KR102268040B1 - Apparatus and method for managing livestock using machine learning - Google Patents

Abstract

The present invention discloses an apparatus and method for managing livestock using machine learning. The livestock management apparatus of the present invention preprocesses the communication unit for receiving image information photographed by a plurality of livestock and the received image information, performs machine learning based on the preprocessed image information to generate learning data, and uses the learning data and a control unit for deriving a countermeasure for selecting the livestock in the abnormal state and managing the livestock in the abnormal state.

Description

Livestock management apparatus and method using machine learning {Apparatus and method for managing livestock using machine learning}

The present invention relates to livestock management technology, and more particularly, to a livestock management apparatus and method using machine learning for managing livestock by performing machine learning based on image information of livestock.

In general, animals raised in barns are useful animals for human life by taming and improving wild animals, and have been mainly used for providing livestock and for ministry. Animals raised in these barns are domestic animals in a narrow sense, meaning only mammals. Livestock includes pigs, cattle, dairy cows, horses, goats, and the like.

As the industry develops, the livestock industry also develops, and livestock is being bred in large quantities in the livestock house according to the development of the livestock industry.

However, when livestock are raised in large quantities in a barn, if an epidemic occurs, the animals raised in the barn are killed at once, which causes a serious financial loss to the livestock farm. In addition, when livestock are bred in large quantities, there is a problem that cannot receive immediate management from the manager, such as childbirth and heat.

Korean Patent Publication No. 10-1694946 (2017.01.04.)

The technical task of the present invention is to capture and monitor livestock in real time, perform machine learning based on the monitored result to determine abnormal symptoms of livestock, and to derive a countermeasure for the symptom when abnormal symptoms occur An object of the present invention is to provide a livestock management device and method using machine learning.

In order to achieve the above object, the livestock management apparatus using machine learning according to the present invention preprocesses a communication unit for receiving image information of a plurality of livestock photographed and the received image information, and machine learning based on the preprocessed image information and a control unit for generating learning data by performing , selecting an abnormal livestock by using the learning data, and deriving a countermeasure for managing the abnormal state livestock.

In addition, the control unit filters unnecessary information among the image information, a preprocessor for object profiling the filtered image information, and learning data that helps determine the state of livestock by machine learning the object-profiled image information A learning unit for generating a state determination unit for determining the status of a plurality of livestock using the generated learning data, and selecting an abnormal state livestock using the determined result, and abnormal symptoms for the selected livestock and a countermeasure derivation unit for detecting the type of , and deriving a countermeasure corresponding to the type of the detected abnormal symptom.

In addition, the learning unit is characterized in that by classifying and predicting the behavior of the livestock based on the image information to generate the learning data representing the distribution and frequency characteristics of the behavior of the livestock.

In addition, the learning unit is characterized in that by comparing the image information with at least one of pre-stored livestock information, behavior pattern information and learning information, the behavior for the abnormal state and the normal state is calculated as characteristic information of distribution and frequency.

In addition, the state determination unit is characterized in that the state of the high characteristic information among the abnormal state and the normal state is determined as the current state of the livestock.

In addition, the countermeasure derivation unit analyzes the behavioral pattern of the livestock in an abnormal state, detects the type of anomaly using the analyzed result, and compares the detected type of the abnormality with pre-stored learning information to obtain the most similarity. It is characterized by deriving a high countermeasure.

In the livestock management method using machine learning according to the present invention, the livestock management device receives image information of a plurality of livestock photographed, the livestock management device pre-processes the received image information, the livestock management device performs the Generating learning data by performing machine learning based on the pre-processed image information, the livestock management device selecting the livestock in an abnormal state using the learning data, and the livestock management device selecting the livestock in the abnormal state It includes the step of deriving a countermeasure for management.

The livestock management apparatus and method using machine learning of the present invention performs machine learning based on the image information photographed in real time to determine the state of the livestock, and if the determined result is an abnormal state, a response appropriate to the state can provide a way.

Through this, the present invention allows the user to check the state of the livestock in real time and at the same time respond in advance to the behavior expected to move the livestock in the future, thereby efficiently managing the livestock.

1 is a configuration diagram for explaining a livestock management system according to an embodiment of the present invention.
2 is a block diagram illustrating a livestock management apparatus according to an embodiment of the present invention.
FIG. 3 is a block diagram illustrating the control unit of FIG. 2 .
FIG. 4 is a view for explaining the preprocessing unit of FIG. 3 .
FIG. 5 is a block diagram illustrating the storage unit of FIG. 2 .
6 is a flowchart for explaining a livestock management method according to an embodiment of the present invention.
7 is a flowchart for explaining step S30 of FIG. 6 .
8 is a flowchart for explaining step S40 of FIG. 6 .

Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings. First, in adding reference numerals to the components of each drawing, it should be noted that the same components are given the same reference numerals as much as possible even though they are indicated on different drawings. In addition, in describing the present invention, if it is determined that a detailed description of a related known configuration or function is obvious to those skilled in the art or may obscure the gist of the present invention, the detailed description thereof will be omitted.

1 is a configuration diagram for explaining a livestock management system according to an embodiment of the present invention.

Referring to FIG. 1 , the livestock management system 500 captures and monitors livestock in real time. The livestock management system 500 performs machine learning based on the monitored result to determine the abnormal symptoms of the livestock, and when abnormal symptoms occur, a countermeasure for the corresponding symptom is derived and a countermeasure is provided to the user. do. The livestock management system 500 includes a livestock management device 100 , an image capturing device 200 , and a user terminal 300 .

The livestock management apparatus 100 receives image information of a plurality of livestock photographed from the image capturing apparatus 200 , and monitors the breeding state of the livestock using the received image information. The livestock management device 100 transmits the monitored result to the user terminal 300 . The livestock management device 100 performs machine learning based on the monitored result, and selects whether the livestock is in a normal state or an abnormal state using the machine-learned learning data. Here, when the livestock determined to be in an abnormal state is selected, the livestock management apparatus 100 detects the type of abnormal symptom corresponding to the livestock. The livestock management apparatus 100 derives a countermeasure corresponding to the type of the detected abnormal symptom, and transmits the derived countermeasure to the user terminal 300 .

The image photographing apparatus 200 collects image information on which a plurality of livestock are photographed, and transmits the collected image information to the livestock management apparatus 100 . The image photographing apparatus 200 may be largely divided into a fixed image photographing apparatus and a mobile image photographing apparatus.

The fixed image photographing apparatus is a device for photographing a plurality of livestock at a fixed location, and is used when photographing livestock raised in a predetermined place. For example, the fixed image photographing apparatus may be implemented with a plurality of cameras such as CCTV. A plurality of cameras capture a predetermined range in real time. At this time, the plurality of cameras are installed in various directions so that images of various angles are captured.

A mobile imaging device is a device for photographing a plurality of livestock while moving, and is used when photographing livestock grazing in a wide range. For example, the mobile imaging device may be implemented as a drone equipped with a camera. Drones can take pictures of livestock while in flight. In this case, the drone can take pictures of livestock from multiple angles while not a single drone, but a plurality of drones flying in groups. In addition, drones flying in groups can increase the rate of image collection by increasing the probability of capturing moments that a single drone could miss while filming.

Preferably, the image photographing apparatus 200 may include an infrared photographing function, an ultra-low-light photographing function, or a thermal image photographing function to enable night photographing.

The user terminal 300 is a terminal used by the user, and receives the monitored result from the livestock management apparatus 100 and outputs the received monitored result. In addition, the user terminal 300 may receive a countermeasure from the livestock management device 100 and output the received countermeasure. Through this, the user can monitor the condition of livestock in real time and, at the same time, quickly respond to livestock with abnormal symptoms. The user terminal 300 includes a smart phone, a tablet PC, a handheld PC, a laptop, a desktop, and the like.

The livestock management system 500 establishes a communication network 400 between the livestock management device 100 , the image photographing device 200 and the user terminal 300 to communicate with each other. The communication network 400 may be composed of a backbone network and a subscriber network. The backbone network may be composed of one or a plurality of integrated networks among an X.25 network, a Frame Relay network, an ATM network, a Multi Protocol Label Switching (MPLS) network, and a Generalized Multi Protocol Label Switching (GMPLS) network. The subscriber network is Zigbee, Bluetooth, Wireless LAN (IEEE 802.11b, IEEE 802.11a, IEEE 802.11g, IEEE 802.11n), Wireless Hart (ISO/IEC62591-1), ISA100.11a (ISO/IEC) 62734), Constrained Application Protocol (COAP), Multi-Client Publish/Subscribe Messaging (MQTT), Wireless Broadband (WIBro), Wimax, 3G, High Speed Downlink Packet Access (HSDPA), 4G, and 5G, which is a next-generation communication network. In some embodiments, the communication network 400 may be a mobile communication network.

2 is a block diagram illustrating a livestock management apparatus according to an embodiment of the present invention.

Referring to FIG. 2 , the livestock management apparatus 100 includes a communication unit 10 and a control unit 20 , and further includes a storage unit 30 .

The communication unit 10 performs communication with the image photographing apparatus 200 and the user terminal 300 . The communication unit 10 receives image information on which a plurality of livestock are photographed from the image photographing device 200 . The communication unit 10 may transmit at least one of the monitored result and the countermeasure to the user terminal 300 .

The controller 20 pre-processes the image information and performs machine learning based on the pre-processed image information. Through this, the control unit 20 generates learning data. The learning data is data that analyzes the breeding status of livestock through learning. The control unit 20 monitors the livestock using the generated learning data. In this case, the control unit 20 selects the livestock in an abnormal state, and derives a countermeasure for managing the selected livestock in an abnormal state. Here, when the user wants to check the result of monitoring the livestock, the control unit 20 controls the monitored result to be transmitted to the user terminal 300 through the communication unit 10 . In addition, when the user wants to check a countermeasure for managing the livestock in an abnormal state, the control unit 20 controls the countermeasure to be transmitted to the user terminal 300 through the communication unit 10 .

The storage unit 30 stores livestock information. The storage unit 30 classifies and stores livestock information for each livestock. The storage unit 30 stores image information received from the image photographing apparatus 200 . In this case, the storage unit 30 may classify and store the image information by at least one of the captured time, date, month, and year. The storage unit 30 stores general livestock behavior pattern information for learning. The storage unit 30 stores the learned learning information. The storage unit 50 is a flash memory type (flash memory type), a hard disk type (hard disk type), a media card micro type (multimedia card micro type), card type memory (for example, SD or XD memory, etc.), Random Access Memory (RAM), Static Random Access Memory (SRAM), Read-Only Memory (ROM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Programmable Read-Only Memory (PROM), Magnetic Memory, It may include at least one storage medium among a magnetic disk and an optical disk.

FIG. 3 is a block diagram illustrating the control unit of FIG. 2 , and FIG. 4 is a diagram illustrating the preprocessor of FIG. 3 .

2 to 4 , the control unit 20 includes a preprocessing unit 21 , a learning unit 23 , a state determination unit 25 , and a countermeasure deriving unit 27 .

The preprocessor 21 filters unnecessary information among the image information. The preprocessor 21 filters unnecessary information such as duplicate information, error information, and distortion information among the image information. That is, the preprocessor 21 normalizes and simplifies the image information. The pre-processing unit 21 performs object profiling of the filtered image information to perform learning based on the image information. That is, the pre-processing unit 21 performs object profiling by blocking specific parts of a plurality of livestock included in the image information. A specific part is an important criterion for behavior analysis and behavior tracking, and may be the head, legs, tail, genitals, and the like. In this case, the preprocessor 21 may classify and generate object profiling into learning object profiling for learning and real-time object profiling representing real-time behavior. The pre-processing unit 21 prevents a later calculation process from being complicated through filtering and object profiling.

The learning unit 23 generates learning data to help determine the state of the livestock by machine learning the object-profiled image information. The learning unit 23 classifies and predicts the behavior of livestock based on the object-profiled image information to generate learning data representing probability information about the behavior of the livestock. At this time, the learning unit 23 calculates the distribution and frequency related to the abnormal state and the behavior for the normal state by comparing the object profiled image information with at least one of pre-stored livestock information, behavior pattern information, and learning information. Through this, the learning unit 23 may generate learning data indicating the characteristics of the behavior distribution and frequency for each livestock.

The state determination unit 25 determines the state of a plurality of livestock using the generated learning data, and selects the animal in an abnormal state using the determined result. The state determination unit 25 may determine the state of the livestock by using characteristic information of the behavior distribution and frequency included in the learning data of each livestock. The state determination unit 25 may determine a state in which the characteristic information is high among the abnormal state and the normal state as the current state of the livestock. In this case, the state determination unit 25 may further use the real-time object profiling generated by the preprocessor 21 to determine the state of the livestock. The state determination unit 25 may determine the state of the livestock by further using the action on a specific part of the livestock in real time.

The countermeasure deriving unit 27 detects an abnormal type for the selected livestock. The countermeasure deriving unit 27 derives a countermeasure corresponding to the detected abnormal type. In detail, the countermeasure deriving unit 27 analyzes the behavioral pattern of the livestock in an abnormal state, and detects an abnormal kind of the livestock using the analyzed result. At this time, the countermeasure deriving unit 27 compares the behavior pattern of the learning data with the previously stored behavior pattern information and the learning information, and determines that the behavior pattern with the highest similarity is the abnormal type. The countermeasure deriving unit 27 compares the detected abnormal type with the previously stored learning information to derive a countermeasure with the highest similarity. The countermeasure deriving unit 27 controls so that the derived countermeasure is transmitted to the user terminal 300 .

FIG. 5 is a block diagram illustrating the storage unit of FIG. 2 .

Referring to FIG. 5 , the storage unit 30 includes livestock information 31 , image information 33 , behavior pattern information 35 , and learning information 37 . The storage unit 30 individually stores livestock information 31 , image information 33 , behavior pattern information 35 , and learning information 37 .

The livestock information 31 is information of all livestock currently being raised, and includes information such as breed, sex, age, weight, and specific matters. The livestock information 31 may be updated at regular intervals through the livestock information received from the user terminal 300 .

The image information 33 is image information received from the image photographing apparatus 100 and is stored in real time. The image information 33 may be stored by classifying at least one of the captured time, date, month, and year. In addition, the image information 33 may be stored by classifying for each photographed camera.

The behavior pattern information 35 is data on the behavior pattern of a general livestock. At this time, the behavior pattern information 35 is stored by dividing the behavior pattern in the abnormal state and the behavior pattern in the normal state. In addition, the behavior pattern information 35 may be stored for each abnormal symptom type of an abnormal behavior pattern. Preferably, the behavior pattern information 35 may store behavior patterns for each part of the livestock.

The learning information 37 includes learned learning data. The learning data is data related to livestock learned through machine learning. The learning information 37 is divided into data indicating the state of livestock and data indicating a countermeasure for abnormal symptoms. The data indicating the state of livestock is data obtained by learning the current state of livestock, and the data indicating the countermeasures for abnormal signs is data obtained by learning the countermeasures for the abnormal symptoms that the livestock will act in the future. Learning information updates the learned data in real time when the learning process is completed.

6 is a flowchart for explaining a livestock management method according to an embodiment of the present invention.

1 and 6, the livestock management method determines the state of the livestock by performing machine learning based on the image information photographed in real time, and if the determined result is an abnormal state, a response appropriate to the state can provide a way. Through this, the livestock management method allows the user to check the status of livestock in real time and at the same time respond in advance to the behavior expected to move the livestock in the future, thereby efficiently managing the livestock.

In step S10 , the livestock management apparatus 100 receives image information from the image photographing apparatus 200 . The livestock management device 100 receives image information of a plurality of livestock photographed.

In step S20, the livestock management apparatus 100 pre-processes the received image information. The livestock management apparatus 100 filters unnecessary information among the image information. The livestock management apparatus 100 filters unnecessary information such as duplicate information, error information, distortion information, etc. among the image information. That is, the livestock management apparatus 100 normalizes and simplifies the image information. The livestock management apparatus 100 performs object profiling of the filtered image information to perform learning based on the image information.

In step S30, the livestock management apparatus 100 generates learning data by performing machine learning based on the pre-processed image information. The livestock management apparatus 100 classifies and predicts the behavior of the livestock based on the object-profiled image information to generate learning data indicating the characteristics of the distribution and frequency of the behavior of the livestock.

In step S40, the livestock management device 100 determines the state of the livestock using the learning data. The livestock management apparatus 100 selects whether the livestock is in a normal state or an abnormal state based on characteristic information related to the characteristics of the behavioral distribution and frequency of the livestock included in the learning data.

In step S50, the livestock management apparatus 100 branches to step S10 if the livestock is in a normal state, and performs step S60 if the livestock is in an abnormal state.

In step S60, the livestock management device 100 derives a countermeasure for managing the livestock in an abnormal state. The livestock management apparatus 100 analyzes the behavioral pattern of the livestock in an abnormal state, and detects an abnormal kind of the livestock using the analyzed result. The livestock management apparatus 100 compares the detected abnormal type with the previously stored learning information to derive a countermeasure with the highest degree of similarity. The livestock management apparatus 100 transmits the derived countermeasure to the user terminal 300 so that the countermeasure can be output from the user terminal 300 .

7 is a flowchart for explaining step S30 of FIG. 6 .

Referring to FIG. 7 , the livestock management device 100 generates learning data through machine learning.

In step S31, the livestock management device 100 analyzes the behavioral profile of the livestock. The livestock management apparatus 100 analyzes the behavioral profile of the livestock based on the object-profiled image information. That is, the livestock management device 100 analyzes the behavior for a specific part of the livestock and determines what kind of behavior the corresponding behavior is.

In step S33, the livestock management apparatus 100 performs machine learning using the behavioral profile of the livestock. The livestock management device 100 classifies by behavior using the analyzed behavioral profiling of the livestock, and predicts the behavior of the livestock through the corresponding behavior. The livestock management apparatus 100 classifies and predicts the behavior of livestock by using at least one of pre-stored behavior pattern information and learning information.

In step S35, the livestock management device 100 generates learning data using the analyzed machine-learning information. The livestock management device 100 represents classification and predicted information on the behavior of livestock by machine learning as characteristic information related to the characteristics of the behavior distribution and frequency of the livestock. The livestock management device 100 generates learning data of livestock by using the characteristic information. That is, the learning data is data that helps determine the state of the livestock, and may be data indicating the characteristics of the behavioral distribution and frequency of the livestock.

8 is a flowchart for explaining step S40 of FIG. 6 .

Referring to FIG. 8 , the livestock management apparatus 100 determines the state of the livestock.

In step S41, the livestock management device 100 analyzes the learning data. The livestock management apparatus 100 analyzes the state of the livestock by using the characteristic information included in the learning data. In this case, the livestock management apparatus 100 may further analyze the state of the livestock by further using the previously generated real-time object profiling.

In step S43, the livestock management apparatus 100 determines the state of the livestock using the analyzed learning data. The livestock management apparatus 100 may determine the state of the livestock by using the probability information included in the learning data. That is, the livestock management apparatus 100 may determine the state in which the characteristic information is high among the abnormal state and the normal state as the current state of the livestock.

Although preferred embodiments of the present invention have been illustrated and described above, the present invention is not limited to the specific preferred embodiments described above, and in the technical field to which the present invention belongs, without departing from the gist of the present invention as claimed in the claims Any person skilled in the art can make various modifications, of course, and such modifications are within the scope of the claims.

10: communication department
20: control unit
21: preprocessor
23: Study Department
25: state judgment unit
27: Response plan derivation unit
30: storage
31: livestock information
33: video information
35: behavior pattern information
37: learning information
100: livestock management device
200: video recording device
300: user terminal
400: communication network
500: livestock management system

Claims (7)

an image photographing device for collecting image information of a plurality of livestock; and
Including; a livestock management device for managing livestock using the collected image information;
The video recording device,
A fixed-type imaging device for photographing a plurality of livestock at a fixed location to photograph livestock raised in a preset place, and a mobile imaging device for photographing a plurality of livestock with a plurality of drones to photograph livestock grazing within a preset range device, including
It includes infrared shooting function, ultra low-light shooting function or thermal imaging function to enable night shooting.
The livestock management device,
a communication unit for receiving image information of a plurality of livestock; and
Pre-processing the received image information, performing machine learning based on the pre-processed image information to generate learning data, selecting an abnormal livestock using the learning data, and managing the abnormal livestock Including; a control unit for deriving a countermeasure for
The control unit is
a preprocessor for normalizing and simplifying image information by filtering duplicate information, error information, and distortion information among the image information, and object profiling for blocking specific parts of a plurality of livestock included in the image information;
a learning unit for generating learning data to help determine the state of livestock by machine learning the object-profiled image information;
a state determination unit that determines the state of a plurality of livestock using the generated learning data, and selects the animal in an abnormal state using the determined result; and
Including; and a countermeasure derivation unit for detecting the type of the abnormal symptom for the selected livestock and deriving a countermeasure corresponding to the detected type of the abnormal symptom.
The preprocessor is
Generated by classifying learning object profiling for learning the object profiling and real-time object profiling representing real-time behavior,
The learning unit,
Classification and prediction of the behavior of livestock based on the object-profiled image information to generate learning data representing probability information about the behavior of the livestock, the object-profiled image information, pre-stored livestock information, behavior pattern information, and Comparing at least one of the learning information, calculating the distribution and frequency related to the behavior for the abnormal state and the normal state based on the compared result to generate learning data indicating the characteristics of the behavior distribution and frequency for each livestock, ,
The status determination unit,
Based on the characteristic information of the behavior distribution and frequency included in the learning data of each livestock, the state with high characteristic information among the abnormal and normal states is determined as the current state of the livestock, but the real-time object profiling and real-time specific part of the livestock Judging the condition of livestock by further using the behavior of
The countermeasure derivation unit,
When the livestock is in an abnormal state, the behavior pattern of the animal in the abnormal state is analyzed, and the behavior pattern with the highest similarity is determined by comparing the behavior pattern of the learning data, the pre-stored behavior pattern, and the learning information based on the analyzed result. Detects by the type of symptom, and compares the detected type of abnormal symptom with pre-stored learning information to derive a countermeasure with the highest degree of similarity,
The specific area is
Livestock management device using machine learning, characterized in that it is a part that is a criterion for judgment for behavior analysis and behavior tracking, and includes a tail and genitals. delete delete delete delete delete delete

Images