, these a1, a2, standard deviation · · · aN ₂ sigma and a maximum value m calculate. The maximum value m is, L2 norm a1, a2, and the average value of · · · aN _2, which is the maximum value of the difference between the L2 norm ai (i = 1, ···, N 2).

  Next, the behavior specifying unit 105 uses the standard deviation σ and the maximum value m calculated by the index value calculation unit 103 and the behavior specifying model 300 to obtain four behaviors (“stand up”, “lying down”, “ruminate”, and “ Activity)) (step S14). That is, the action specifying unit 105 specifies which of the areas D1 to D4 on the action specifying model 300 contains the standard deviation σ and the maximum value m calculated by the index value calculating unit 103, Identify actions.

  Next, the action specifying unit 105 determines whether the action specified in step S14 is “standing”, “lying down”, or “rubbing” (step S15).

  When it is determined in step S15 that the specified action is “standing”, “lying”, or “rubbing”, the action specifying processing unit 100 specifies a more accurate action by the atmospheric pressure analysis (step S16). Details of the processing in this step will be described later.

  On the other hand, in step S15, when it is determined that the specified action is not “standing”, “lying” or “ruminating” (that is, when the specified action is “activity”), or after step S16, the action The specifying unit 105 stores the behavior data indicating the specified behavior in the behavior data storage unit 500 (Step S17). At this time, the behavior specifying unit 105 causes the behavior data storage unit 500 to store, for example, behavior data in which the identified behavior is associated with the current date and time.

  That is, when it is determined in step S15 that the specified action is not “standing”, “lying”, or “rubbing”, the action specifying unit 105 stores the action data in which the action “activity” is associated with the current date and time. The data is stored in the data storage unit 500. Similarly, after step S16, the action specifying unit 105 causes the action data storage unit 500 to store action data in which the action specified in step S16 is associated with the current date and time.

行動 Behavior identification processing by barometric pressure analysis≫
Here, the details of the process of step S16 (behavior specifying process by atmospheric pressure analysis) will be described with reference to FIG. FIG. 7 is a flowchart illustrating an example of the action specifying process based on the atmospheric pressure analysis.

  Here, as described above, “rumination” is divided into “standing rumination” and “lying rumination”. Therefore, in FIG. 7, the description will be made assuming that “standing” includes “rutting during standing” and “lying” includes “ruminating during lying”.

First, the acquiring unit 101 acquires the reference atmospheric pressure data during the same time as in step S11 from the reference atmospheric pressure data storage unit 400 (step S21). That is, for example, when time Tl to Tn ₃ in step S11 described above _{(N 3} is an integer of 1 or more) measurement data between has been acquired, the acquisition section 101, likewise, between the time Tl to Tn ₃ The reference pressure data is acquired from the reference pressure data storage unit 400.

  Next, the difference minimum value calculation unit 104 calculates the first pressure difference minimum value during the latest predetermined time (for example, the latest 5 minutes) from the measurement data and the reference air pressure data acquired by the acquisition unit 101. Then, a second pressure difference difference minimum value during a predetermined time period before (for example, 5 minutes before) is calculated (step S22).

  That is, the minimum difference value calculating unit 104 calculates, for example, the difference between each barometric pressure sensor value and each reference barometric pressure sensor value in the last five minutes, and determines the smallest one of the calculated differences as the first difference. The minimum pressure difference is used. Similarly, the minimum difference value calculating unit 104 calculates a difference between each barometric pressure sensor value and each reference barometric pressure sensor value in the preceding 5 minutes (that is, 5 minutes before the latest 5 minutes), for example, Of the calculated differences, the smallest one is set as a second pressure difference minimum value.

  Note that the minimum difference value calculating unit 104 can also use, for example, the average or the maximum value of the difference between each barometric pressure sensor value and each reference barometric pressure sensor value during a predetermined time. More specifically, for example, an “atmospheric pressure difference average value” indicating an average of a difference between each atmospheric pressure sensor value and each reference atmospheric pressure sensor value during a predetermined time may be used instead of the “atmospheric pressure difference minimum value”. it can. Similarly, for example, "the maximum pressure difference" indicating the maximum value of the difference between each pressure sensor value and each reference pressure sensor value during a predetermined time can be used instead of "the minimum pressure difference". However, the accuracy of action identification is lower when using the “atmospheric pressure difference average value” or “atmospheric pressure difference maximum value” than when using the “atmospheric pressure difference minimum value”.

  Next, the acquiring unit 101 acquires the latest behavior data (that is, the behavior data indicating the behavior identified immediately before) from the behavior data stored in the behavior data storage unit 500 (Step S23). .

  Next, the action specifying unit 105 determines whether the action indicated by the action data acquired in step S23 is “lying down” (step S24). That is, the action specifying unit 105 determines whether the action specified immediately before is “lying down”.

If it is determined in step S24 that the immediately preceding action is “lying down”, the action specifying unit 105 determines that the first minimum pressure difference is a predetermined threshold value TH ₁ (>) from the second minimum pressure difference. 0) It is determined whether or not it has decreased (step S25). In other words, the behavior specifying unit 105 determines whether the amount of change of the first minimum pressure difference with respect to the second minimum pressure difference is equal to or less than a predetermined threshold TH ₁ ′ (<0). Note that the threshold value TH ₁ ′ is obtained by reversing the sign of the threshold value TH ₁ .

In step S25, when it is determined that the amount of change of the first minimum pressure difference with respect to the second minimum pressure difference is equal to or less than a predetermined threshold value TH ₁ ′, the action specifying unit 105 sets the action to “stand up”. It is specified (Step S26). When the amount of change of the first minimum pressure difference with respect to the second minimum pressure difference is equal to or less than the predetermined threshold value TH ₁ ′, the altitude of the tag 20 attached to the cow increases by the predetermined threshold value TH ₁ or more. Because if you do.

For example, as shown in FIG. 8A, the second pressure difference minimum value between 10:00 and 10:05 is “13 Pa” (reference pressure sensor value “980 Pa”, pressure sensor value “993.0 Pa”). Suppose that the first pressure difference minimum value between 10:05 and 10:10 is “12.2 Pa” (reference pressure sensor value “978 Pa”, pressure sensor value “990.2 Pa”). In this case, the variation “−0.8” of the first minimum pressure difference “12.2” with respect to the second minimum pressure difference “13” is equal to or less than the threshold value TH ₁ ′ −0.5. Therefore, it is specified as “standing”.

On the other hand, if it is not determined in step S25 that the amount of change of the first minimum pressure difference with respect to the second minimum pressure difference is equal to or less than the predetermined threshold TH ₁ ′, the action specifying unit 105 determines that the action is 1 The same action as the previous action (that is, “lying down”) is specified (step S27). When the amount of change of the first minimum pressure difference with respect to the second minimum pressure difference does not fall below the predetermined threshold value TH ₁ ′, the altitude of the tag 20 attached to the cow rises above the predetermined threshold value TH _1. This is because it is considered not to have been performed, and it is considered that the state is still "lying down".

For example, as shown in FIG. 8B, the second minimum pressure difference between 10:00 and 10:05 is “13 Pa” (reference pressure sensor value “980 Pa”, pressure sensor value “993 Pa”). It is assumed that the first minimum pressure difference between 10:05 and 10:10 is “12.8 Pa” (reference pressure sensor value “978 Pa”, pressure sensor value “990.8 Pa”). In this case, the amount of change “−0.2” of the first minimum pressure difference “12.8” with respect to the second minimum pressure difference “13” does not fall below the threshold value TH ₁ ′ −0.5. Therefore, it is specified that the previous action “lying down” remains.

When it is determined in step S24 that the immediately preceding action is other than “lying down” (ie, “standing”), the action specifying unit 105 determines that the first minimum pressure difference is the second minimum pressure difference. Is determined to have increased by more than a predetermined threshold TH ₂ (<0) (step S28). In other words, the behavior specifying unit 105 determines whether or not the amount of change of the first minimum pressure difference with respect to the second minimum pressure difference is equal to or greater than a predetermined threshold TH ₂ ′ (> 0). Note that the threshold value TH ₂ ′ is obtained by reversing the sign of the threshold value TH ₂ .

In step S28, if a variation of the first pressure difference minimum value for the second pressure difference minimum value is judged to be a predetermined threshold value TH 2 _'or, behavior identification unit 105, the action as a "lying" It is specified (step S29). When the amount of change of the first minimum pressure difference with respect to the second minimum pressure difference is equal to or greater than the predetermined threshold TH ₂ ′, the altitude of the tag 20 attached to the cow is equal to or greater than the predetermined threshold TH ₂ ′, This is because it descends.

For example, as shown in FIG. 8C, the second pressure difference minimum value between 10:00 and 10:05 is “12.5 Pa” (reference pressure sensor value “980 Pa”, pressure sensor value “992. 5Pa "), and the first minimum pressure difference between 10:05 and 10:10 was" 13.3 Pa "(reference pressure sensor value" 978 Pa ", pressure sensor value" 991.3 Pa "). I do. In this case, the amount of change “0.8” of the first minimum pressure difference “13.3” with respect to the second minimum pressure difference “12.5” is equal to or greater than the threshold value TH ₂ ′ “0.5”. Therefore, it is specified as “lying down”.

On the other hand, in step S28, if a variation of the first pressure difference minimum value for the second pressure difference minimum value is not determined to be the predetermined threshold value TH ₂ or more, the behavior identification unit 105, one action The same action as the previous action (that is, “standing”) is specified (step S30). When the amount of change of the first minimum pressure difference with respect to the second minimum pressure difference is not equal to or greater than the predetermined threshold value TH ₂ ′, the altitude of the tag 20 attached to the cow is equal to or greater than the predetermined threshold value TH ₂ ′. This is because it does not descend, and it is considered that it is still in the “standing” state.

For example, as shown in FIG. 8D, the second pressure difference minimum value between 10:00 and 10:05 is “12.5 Pa” (reference pressure sensor value “980 Pa”, pressure sensor value “992. 5Pa "), and the first minimum pressure difference between 10:05 and 10:10 was" 12.8 Pa "(reference pressure sensor value" 978 Pa ", pressure sensor value" 990.8 Pa "). I do. In this case, the amount of change “0.3” of the first minimum pressure difference “12.8” with respect to the second minimum pressure difference “12.5” is equal to or greater than a predetermined threshold TH ₂ ′ “0.5”. Therefore, it is specified that the previous action “Standing up” remains.

  As described above, the behavior identification system 1 according to the present embodiment can identify the behavior of the cow from the measurement data received from the tag 20 worn by the cow. At this time, the behavior identification system 1 according to the present embodiment uses the index value calculated from the acceleration sensor value included in the measurement data and the behavior identification model 300 created in advance by a machine learning method, thereby (Behavior identification by behavior intensity analysis). In particular, by using the standard deviation and the maximum value of the L2 norm as the index values, it is possible to specify “rumination”, which is a behavior specific to cattle, with high accuracy. In other words, it is possible to distinguish "ruminating", which is a behavior specific to cattle, from behaviors such as "standing" and "lying down" with high accuracy.

  In addition, the behavior identification system 1 according to the present embodiment, when the behavior of the cow is identified as “standing”, “lying down” or “rubbing” in the behavior identification by the activity intensity analysis, the barometric pressure sensor value and the reference atmospheric pressure From the sensor value, more accurate behavior is specified (behavior specification by barometric pressure analysis). This makes it possible to specify “stand up” (including “ruining while standing”) or “lying” (including “ruminating while lying down”) with high accuracy. In addition, even in a barn having a hill or the like such as a free-burn barn, for example, "standing" (including "ruminating while standing") or "lying" (including "ruminating while lying") is highly accurate. Can be specified.

  The behavior of the cow identified in this way is displayed on, for example, a display device (display or the like) of the behavior identification device 10 and provided to a dairy farmer or the like, so that it is easy to confirm the behavior often seen when the cow is in estrus. Will be able to do it. In addition, it leads to early detection of abnormal behavior due to illness, injury, and the like, and health management of cattle can be easily performed.

<Specification of behavior in a freeburn barn>
Here, a case where the behavior is specified by the barometric pressure analysis according to the present embodiment in a free-burn barn will be described with reference to FIG. FIG. 9 is a diagram illustrating an example of specifying a behavior by barometric pressure analysis in a freeburn barn.

  As shown in FIG. 9, in the freeburn type barn, a hill exists in the barn. For this reason, the altitude of the tag 20 changes as the cow climbs the hill. In the barometric pressure analysis according to the present embodiment, even when the cow climbs the hill, the behavior of the cow (“Standing” (including “Rumming while standing”) or “Lying down” (including “Rumming while lying”) is highly accurate. Either) can be specified.

  For example, as shown in FIG. 9, the case where the altitude of the barometric pressure sensor when the cow is lying on the hill is higher than the altitude of the barometric sensor when the cow is standing below the hill. Think. At this time, for example, when the cow lies down after climbing the hill from the bottom of the hill, in the method of identifying the behavior of the cow by comparing the barometric pressure sensor value with a predetermined threshold, the cow lies on the hill. The behavior of a cow that is running may be incorrectly identified as “standing”. On the other hand, in the barometric pressure analysis according to the present embodiment, the behavior of the cow is specified based on whether the minimum value of the pressure difference has increased (or decreased) by a predetermined threshold or more. Even when lying down, the behavior of the cow can be correctly specified as “lying down”.

  As described above, in the behavior identification system 1 according to the present embodiment, high accuracy can be obtained by barometric analysis even when a hill exists in a barn such as a free-burn barn and a cow climbs the hill. Can identify the behavior of the cow.

  The present invention is not limited to the above-described embodiments specifically disclosed, and various modifications and changes can be made without departing from the scope of the claims.

REFERENCE SIGNS LIST 1 behavior identification system 10 behavior identification device 20 tag 30 reference atmospheric pressure sensor 100 behavior identification processing unit 101 acquisition unit 102 preprocessing unit 103 index value calculation unit 104 minimum difference value calculation unit 105 behavior identification unit 200 measurement data storage unit 300 behavior identification model 400 Reference atmospheric pressure data storage unit 500 Action data storage unit

Claims (9)

A behavior identification device that identifies the behavior of livestock,
Storage means for storing three-dimensional acceleration data measured by a three-axis acceleration sensor attached to the livestock, and pressure data measured by a pressure sensor;
Acquiring means for acquiring one or more three-dimensional acceleration data and one or more air pressure data during a predetermined time from the storage means;
An L2 norm of each of the one or more three-dimensional acceleration data acquired by the acquisition means is calculated, and a maximum value of the calculated one or more L2 norms and a standard deviation of the one or more L2 norms are used as index values. An index value calculating means for calculating as
Based on the index value calculated by the index value calculating means, and a specific model for specifying the behavior of the livestock created in advance, a first specifying means for specifying the behavior of the livestock,
When the action specified by the first specifying means is a predetermined action, a second specifying means for specifying the action of the livestock based on the one or more air pressure data obtained by the obtaining means,
Have a,
The specific model is
An action specifying device , which is represented in a two-dimensional space in which the maximum value is a first axis and the standard deviation is a second axis, in an area where the index value and the action are associated with each other . The second specifying means includes:
The action specified by the first specifying means is a standing action indicating that the livestock is standing, a lying action indicating that the livestock is lying, or a ruminating action indicating that the livestock is ruminating. The behavior specifying device according to claim 1, wherein in a case of any of the behaviors, the behavior of the livestock is specified based on the one or more atmospheric pressure data acquired by the acquisition unit. The second specifying means includes:
Among the one or more atmospheric pressure data, a first value based on a difference between an atmospheric pressure value indicated by the atmospheric pressure data during a first predetermined time and a reference atmospheric pressure value corresponding to the atmospheric pressure value;
Among the one or more air pressure data, an air pressure value indicated by the air pressure data during a second predetermined time indicating a time zone before the first predetermined time, and a reference air pressure value corresponding to the air pressure value And a second value is calculated from the difference between
The behavior specifying device according to claim 1 or 2, wherein the behavior of the livestock is specified according to a comparison between the calculated first value and the second value. The second specifying means includes:
In a case where the behavior at the time immediately before the livestock is other than the lying behavior, when the first value increases by a predetermined threshold or more than the second value, the behavior of the livestock is specified as the lying behavior. And
In a case where the behavior at the time immediately before the livestock is a lying behavior, when the first value is reduced by a predetermined threshold or more from the second value, the behavior of the livestock is specified as a standing behavior. The behavior specifying device according to claim 3. The second specifying means includes:
In the case where the behavior at the time immediately before the livestock is other than the recumbent behavior, when the first value does not increase by a predetermined threshold value or more than the second value, the behavior of the livestock is Identify the action at the previous time,
In a case where the behavior at the time immediately before the livestock is a lying behavior, when the first value does not decrease by a predetermined threshold value or more from the second value, the behavior of the livestock is changed to the 1st behavior. The action specifying device according to claim 3, wherein the action specifying apparatus specifies the action at the immediately preceding time point. The first value is
Among the one or more air pressure data, a minimum value, an average value, or a maximum value of a difference between an air pressure value indicated by the air pressure data during the first predetermined time and a reference air pressure value corresponding to the air pressure value. Yes,
The second value is
Among the one or more air pressure data, a minimum value, an average value, or a maximum value of a difference between an air pressure value indicated by the air pressure data during the second predetermined time and a reference air pressure value corresponding to the air pressure value. The behavior specifying device according to any one of claims 3 to 5. The behavior identification device according to any one of claims 1 to 6 , wherein the livestock is a cow. Behavior identification device that identifies the behavior of livestock,
A storage procedure for storing, in a storage unit, three-dimensional acceleration data measured by a three-axis acceleration sensor attached to the livestock and pressure data measured by a pressure sensor,
An acquisition procedure for acquiring one or more three-dimensional acceleration data and one or more air pressure data during a predetermined time from the storage unit;
An L2 norm of each of the one or more three-dimensional acceleration data acquired by the acquisition procedure is calculated, and a maximum value of the calculated one or more L2 norms and a standard deviation of the one or more L2 norms are used as index values. Index value calculation procedure to calculate as
Based on the index value calculated by the index value calculation procedure and a specific model for specifying the behavior of the livestock created in advance, a first specification procedure for specifying the behavior of the livestock,
When the action specified by the first specifying procedure is a predetermined action, a second specifying procedure for specifying the action of the livestock based on the one or more pressure data acquired by the acquiring procedure,
The execution,
The specific model is
An action identification method , wherein the action value is represented by an area in which the index value and the action are associated in a two-dimensional space in which the maximum value is a first axis and the standard deviation is a second axis . A program for causing a computer to function as each unit in the behavior specifying device according to any one of claims 1 to 7 .

Images