WO2020005038A1

WO2020005038A1 - System and terminal for managing environment of breeding place and method thereof

Info

Publication number: WO2020005038A1
Application number: PCT/KR2019/007935
Authority: WO
Inventors: 김일봉; 류승훈; 신제용
Original assignee: 엘지이노텍 주식회사
Priority date: 2018-06-29
Filing date: 2019-06-28
Publication date: 2020-01-02

Abstract

A terminal for managing an environment of a breeding place according to an embodiment of the present invention comprises: an image analysis unit for detecting at least one of the position and movement of a moving individual animal in a breeding place, from image data including information on an individual animal, and thus generating behavior inclination data including at least one piece of data among position distribution data and movement data on the moving individual animal; and an environment parameter estimation unit for applying the behavior inclination data and environmental data of the breeding place to a machine learning-based optimal environment estimation algorithm, and thus estimating an optimal environment parameter optimized for growth of the moving individual animal.

Description

Farm environment management system, terminal and method

Embodiments relate to a kennel environmental management system, a terminal, and a method thereof.

Recently, diseases targeting livestock farms such as mad cow disease, foot and mouth disease, and bird flu have been spreading. Livestock diseases such as mad cow disease, foot-and-mouth disease, and bird flu have very high mortality and transmission rates, but most treatments have been solved by killing because no treatment has been found. These livestock diseases are becoming a bigger problem because of the potential for human transmission.

In particular, livestock diseases such as mad cow disease, foot-and-mouth disease, and bird flu are intensively occurring in Korea around the world. Pointing out. However, it is difficult to get out of the factory farming system right away in the present situation of our country where farmland is scarce.

Therefore, there is a need for a system that can improve the production of livestock diseases while reducing the incidence of livestock diseases by improving the environment of a livestock farm.

Embodiments provide a farm environment management system, a terminal, and a method for optimizing a livestock farming environment of a farm to improve a production capacity.

The problem to be solved in the embodiment is not limited thereto, and also includes a purpose or effect which can be grasped from the solution means or the embodiment described below.

The kennel environment management terminal according to an embodiment of the present invention detects at least one of the position or movement of the moving object in the kennel from the image data including the information on the object to generate the behavioral propensity data for the moving object And an environment parameter estimator for estimating an optimal environment parameter optimized for growth of the moving object by applying the behavioral propensity data and the environment data of the kennel to an optimal environment estimation algorithm based on a machine learning.

The behavior propensity data may include at least one of position distribution data and motion data for the motion entity.

The image analyzer may divide the image data into a plurality of regions, and generate the behavior tendency data by calculating at least one of an activity amount and a distribution amount of the moving object for each of the plurality of regions.

The image analyzer may classify the image data into a plurality of regions such that the actual areas corresponding to each of the plurality of regions are equal to each other based on at least one of height, direction, and lens distortion information of the imaging apparatus generating the image data. .

The image analyzer may detect the edge of the object from the image data and generate the position distribution data by calculating a distribution of the object for each of the plurality of regions based on the detected edge of the object and a previously stored comparison database. It may comprise one analysis unit.

The comparison database may include an animal subject DB trained by applying a training DB constructed from images of pre-photographed animal subjects to a machine learning algorithm.

The image analyzer may include a second analysis unit that calculates a motion vector of a pixel of the image data and generates the motion data by calculating the activity of the object for each of the plurality of regions based on the motion vector.

The environment parameter estimator may generate a training data set by combining the behavioral propensity data and the environment data, and learn the optimal environment estimation algorithm using the training data set.

The environmental data may include at least one of feed amount, feed intake, age, moving object weight, temperature, humidity, carbon dioxide, and outside temperature of the kennel.

The environmental parameter estimating unit is based on a correlation between at least one of the feed amount, feed intake, age, moving object weight, temperature, humidity, carbon dioxide, and outside temperature of the breeding ground and at least one of the location distribution data and the operation data. The training dataset can be generated.

The optimum environment parameter may include at least one of a target temperature, a target humidity, a target feed amount and a target water supply amount of the kennel.

The environmental parameter estimating unit calculates data reliability based on the behavioral propensity data and the environmental data and a confidence range corresponding to each of the environmental propensity data, and the behavioral propensity data and the environmental data according to a result of comparing the data reliability with a preset threshold. It can generate an alarm signal for.

The environment parameter estimator may store a result of the user's determination of the alarm signal, and calculate the data reliability using the stored user's decision result when the behavior tendency data and the environment data corresponding to the alarm signal are input. Can be.

The environment parameter estimator may calculate a final environment parameter based on at least one of a recommended environment parameter and the optimum environment parameter based on the environment data.

The environment parameter estimator determines whether the environment of the kennel deteriorates when the recommended environment parameter is applied based on the recommended environment parameter and the environment data, and determines that the optimal environment parameter is the final environment parameter when it is determined that the environment of the kennel deteriorates. If it is determined that the environment of the kennel improves, the final environmental parameter may be calculated by combining the optimal environment parameter with the recommended parameter.

The recommendation environment parameter may be generated by a learning server including a machine learning based recommendation environment estimation algorithm.

The learning server generates a second training dataset by combining the environment data and the production amount of the moving object in the kennel, and the environment parameter estimating unit learns the recommended environment estimation algorithm using the second training dataset. You can.

The learning server generates a second training dataset by combining environmental data for other breeding grounds other than the breeding ground and a moving object production amount, and the environment parameter estimating unit uses the second training dataset to generate the recommended environment. Train the estimation algorithm.

The learning server may learn the recommended environment estimation algorithm according to at least one of a date, a region, and a place.

The kennel environment management system according to the embodiment of the present invention is disposed at a preset position in a kennel, an imaging apparatus for generating an image data including information on an object by photographing the foreground in the kennel, and arranged at a preset position in the kennel. A sensing device for measuring environment in the kennel and generating environmental data, and detecting position and movement of the movement entity in the kennel from the image data to generate behavior propensity data for the movement entity, and the behavior propensity data And a kennel environment management terminal for applying the environment data to a machine learning-based optimal environment estimation algorithm to estimate an optimal environment parameter optimized for growth of the moving object.

The air conditioning apparatus may further include an air conditioning apparatus configured to receive the optimum environmental parameters from the kennel environment management terminal and to adjust at least one of temperature and humidity in the kennel based on the optimum environment parameters.

The sensing device may be disposed at a position higher than a predetermined height from the bottom surface of the kennel.

In the kennel environment management method according to an embodiment of the present invention, the kennel environment management terminal detects at least one of the position or the movement of the moving object in the kennel from the image data including the information on the individual to the behavior propensity data for the moving object Generating, and estimating, by the kennel environment management terminal, the behavioral propensity data and the environment data of the kennel to an optimal environment estimation algorithm based on a machine learning to estimate an optimal environment parameter optimized for growth of the moving object. .

In the kennel environment management method according to an embodiment of the present invention, the image pickup apparatus photographs the foreground in the kennel to generate the image data including the information on the object, the sensing device to measure the environment in the kennel to generate environmental data Step, the kennel environment management terminal detects at least one of the position or movement of the moving object in the kennel from the image data to generate behavioral propensity data for the moving object, the kennel environment management terminal and the behavioral propensity of the individual Estimating an optimal environment parameter optimized for growth of the moving object by applying environmental data of a kennel to a machine learning based optimal environment estimation algorithm, and an air conditioning apparatus receives the optimum environment parameter from the kennel environment management terminal, Based on the optimum environmental parameters Adjusting at least one of temperature and humidity in the kennel.

Generating, by the learning server, a recommended environment parameter according to at least one of a current date, a region and a place where the kennel is located, and transmitting the recommended environment parameter to the kennel environment management terminal; and at least one of the recommended environment parameter and the optimal environment parameter based on the environment data. The method may further include calculating a final environmental parameter based on one.

Communication terminal according to an embodiment of the present invention is a communication processor for communicating with the air conditioning apparatus; And receiving a captured image, outputting an image in which the pixel region is masked to instruct the position of the object in the captured image and whether the object is in motion, and outputting at least one of the masked image and the current temperature and the current humidity. And a controller configured to generate a control signal for adjusting at least one of a target temperature and a target humidity of the air conditioner, and transmit the control signal to the air conditioner.

By managing the environment of the kennel by reflecting the state of the kennel in real time, there is an advantage that can improve livestock production and prevent livestock disease.

Various and advantageous advantages and effects of the present invention are not limited to the above description, and will be more readily understood in the course of describing specific embodiments of the present invention.

1A is a conceptual diagram of a kennel environmental management system according to an embodiment of the present invention.

1B is a conceptual diagram of a kennel environmental management system according to another embodiment of the present invention.

2 is a configuration diagram of an imaging device according to an embodiment of the present invention.

3 is a view for explaining an arrangement of an imaging device according to an embodiment of the present invention.

4 is a view for explaining the arrangement of the sensing device according to an embodiment of the present invention.

5 is a block diagram of a kennel environment management terminal according to an embodiment of the present invention.

6A is a flowchart illustrating a method for managing a kennel environment of a kennel environment management terminal according to an exemplary embodiment of the present invention.

6B is a flowchart illustrating a kennel environment management method of the kennel environment management terminal according to another embodiment of the present invention.

7A is a flowchart illustrating a kennel environment management method of a kennel environment management system according to an exemplary embodiment of the present invention.

7B is a flowchart illustrating a kennel environment management method of the kennel environment management system according to another embodiment of the present invention.

8 is a diagram for explaining a machine learning based distribution amount prediction algorithm according to an embodiment of the present invention.

9A is a conceptual diagram illustrating a data flow according to calculation of an optimal environment parameter according to an embodiment of the present invention.

9B is a conceptual diagram illustrating a data flow according to calculation of an optimal environment parameter according to an embodiment of the present invention.

As the invention allows for various changes and numerous embodiments, particular embodiments will be illustrated and described in the drawings. However, this is not intended to limit the present invention to specific embodiments, it should be understood to include all modifications, equivalents, and substitutes included in the spirit and scope of the present invention.

Terms including ordinal numbers, such as second and first, may be used to describe various components, but the components are not limited by the terms. The terms are used only for the purpose of distinguishing one component from another. For example, without departing from the scope of the present invention, the second component may be referred to as the first component, and similarly, the first component may also be referred to as the second component. The term and / or includes a combination of a plurality of related items or any item of a plurality of related items.

When a component is referred to as being "connected" or "connected" to another component, it may be directly connected to or connected to that other component, but it may be understood that other components may be present in between. Should be. On the other hand, when a component is said to be "directly connected" or "directly connected" to another component, it should be understood that there is no other component in between.

The terminology used herein is for the purpose of describing particular example embodiments only and is not intended to be limiting of the present invention. Singular expressions include plural expressions unless the context clearly indicates otherwise. In this application, the terms "comprise" or "have" are intended to indicate that there is a feature, number, step, operation, component, part, or combination thereof described in the specification, and one or more other features. It is to be understood that the present invention does not exclude the possibility of the presence or the addition of numbers, steps, operations, components, components, or a combination thereof.

Unless defined otherwise, all terms used herein, including technical or scientific terms, have the same meaning as commonly understood by one of ordinary skill in the art. Terms such as those defined in the commonly used dictionaries should be construed as having meanings consistent with the meanings in the context of the related art and shall not be construed in ideal or excessively formal meanings unless expressly defined in this application. Do not.

DETAILED DESCRIPTION Hereinafter, exemplary embodiments will be described in detail with reference to the accompanying drawings, and the same or corresponding components will be given the same reference numerals regardless of the reference numerals, and redundant description thereof will be omitted.

As shown in FIG. 1, the kennel environment management system according to the embodiment of the present invention may include an imaging device 100, a sensing device 200, a kennel environment management terminal 300, and an air conditioning device 400. . Here, the kennel means a livestock farm. The livestock may be not only poultry such as chickens and ducks, but also various kinds of animals bred in groups such as cattle and pigs.

First, the imaging apparatus 100 may photograph an environment in a kennel, and transmit image data photographed by an individual to the kennel environment management terminal 300. To this end, the imaging apparatus 100 may communicate with the kennel environment management terminal 300 by wire or wirelessly.

A plurality of imaging devices 100 may be disposed in the kennel. The imaging apparatus 100 may transmit the image data photographed by the image sensor to the kennel environment management terminal 300. Alternatively, a separate data collection device may be disposed in the kennel to collect image data from the plurality of imaging devices 100 and transmit the image data to the kennel environment management terminal 300. In order to accurately analyze the state inside the kennel, the image data of the imaging apparatus 100 may have high quality and high capacity. Therefore, a large bandwidth may be required to transmit the image data photographed by the plurality of imaging devices 100 to the remote kennel environment management terminal 300 in real time, and thus the transmission speed may be reduced. Therefore, in a separate data collection device, by collecting and encoding the image data from the image capture device 100 in the kennel and transmitting the data to the kennel environment management terminal 300, the communication bandwidth can be reduced and the transmission speed can be improved. Alternatively, the imaging apparatus 100 performs the first analysis for analyzing the state of the kennel using the image data, and then transfers the processed data to a data collection device (not shown) or to the kennel environment management terminal 300. Can transmit That is, the imaging apparatus 100 may efficiently manage communication resources by selecting only the data necessary for analyzing the state of the kennel and transferring the data to the data collector or the kennel environment management terminal 300. Alternatively, the imaging apparatus 100 may primarily analyze the internal state of the kennel using the image data photographing the kennel and transmit the result to the data collection device or transmit the result to the kennel environment management terminal 300.

Next, the sensing device 200 may measure environment information in the bathing area, generate environment data including the environment information, and transmit the generated environment data to the kennel environment management terminal 300. To this end, the sensing device 200 may communicate with the kennel environment management terminal 300 by wire or wirelessly. A plurality of sensing devices 200 may be disposed in the kennel. The plurality of sensing devices 200 may directly transmit environment data measuring the environment of the kennel to the kennel environment management terminal 300. Alternatively, a separate data collection device may be disposed in the kennel, and the data collector may collect environmental data from the plurality of sensing devices 200 and transmit the environmental data to the kennel environment management terminal 300.

The sensing device 200 may be a device in which sensors for measuring various kinds of environmental information are modularized. For example, the sensing device 200 may be a device in which a sensor for measuring environmental information such as a temperature sensor, a humidity sensor, and a gas sensor is modularized. The plurality of sensing devices 200 in which a plurality of sensors are modularized may be disposed at various locations in the kennel. Alternatively, the plurality of sensing devices 200 may measure different environment information. For example, the first sensing device 200 may measure the temperature in the kennel, and the second sensing device 200 may measure the humidity in the kennel. In this case, the sensing device 200 measuring the same kind of environmental information may be several. For example, the first and fifth sensing apparatuses 200 may measure the temperature in the kennel, and the second to fourth sensing apparatuses 200 may measure the humidity in the kennel.

Next, the kennel environment management terminal 300 uses the image data received from the imaging device 100 and the environment data received from the sensing device 200 to generate an optimal environment parameter optimized for the growth of livestock in the kennel.

The kennel environment management terminal 300 detects at least one of the position or movement of the livestock in the kennel from the image data including the information on the individual to generate behavioral propensity data for the livestock in the kennel. In addition, the kennel environment management terminal 300 estimates an optimal environmental parameter optimized for livestock growth by applying behavioral propensity data and environmental data to an optimal environment estimation algorithm based on machine learning. In addition, the kennel environment management terminal 300 may calculate the final environment parameter based on at least one of the recommendation environment parameter and the optimal environment parameter based on the kennel.

The kennel environment management terminal 300 may be implemented as a personal computer (PC), a tablet PC, a portable terminal, or the like, or may be implemented in the form of a server or a cloud. When the kennel environment management terminal 300 is a central server that receives information from a plurality of kennels, the manager may not have easy access to the kennel environment management terminal 300. In this case, the kennel environment management terminal 300 may transmit information on the kennel management through a portable terminal of a separate administrator. Details of the kennel environment management terminal 300 will be described in detail with reference to the accompanying drawings.

Next, the air conditioner 400 is a device for adjusting the temperature of the kennel. The air conditioning apparatus 400 may adjust at least one of temperature and humidity in the kennel based on the optimum environment parameter received from the kennel environment management terminal 300. For example, assume that an optimal environmental parameter of "target temperature 20 degrees" has been received. At this time, if the temperature of the kennel is 25 degrees, the air conditioning apparatus 400 may operate the cold heat source so that the room temperature of the kennel reaches 20 degrees. As another example, assume that an optimal environmental parameter of "target humidity 40%" has been received. At this time, if the humidity of the kennel is 70%, the air conditioning apparatus 400 may operate the blower so that the indoor humidity of the kennel reaches 40%.

On the other hand, the kennel environment management system according to an embodiment of the present invention may further include a communication terminal (not shown).

The communication terminal may include a communication processor and a controller.

In detail, the communication processor communicates with the air conditioning apparatus 400 and transmits an adjustment signal generated by the controller of the communication terminal to the air conditioning apparatus. In addition, the communication processor may communicate with the imaging apparatus 100, the sensing apparatus 200, and the kennel environment management terminal 300, and may receive data such as a captured image.

The controller may receive a captured image and output an image in which the pixel area is masked to instruct the position of the object in the captured image and whether the object moves. For example, the controller may output an image in which motion data and position distribution data of the captured image are masked in a pixel area of the captured image. In this case, the motion data and the position distribution data may be represented by region, block, or pixel. In the present specification, the abnormal entity information may be mixed with the abnormal entity data. Alternatively, the abnormal entity information may be information obtained from abnormal entity data. For example, the higher the probability that an object exists in a pixel of the original captured image, the higher the value of the corresponding pixel of the masked image. As another example, it may indicate whether a motion exists in an area or a block corresponding to each divided area or block of the image data, or indicate whether a motion exists in a pixel corresponding to each pixel.

The controller may generate an adjustment signal for adjusting at least one of a target temperature and a target humidity of the air conditioning apparatus by using the masked image and at least one of the current temperature and the current humidity.

As shown in FIG. 1B, the kennel environment management system according to another embodiment of the present invention includes an imaging device 100, a sensing device 200, a kennel environment management terminal 300, and an air conditioning device 400. It may include a communication terminal (not shown), and may further include a learning server 500. Since the imaging apparatus 100, the sensing apparatus 200, the kennel environment management terminal 300, the air conditioning apparatus 400, and the communication terminal are the same as those described with reference to FIG. 1A, detailed descriptions thereof will be omitted.

The learning server 500 is an apparatus including a machine learning based recommendation environment estimation algorithm for generating a recommendation environment parameter. The learning server 500 may generate a recommendation environment parameter through a recommendation environment estimation algorithm according to at least one of a current date, a region and a place where the kennel is located. The learning server 500 may improve the appropriateness of the recommendation environment parameter by learning the recommendation environment estimation algorithm through the second training data. The second training data may be generated from environmental data and livestock production information obtained from a plurality of farm environment management terminals 300 connected to the learning server 500. The learning server 400 may be implemented as a personal computer (PC), a tablet PC, a portable terminal, or the like, or may be implemented in the form of a cloud. Details of the learning server 500 will be described in detail with reference to the following drawings.

Referring to FIG. 2, the imaging apparatus 100 may include a photographing unit 111, a control unit 112, and a first communication unit 113, and an encoding unit 114, a database 115, and a light source unit 116. , May further include a pan tilt portion 117. At least one of the control unit 112 and the encoding unit 114 may be implemented by a computer processor or a chip, the database 115 may be mixed with a memory, and the communication unit 113 may be mixed with an antenna or a communication processor. have.

First, the photographing unit 111 may photograph an image including a plurality of objects. In the exemplary embodiment of the present invention, the plurality of individuals may refer to poultry farmed in a kennel. The photographing unit 111 may generate image data by using an image including a plurality of objects. One first image data may mean a single frame. The photographing unit 111 may generate a plurality of image data using images sequentially photographed.

The photographing unit 111 may be an image sensor photographing a subject by using a complementary metal-oxide semiconductor (COMS) module or a charge coupled device (CCD) module. At this time, the input image frame is provided to the COMS module or CCD module in the photographing unit 111 through the lens, the COMS module or CCD module converts the optical signal of the subject passing through the lens into an electrical signal (image data) and outputs it. do.

The photographing unit 111 may include a fisheye lens or a wide angle lens having a wide viewing angle. Accordingly, it is possible for one imaging device 100 to photograph the entire space inside the kennel.

In addition, the photographing unit 111 may be a depth camera. The photographing unit 111 may be driven by any one of various depth recognition methods, and the depth information may be included in the image photographed by the photographing unit 111. The photographing unit 111 may be, for example, a Kinect sensor. Kinect sensor is a depth camera of the structured light projection method, it is possible to obtain a three-dimensional information of the scene by projecting a pattern image defined using a projector or a laser, and obtains the image projected pattern through the camera. These Kinect sensors include infrared emitters that irradiate patterns using infrared lasers, and infrared cameras that capture infrared images. An RGB camera that functions like a typical webcam is disposed between the infrared emitter and the infrared camera. In addition to this, the Kinect sensor may further include a pan tilt unit 117 for adjusting the angle of the microphone array and the camera.

The basic principle of the Kinect sensor is that when the laser pattern irradiated from the infrared emitter is projected and reflected on the object, the distance to the surface of the object is obtained using the position and size of the pattern at the reflection point. According to this principle, the photographing unit 111 may generate the image data including depth information for each object by irradiating a laser pattern into the space in the barn and sensing the laser pattern reflected from the object.

Next, the control unit 112 controls the overall operation of the imaging apparatus 100. For example, the controller 112 may control the photographing unit 111 to follow-up the specific region in which the abnormal object is located. The tracking photographing target may be set through a user interface unit (not shown) or through a control command of a kennel environment management terminal. The control unit 112 controls the photographing unit 111 to track and photograph a specific area in which an abnormal object exists to generate image data so that continuous monitoring can be performed.

In this case, the controller 112 may control the pan tilt unit 117 to perform tracking shooting. The pan tilt unit 117 may control a photographing area of the photographing unit 111 by driving two motors, a pan and a tilt. The pan tilt unit 117 may adjust the directing direction of the photographing unit 111 to photograph a specific area under the control of the controller 112. In addition, the pan tilt unit 117 may adjust the directing direction of the photographing unit 111 to track a specific object under the control of the controller 112.

Next, the first communication unit 113 may perform data communication with at least one of the other imaging apparatus 100 and the kennel environment management terminal. For example, the first communication unit 113 may include a wireless LAN (WLAN), a Wi-Fi, a Wi-Fi, a WiMAX, a World Interoperability for Microwave Access (Wimax), and an HSDPA (High). Data communication may be performed using telecommunication technologies such as Speed Downlink Packet Access), IEEE 802.16, Long Term Evolution (LTE), and Wireless Mobile Broadband Service (WMBS).

Alternatively, the first communication unit 113 may include Bluetooth, RadioFrequency Identification (RFID), Infrared Data Association (IrDA), Ultra Wideband (UWB), Zigbee, and Near Field Communication (NFC). In addition, as a wired communication technology, data communication may be performed using a short-range communication technology such as USB communication, Ethernet, serial communication, and optical / coaxial cable.

For example, the first communication unit 113 may perform data communication with another abnormality object detecting apparatus using a short range communication technique, and may perform data communication with a kennel environment management terminal using a telecommunication technique, but is not limited thereto. Various communication technologies may be used in consideration of various matters of the kennel.

The first communication unit 113 may transmit the image data photographed by the photographing unit 111 to the kennel environment management terminal. The image data transmitted through the first communication unit 113 may be compressed data encoded through the encoding unit 114.

Next, the encoder 114 encodes the image data photographed by the photographing unit 111 into a digital signal. For example, the encoding unit 114 may encode image data according to H.264, H.265, Moving Picture Experts Group (MPEG), and Motion Joint Photographic Experts Group (M-JPEG) standards.

Next, the database 115 includes a flash memory type, a hard disk type, a multimedia card micro type, and a card type memory (for example, SD or XD memory). Etc.), magnetic memory, magnetic disk, optical disk, random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EPM), PROM It may include at least one storage medium of (Programmable Read-Only Memory). In addition, the imaging apparatus 100 may operate a web storage that performs a storage function of the database 115 on the Internet, or may operate in connection with the web storage.

The database 115 may store image data photographed by the photographing unit 111 and may store image data for a predetermined period of time. In addition, the database 115 may store data, a program, and the like necessary for the imaging apparatus 100 to operate. In addition, the database 115 may store various user interfaces (UIs) or graphical user interfaces (GUIs).

Next, the light source unit 116 may irradiate light in a direction directed under the control of the control unit 112. For example, the light source unit 116 may include at least one laser diode (LD) and a light emitting diode (LED). The light source unit 116 may emit light of various wavelength bands under the control of the controller 112.

For example, the light source unit 116 may irradiate light in the infrared wavelength band for night photographing. Alternatively, the light source unit 116 may irradiate light in the ultraviolet wavelength band for photochemotherapy of livestock in the kennel.

According to an embodiment, as shown in FIG. 3A, the imaging apparatus 100 may be implemented as one, and may be disposed at a predetermined position in the kennel 10. The imaging apparatus 100 is disposed in consideration of the viewing angle of the imaging apparatus 100 to generate image data including information about an object by photographing the foreground in the kennel 10. That is, the imaging apparatus 100 may be disposed at a position where the entire foreground of the kennel 10 may be photographed. For example, it may be disposed above the kennel 10, it may be arranged on the side of the kennel 10.

As another example, as illustrated in FIGS. 3B and 3C, the imaging apparatus 100 may be implemented in plural and disposed at various preset locations in the kennel 10. The plurality of imaging apparatuses 100 are disposed at various places in the kennel 10 in consideration of the viewing angle of the imaging apparatus 100 to generate image data including information about an object by photographing the foreground in the kennel 10. do. That is, the imaging apparatus 100 may be disposed at a position where the entire foreground of the kennel 10 may be photographed. For example, the plurality of imaging devices 100 may be disposed on the upper and side portions of the kennel 10.

Meanwhile, when there are a plurality of kennels 10, the imaging apparatus 100 may be arranged for each of the kennels 10, and the plurality of imaging apparatuses 100 may be arranged for each of the kennels 10.

The sensing device 200 is disposed at a predetermined position in the kennel 10 to measure environment information in the kennel 10 to generate environment data. In this case, as illustrated in FIG. 4, the sensing device 200 may be disposed at a position higher than a predetermined height from the bottom surface of the kennel 10. The predetermined height may be determined based on the radius of action of the livestock in the kennel 10. For example, when the livestock is a chicken, the sensing device 200 may be disposed at a position higher than a height at which the chicken may reach the wing.

In the case of the sensing device 200, since the size is small and the durability is not high, breakage or failure may easily occur. When the sensing device 200 is installed within the action radius of the livestock in the kennel 10, the livestock may damage the sensing device 200 by misunderstanding the sensor as food. Therefore, the sensing device 200 may be disposed at a position higher than a predetermined height from the bottom surface of the kennel 10 to prevent breakage.

As shown in FIG. 5, the kennel environment management terminal 300 according to the embodiment of the present invention includes an image analyzer 310 and an environment parameter estimator 320, and may further include a communication unit. The image analyzer 310 may include a first analysis unit 311 and a second analysis unit 312.

First, the image analyzer 310 detects at least one of the position or movement of the livestock in the kennel from the image data including the information on the individual to generate behavioral propensity data for the livestock in the kennel. Behavioral propensity data includes at least one of location distribution data and behavioral data for livestock. The position distribution data is generated by detecting the position of the livestock from the image data, and the motion data is generated by detecting the movement of the livestock from the image data. The image analyzer 310 divides the image data into a plurality of regions, and calculates at least one of an activity amount and a distribution amount of livestock for each of the plurality of regions. For example, when the image data is divided into 24 regions, the activity and distribution of each of the 24 regions may be generated from one frame. Here, the activity of the livestock may mean the amount of motion vector in the divided region, and the distribution of the livestock may mean the density of the livestock detected in the divided region.

In this case, the image analyzer 310 may classify the image data into a plurality of regions such that the actual area of each of the plurality of regions is the same based on at least one of height, direction, and lens distortion information of the imaging apparatus generating the image data. Even if image data is divided into the same area due to distortion of a lens or the like, the actual area of each area may not be the same. Therefore, when calculating the amount of activity and distribution of each area by dividing the image data itself into the same area, the actual area of each area is different, which may cause a problem in data reliability. For example, it is assumed that image data is divided into the same area in an image, and an actual area corresponding to an area A of the plurality of areas is 10 m ² and an actual area corresponding to an area B is 20 m ² . In this case, when the number of individuals detected in the area A is 30 and the number of individuals detected in the area B is 40, the distribution of the area A is high, but the calculation result is calculated to be the distribution of the area B high. To solve this problem, the image analyzer 310 according to an embodiment of the present invention may divide the image data into a plurality of regions such that the actual areas of the plurality of regions are the same.

Next, the process of generating position distribution data will be described in detail. The first analysis unit 311 included in the image analyzer 310 may generate location distribution data by detecting the livestock individual on the image data. That is, the first analysis unit 311 may generate position distribution data by inputting image data to a distribution amount prediction algorithm that detects a livestock individual and calculates a distribution amount of the livestock. The distribution prediction algorithm may be a machine learning based algorithm. In detail, the first analysis unit 311 may detect the edge of the object from the image data and generate position distribution data by calculating a distribution amount of the object for each of the plurality of regions based on the detected edge of the object and a previously stored comparison DB. have. Here, the individual may mean a background of a kennel including a feeder as well as a livestock.

In one embodiment, the first analysis unit 311 detects the outline of the object in the image data, compares the detected outline with the appearance of the animal object previously stored in the comparison DB to determine the outline matching the appearance of the animal object previously stored. Livestock can be detected with livestock. In this case, the appearance of the livestock stored in the comparison DB may be the appearance of at least one or more livestock, and the abnormal individual analysis unit detects the individual having the matching outline as the livestock as described above and may also determine the type of the livestock. .

In another embodiment, the first analysis unit 311 extracts a feature point of the entity in the image data, and if the extracted feature point matches the feature point of the animal previously stored in the comparison DB with a proximity greater than or equal to a threshold, the entity in the image data is domesticated. Can be detected as. In this case, the first analysis unit 311 extracts feature points from images of two objects to be compared, and scale invariant feature transform (SIFT) or speeded up robust (SURF) matching the extracted feature descriptors of the extracted two objects. Features) Algorithms can be used.

In another embodiment, the first analysis unit 311 may detect livestock based on the contours of the objects in the image data. More specifically, the first analysis unit 311 generates an edge image by detecting the contours of the objects in the image data, and detects the contour from the foreground image data, which is a background image of the kennel, which is pre-stored in the comparison DB 118, and thus the background edge. An image may be generated and livestock may be detected from a differential image obtained by subtracting a background edge image from an edge image. In this case, the first analyzing unit 311 generates an edge image by detecting an edge of an object appearing in the frame using the gradient information of the image data frame. Here, the gradient information is a value generated from a difference value between adjacent pixels among predetermined pixels in a frame, and means a sum of absolute values of differences, and an edge means a boundary between objects using gradient information.

In another embodiment, the first analysis unit 311 may generate a background edge image by detecting an edge of an object corresponding to a background from image data of a foreground in a pre-scanned farm. In this case, the background edge image may be an image obtained by detecting the contours of the objects of the preset area with the background edge, but comparing the plurality of image data frames of the foreground in the pre-scanned kennel and repeating the predetermined number of times or more. The image may be an image of which the contour is detected as a background edge.

In another embodiment, the first analysis unit 311 may detect livestock from the image data using the object detection classifier. In this case, the object detection classifier is trained by constructing a training DB from livestock images taken by changing the posture or the external environment of the animal. The object detection classifier uses a support vector machine (SVM), a neural network, an AdaBoost algorithm, and the like. A comparison DB of livestock is generated through various learning algorithms. In detail, the first analysis unit 311 detects an edge of the object corresponding to the foreground from the image data of the background in the previously photographed kennel, applies an edge of the foreground object detected from the image data, Livestock can be detected by applying the object detection classifier to the area of the applied image data.

Next, a process of generating motion data will be described in detail. The second analysis unit 312 included in the image analyzer 310 may generate motion data indicating the movement of the livestock using the image data. That is, the second analysis unit 312 may generate motion data by inputting image data into an activity prediction algorithm that calculates the activity of the individual. In detail, the second analysis unit 312 may generate motion data by calculating a motion vector for a pixel of the image data and calculating an activity amount of the object for each of the plurality of regions based on the motion vector.

In one embodiment, the second analysis unit 312 may detect the movement of the moving object using the Dense Optical Flow method. The second analysis unit 312 may calculate a motion vector for all the pixels on the image data to detect a motion for each pixel. In the case of the Dense Optical Flow method, since the motion vector is calculated for all pixels, the detection accuracy is improved, but the amount of computation is relatively increased. Therefore, the Dense Optical Flow method can be applied to a specific area where detection accuracy is very high, such as a kennel where an abnormal situation is suspected or a kennel with a large number of individuals.

In another embodiment, the second analysis unit 312 may detect the movement of the moving object using a sparse optical flow method. The second analysis unit 312 may detect a motion by calculating a motion vector only for some of the characteristic pixels that are easy to track, such as edges in the image. The sparse optical flow method reduces the amount of computation, but only results for a limited number of pixels. Therefore, the sparse optical flow method may be applied to a kennel having a small number of individuals or to a specific area where the objects do not overlap.

In another embodiment, the second analyzing unit 312 may detect movement of the moving object using block matching. The second analysis unit 312 may divide the image evenly or unequally, calculate a motion vector with respect to the divided region, and detect a motion. Block Matching reduces the amount of computation because it calculates the motion vector for each partition, but it can have a relatively low detection accuracy because it calculates the results for the motion vector for each region. Accordingly, the block matching method may be applied to a kennel with a small number of individuals or to a specific area where the objects do not overlap.

In another embodiment, the second analysis unit 312 may detect the movement of the moving object by using a continuous frame difference method. The second analyzing unit 312 may compare successive image frames for each pixel and calculate a value corresponding to the difference to detect motion. Since the Continuous Frame Difference method detects motion by using the difference between frames, the overall computational amount is reduced, but the detection accuracy of a large object or a duplicate object may be relatively low. In addition, the Continuous Frame Difference method can not distinguish between the background image and the moving object may have a relatively low accuracy. Therefore, the Continuous Frame Difference method may be applied to a kennel with a small number of objects or to a specific area where the objects do not overlap.

In another embodiment, the second analysis unit 312 may detect the movement of the moving object by using a background subtraction method. The second analysis unit 312 may detect the motion by comparing successive image frames for each pixel in a state where the background image is initially learned, and calculating a value corresponding to the difference. The background subtraction method is to learn the background image in advance so that the background image can be distinguished from the moving object. Therefore, a separate process of filtering the background image is required, thereby increasing the amount of computation but improving the accuracy. Therefore, the background subtraction method can be applied to a specific area where detection accuracy is very high, such as a kennel suspected of an abnormal situation or a kennel with a large number of individuals. In the Background Subtraction method, the background image can be updated continuously.

Next, the environment parameter estimator 320 includes a machine learning based optimal environment estimation algorithm, and calculates an optimal environment parameter using the optimal environment estimation algorithm. In detail, the environmental parameter estimator 320 estimates an optimal environmental parameter optimized for livestock growth by applying behavioral propensity data and environment data of a kennel to an optimal environment estimation algorithm based on machine learning.

Here, the environmental data may include at least one of the amount of feed provided, feed intake, age, livestock weight, temperature, humidity, carbon dioxide amount, outside temperature of the kennel. Among them, the temperature, humidity, and carbon dioxide amount of the kennel may be received from the sensing device, and the amount of feed provided, the amount of feed intake, and the amount of livestock gain may be input from the user. The moon age and the outside temperature can be received from an external server such as the Meteorological Agency server.

The environment parameter estimator 320 may calculate the final environment parameter based on at least one of the recommended environment parameter and the optimal environment parameter based on the environment data. When controlling the air conditioning apparatus according to the recommended environmental parameters, the environmental parameter estimating unit 320 determines whether the environment of the kennel deteriorates and determines whether to consider the recommended environmental parameters when calculating the final environmental parameters. If it is determined that the environment of the kennel deteriorates according to the recommended environmental parameter, the environmental parameter estimator 320 determines the optimum environmental parameter as the final environmental parameter. On the other hand, if it is determined that the environment does not deteriorate according to the recommended environmental parameters, the environmental parameter estimator 320 calculates the final environmental parameters in consideration of the optimal environmental parameters and the recommended environmental parameters.

The optimum environmental parameter may include at least one of an optimum temperature, an optimum humidity, an optimum feeding amount, and an optimum water supply amount of the kennel. The recommended environment parameter includes at least one of the recommended temperature, recommended humidity, recommended feeding amount and recommended water supply amount of the kennel. The final environmental parameter includes at least one of the final temperature, final humidity, final feed amount and final feed amount of the kennel.

The environmental parameter estimator 320 may generate a training data set by combining behavioral propensity data and environmental data, and may learn an optimal environmental estimation algorithm using the training data set. In this case, the training data set may be generated by combining a training data generation algorithm included in the environment parameter estimator 320 with behavioral propensity data and environmental data. The environmental parameter estimator 320 may include training data based on a correlation between at least one of feed supply, feed intake, age, livestock weight, temperature, humidity, carbon dioxide, and external temperature of the kennel and at least one of location distribution data and motion data. You can create a set.

The environmental parameter estimator 320 may determine the reliability of the behavioral propensity data and the environmental data, and may transmit an alarm signal to the user according to the determined reliability. In detail, the environmental parameter estimator 320 calculates data reliability based on the behavioral propensity data and the environmental data and the confidence range corresponding to each of the environmental parameter estimator 320, and the data reliability is smaller than the preset threshold. Alarm signals can be generated. In response to the alarm signal, when the user inputs the determination result on the behavior propensity data and the environment data, the environment parameter estimator 320 stores the determination result of the user on the alarm signal, and the behavior propensity data corresponding to the alarm signal; When the environment data is input, data reliability may be calculated using the stored user's determination result. For example, the environmental parameter estimator 320 may calculate, as data reliability, the standard deviation of the current temperature with respect to the kennel temperature for the previous hour from the current time point, and if the standard deviation is higher than the threshold, Can generate an alarm signal. Then, the user may input whether the current temperature is an error or a correction temperature in response to the alarm signal, and the environmental parameter estimator 320 may store the received information. If a similar situation occurs later, the environmental parameter estimator 320 may determine data reliability according to the stored information.

Next, the second communication unit 330 may perform data communication with at least one of the imaging apparatus and the air conditioning apparatus. For example, the second communication unit 330 may include a wireless LAN (WLAN), a Wi-Fi, a Wi-Fi, a WiMAX, a World Interoperability for Microwave Access (Wimax), and an HSDPA (High). Data communication may be performed using telecommunication technologies such as Speed Downlink Packet Access), IEEE 802.16, Long Term Evolution (LTE), and Wireless Mobile Broadband Service (WMBS).

Alternatively, the second communication unit 330 may include Bluetooth, RadioFrequency Identification (RFID), Infrared Data Association (IrDA), Ultra WideBand (UWB), Zigbee, and Near Field Communication (NFC). In addition, as a wired communication technology, data communication may be performed using a short-range communication technology such as USB communication, Ethernet, serial communication, and optical / coaxial cable.

For example, the second communication unit 330 may perform data communication with another abnormality object detecting apparatus using a short range communication technology, and may perform data communication with the kennel environment management terminal 300 using a long distance communication technology. However, the present invention is not limited thereto, and various communication technologies may be used in consideration of various matters of the kennel.

The second communication unit 330 may transmit the image data photographed by the photographing unit to the kennel environment management terminal 300. The image data transmitted through the second communication unit 330 may be compressed data encoded through the encoding unit.

First, the image analyzer receives the image data photographing the foreground of the kennel and the environmental data measuring the environment information of the kennel (S610).

The image analyzer detects at least one of the position or movement of the livestock in the kennel from the image data to generate behavioral propensity data for the livestock in the kennel (S611). Behavioral propensity data may include at least one of location distribution data and motion data for livestock.

Then, the environmental parameter estimator estimates the optimal environmental parameters optimized for the growth of the livestock by applying the behavioral propensity data and the environment data of the kennel to the optimal environment estimation algorithm based on the machine learning (S612).

The environment parameter estimator transmits the optimum environment parameter to the air conditioning apparatus (S613).

In addition, the environmental parameter estimator generates a training data set by combining behavioral propensity data and environmental data (S614). The environment parameter estimator learns an optimal environment estimation algorithm using the training data set (S615).

Detailed description of the configuration described with reference to FIG. 6A will be omitted. S620 to S622 of FIG. 6B correspond to S610 to S612 of FIG. 6A, respectively. S625 to S627 in FIG. 6B correspond to S613 to S615 in FIG. 6A, respectively.

After the steps S620 to S622, the environmental parameter estimator estimates an optimal environmental parameter optimized for livestock growth by applying behavioral propensity data and environment data of the kennel to an optimal environment estimation algorithm based on the machine learning (S622).

The environment parameter estimator determines whether the environment of the kennel deteriorates when the recommended environment parameter is applied based on the environment data (S623).

Then, the environmental parameter estimating unit calculates the final environmental parameter based on at least one of the optimal environmental parameter and the recommended environmental parameter according to the determination result of the deterioration of the kennel environment (S624).

Then, steps S625 to S627 are performed.

First, the imaging apparatus 100 photographs the foreground of the kennel to generate the image data including the information on the individual in the kennel (S710), and then transmits the image data to the kennel environment management terminal 300 (S711).

Then, the sensing device 200 measures the environment of the kennel to generate environment data (S712), and transmits the environment data to the kennel environment management terminal 300 (S713).

Then, the kennel environment management terminal 300 detects at least one of the position or movement of the livestock in the kennel from the image data to generate behavioral propensity data for the livestock in the kennel (S714). Behavioral propensity data may include at least one of location distribution data and motion data for livestock.

Next, the kennel environment management terminal 300 generates a training data set by combining behavioral propensity data and environmental data (S715). The kennel environment management terminal 300 learns an optimal environment estimation algorithm using the training data set (S716).

In addition, the kennel environment management terminal 300 estimates an optimal environment parameter optimized for livestock growth by applying behavioral propensity data and environment data of the kennel to an optimal environment estimation algorithm based on machine learning (S717).

The environmental parameter estimator transmits the optimum environmental parameter to the air conditioning apparatus 400 (S718).

Then, the air conditioning apparatus 400 controls the environment of the kennel based on the optimum environmental parameters (S719).

Detailed description of the configuration described with reference to FIG. 7A will be omitted. S720 to S727 of FIG. 7B correspond to S710 to S717 of FIG. 7A, respectively. S730 and S731 of FIG. 7B correspond to S718 and S719 of FIG. 7A, respectively.

After the steps S720 to S727, the kennel environment management terminal 300 determines whether the kennel environment is deteriorated when the recommended environment parameter is applied based on the environment data (S728).

Then, the kennel environment management terminal 300 calculates the final environment parameter based on at least one of the optimal environment parameter and the recommended environment parameter according to the determination result of the deterioration of the kennel environment (S729).

Then, steps S730 and S731 are performed.

Here, the distribution prediction algorithm is an example of a deep learning algorithm designed to display density information of a region where an object is located. The distribution amount prediction algorithm may be an algorithm for inputting an original image to a convolution network-based learning machine and then outputting a density image represented by a gray scale probability map. According to this, abnormal object information can be easily extracted even in a kennel environment in which objects having a similar shape, such as a chicken house, are accommodated at high density.

As shown in FIG. 8, the distribution prediction algorithm may include at least one convolution network (layer).

The convolutional network may classify the feature points of the image using at least one feature map (W).

The convolutional network may improve the performance of the distribution prediction algorithm through at least one pooler and / or activator.

The distribution prediction algorithm further includes a concatenator, and the concatenator may concatenate and rearrange output results of at least one convolutional network to output density (distribution) information of an object by using feature points of the original image. .

The kennel environment management terminal can learn the distribution prediction algorithm using the original image and the density image, and at least one feature map (W) is trained (tuned) to output the density information of the object. Can be

In an embodiment of the present invention, the density prediction algorithm is configured such that the density image has a size (Width x Highet) that is the same as or similar to the original image, and the positions of each pixel (or block) correspond to each other, and the pixels of the density image are The value of may represent the probability that an object (eg, poultry) may exist in a corresponding pixel of the original image. For example, the higher the probability that an object exists in a pixel of the original image, the higher the value of the pixel of the density image may be.

In FIG. 9, the flow of data will be described based on a process of estimating an optimal environment parameter through an optimal environment estimation algorithm.

First, the n th image data is input to the distribution amount prediction algorithm and the active amount prediction algorithm, respectively, and the n-1 th image data is input to the activity amount prediction algorithm. The image data may be RGB data having a W X H size, and the n th image data may be mixed with the n th original data, the n th original image, the n th original image data, and the like.

The image analyzer 310 of the kennel environment management terminal 300, for example, the first analysis unit 311, detects an object from the nth image data and generates position distribution data of the object with respect to the nth image data. Here, the position distribution data may be generated for each region, for each block, or for each pixel, and the image analyzer 310 may use a learned distribution prediction algorithm to suggest a region in which the object is located in the image data. . Here, the distribution amount prediction algorithm may use a real-time object detection method using an algorithm indicating an area where an object exists as described above, or the distribution amount prediction algorithm may use distribution information of an object, that is, an area of an area where an object is located. It is also possible to use a method of representing density information. Here, the position distribution data may be a density map.

The image analyzing unit 310 of the kennel environment management terminal 300, for example, the second analysis unit 312 compares the n-1th image data and the nth image data to determine the movement of the object with respect to the nth image data. Detect. In this case, the n−1th image data may be stored in the latch circuit or the buffer circuit. The image analyzer 310, for example, the second analysis unit 312 of the farm environment management terminal 300 generates motion data according to the detected movement. Here, the motion data may be a motion map. Here, the motion data may be generated for each region, block, or pixel, and the image analyzer 310 may use an activity amount prediction algorithm trained to detect motion by using an optical flow.

Then, the kennel environment management terminal 300 inputs the n-th position distribution data, the n-th operation data and the n-th environment data received from the sensor to the optimum environment estimation algorithm to extract the optimal environment parameters. The optimum environmental parameters are transferred to the air conditioning apparatus 400, which operates the air conditioning apparatus 400 using the optimal environmental parameters.

Meanwhile, the kennel environment management terminal 300 may train, that is, update the optimal environment estimation algorithm.

The optimal environment estimation algorithm may be at least one of a supervised learning based machine learning algorithm, an unsupervised learning based machine learning algorithm, and a reinforcement learning based machine learning algorithm.

Optimal environmental estimation algorithms include logistic regression, k-nearest neighbor (KNN), support vector machine (SVM), generalized linear models (GLM), and pseudo Decision trees, random forests, gradient boosting machines (GBMs), deep learning, convolutional neural networks (CNNs), recurrent neural networks (RNNs) ), Clustering, Anomaly detection, Dimension reduction, Underlying Probability Density Estimation, Autoencoders, Q-Learning, Deep-Q-Network, DQN), Naive Bayes Classifier (NBC), Hidden Markov Model (HMM), K-Means Clustering, K-Nearest Neighbors, k -NN), ensemble learning (Ens) emble learning).

Specifically, the kennel environment management terminal 300 according to an embodiment of the present invention may generate a training data set by combining motion data, location data, and environmental data through a training data generation algorithm.

The training data generation algorithm may generate the training dataset according to the type of the optimal environment estimation algorithm.

According to an embodiment, when the optimal environment estimation algorithm is a supervised learning-based machine learning algorithm, the training data generation algorithm may generate the training data set by labeling the environmental data in at least one of the position distribution data and the motion data. Can be. The training data generation algorithm may generate at least one of position distribution data and motion data in the environmental data to generate a training data set. For example, the training data set may be [Data (temperature of the nth environmental data); Label (nth motion data)] or [Data (nth motion data, nth position distribution data); Label (temperature of the nth environmental data, humidity of the nth environmental data, CO2 of the nth environmental data).

According to another embodiment, when the optimal environment estimation algorithm is an unsupervised learning-based machine learning algorithm, the training data generation algorithm may generate a training data set by combining location distribution data, motion data, and environment data. For example, the training data set may be in the form of [Data (n th motion data, n th position distribution data, n th environmental data temperature, n th environmental data humidity)].

According to another embodiment, when the optimal environment estimation algorithm is an unsupervised learning-based machine learning algorithm, the training data generation algorithm may generate a training data set by combining location distribution data, motion data, and environment data. For example, the training data set is [state (nth motion data, nth position distribution data, nth environmental data temperature, nth environmental data humidity), action (nth optimal environmental parameter control temperature, nth Control humidity of the optimal environmental parameters), reward (production, feed intake, weight gain)]. In this case, the action (the control temperature of the n-th optimal environmental parameter, the control humidity of the n-th optimal environmental parameter) may be a value of the optimum environmental parameter, and may be an actual control value of the air conditioning apparatus.

In addition, the kennel environment management terminal 300 according to an embodiment of the present invention may learn an optimal environment estimation algorithm using a training data set.

For example, assume that the optimal environment estimation algorithm is a neural network circuit composed of an input layer, a hidden layer, and an output layer.

Then, the n th operation data, the n th position distribution data, and the n th temperature data are input to three nodes of an input layer, respectively. The node of each input layer may extract feature values from the n th operation data, the n th position distribution data, and the n th temperature. For example, the sum of the magnitude values of the entire motion vectors of the n th motion data may be extracted as a feature value, and various feature values may be extracted.

Each feature value is input to each node of the hidden layer, and each node of the hidden layer adds the value multiplied by the corresponding weight (w) to the feature value input in Equation 1 and Equation 1 below and The output may be determined by applying to a sigmoid fuction, which is an activation function such as Equation 2.

[Equation 1]

w ₁ is a weight corresponding to the n th motion data, x ₁ is a feature value of the n th motion data, w ₂ is a weight corresponding to the n th position distribution data, and x ₂ is a feature value of the n th position distribution data , W ₃ is a weight corresponding to the n th environmental data, and x ₃ is a feature value of the n th environmental data.

[Equation 2]

The hidden layers other than the first hidden layer connected to the input layer may be modified by Equation 1 according to the input node.

Then, the kennel environment management terminal 300 according to the embodiment of the present invention adjusts the weight (w) set in each node according to the error of the value output through the output layer. In this case, the weight w may be adjusted using backpropagation.

9B is a conceptual diagram illustrating a data flow according to calculation of an optimal environment parameter according to another embodiment of the present invention.

FIG. 9B may include all of the configurations described with reference to FIG. 9A, and may further include contents described below.

The final environment estimation algorithm generates a final environment parameter based on at least one of the optimal environment parameter and the recommended environment parameter.

First, the final environment estimation algorithm determines whether the recommended environment parameter is appropriate in order to determine whether to consider only the optimal environment parameter or the recommended environment parameter and the recommended environment parameter together when calculating the final environment parameter. The final environment estimation algorithm compares the environment data for the kennel with the recommended environment parameter to determine whether productivity is deteriorated when the environment is controlled according to the recommended environment parameter. In one embodiment, when the difference value between the value of the environmental data and the value of the recommended environment parameter exceeds the threshold range, the final environment estimation algorithm may determine that the kennel environment is deteriorated when the recommended environment parameter is applied. For example, if the critical range for temperature is set between -5 and 5 degrees, and the difference between the current temperature in the environmental data and the recommended temperature in the kennel recommended environmental parameter is 10 degrees, then the final environmental estimation algorithm will determine the recommended environmental parameter. It can be judged that the application environment deteriorates when applied.

If it is determined that the environment of the kennel deteriorates when the recommended environment parameter is applied, the final environment estimation algorithm determines the optimal environment parameter as the final environment parameter without considering the recommended environment parameter.

On the other hand, if it is determined that the environment of the kennel does not deteriorate when the recommended environment parameter is applied, the final environment estimation algorithm determines the optimum environment parameter as the final environment parameter without considering the recommended environment parameter. In an embodiment, the final environment estimation algorithm may calculate an average value of the recommended environment parameter and the optimal environment parameter as the final environment parameter. In another embodiment, the final environment estimation algorithm may generate an offset value by adding a preset weight to the recommended environment parameter, and calculate or add a final value to the optimum environment parameter by adding or subtracting the offset value. have.

Meanwhile, the recommended environment estimation algorithm included in the learning server 400 may set at least one of a date, a region, and a place as a category, and learn by using the second training data set for each category. For example, the recommended environment estimation algorithm may have a category of <Gangwon-do Cheorwon-gun, April>. At this time, if the area of the kennel where the environmental data used for the second training data set is generated is Cheorwon-gun, Gangwon-do, and the date is April 15, the second training dataset is the recommended environment in the category <Gewon, Cheorwon-gun, April>. Used in the estimation algorithm.

The second training dataset is used to train the machine learning based recommended environment estimation algorithm. The recommendation environment estimation algorithm may generate a recommendation environment parameter according to at least one of a current date, a region, and a place where the kennel is located. As an example, when generating recommended environment parameters for a kennel in a coastal area of Gimhae, Gyeongsang-do, October, the recommended environment estimation algorithm may generate recommended environment parameters within a category of <Gyeongsang-do, Gimhae, coast, autumn>. have.

The recommended environment estimation algorithm may transmit the recommended environment parameter to the kennel environment management terminal at regular intervals or, if there is a request for transmission, the recommended environment parameter to the kennel environment management terminal. Since the type and learning process of the recommended environment estimation algorithm differ only from the information included in the training data set, the detailed environment description is omitted.

The term '~ part' used in the present embodiment refers to software or a hardware component such as a field-programmable gate array (FPGA) or an ASIC, and '~ part' performs certain roles. However, '~' is not meant to be limited to software or hardware. '~ Portion' may be configured to be in an addressable storage medium or may be configured to play one or more processors. Thus, as an example, '~' means components such as software components, object-oriented software components, class components, and task components, and processes, functions, properties, procedures, and the like. Subroutines, segments of program code, drivers, firmware, microcode, circuits, data, databases, data structures, tables, arrays, and variables. The functionality provided within the components and the 'parts' may be combined into a smaller number of components and the 'parts' or further separated into additional components and the 'parts'. In addition, the components and '~' may be implemented to play one or more CPUs in the device or secure multimedia card.

Although the above description has been made based on the embodiments, these are merely examples and are not intended to limit the present invention. Those skilled in the art to which the present invention pertains may not have been exemplified above without departing from the essential characteristics of the present embodiments. It will be appreciated that many variations and applications are possible. For example, each component specifically shown in the embodiment can be modified. And differences relating to such modifications and applications will have to be construed as being included in the scope of the invention defined in the appended claims.

Claims

An image analyzer which detects at least one of the position or movement of the moving object in the kennel from the image data including the information on the object and generates behavior tendency data including at least one of position distribution data and motion data of the moving object. , And

And an environment parameter estimator for estimating an optimal environment parameter optimized for growth of the moving object by applying the behavioral propensity data and the environment data of the kennel to an optimal environment estimation algorithm based on a machine learning.
The method of claim 1,

The image analyzer,

And dividing the image data into a plurality of areas, and calculating at least one of an activity amount and a distribution amount of the moving object for each of the plurality of areas to generate the behavior tendency data.
The method of claim 2,

The image analyzer,

Detecting the edge of the object from the image data, and calculating the distribution of the object for each of the plurality of regions based on the detected edge and the previously stored comparison database to generate the position distribution data. Farm environment management terminal comprising an analysis unit.
The method of claim 2,

The image analyzer,

And a second analysis unit configured to calculate a motion vector for the pixel of the image data, calculate activity of the object for each of the plurality of regions based on the motion vector, and generate the motion data.
The method of claim 1,

The environmental parameter estimator,

Generating a training data set by combining the behavioral propensity data and the environment data, and learning the optimal environment estimation algorithm using the training data set,

The environmental data,

At least one of the amount of feed, feed intake, age, moving object weight, temperature, humidity, carbon dioxide, outside temperature of the kennel,

The environmental parameter estimator,

Generate the training dataset based on a correlation between at least one of the feeding amount, feeding intake, age, moving object gain, temperature, humidity, carbon dioxide, and external temperature of the kennel and at least one of the location distribution data and the motion data. Farm environment management terminal to say.
The method of claim 1,

The environmental parameter estimator,

Computing data reliability based on the behavior propensity data and environmental data and a confidence range corresponding to each, and generating an alarm signal for the behavior propensity data and environmental data according to a result of comparing the data reliability with a predetermined threshold value. ,

A storage environment management terminal for storing the determination result of the user for the alarm signal and calculating the data reliability using the stored determination result of the user when the behavior tendency data and the environment data corresponding to the alarm signal are input.
The method of claim 1,

The environmental parameter estimator,

And a kennel environment management terminal for calculating a final environment parameter based on at least one of a recommended environment parameter and the optimum environment parameter based on the environment data.
The method of claim 7, wherein

The environmental parameter estimator,

Determining whether the kennel environment is deteriorated when the recommended environment parameter is applied based on the recommended environment parameter and the environment data;

If it is determined that the environment of the kennel deteriorates, the optimal environmental parameter is determined as the final environmental parameter,

If it is determined that the environment for the kennel is improved, the final environmental parameter is calculated by combining the optimal environment parameter with the recommended environment parameter.

The recommended environment parameter is

Farm environment management terminal generated by the learning server including a machine learning-based recommendation environment estimation algorithm.
The method of claim 8,

The learning server,

Combining the environmental data with the production of moving individuals in the kennel to generate a second training dataset,

The environmental parameter estimator,

And a kennel environment management terminal learning the recommendation environment estimation algorithm using the second training dataset.
The method of claim 8,

The learning server,

Generate a second training dataset by combining environmental data for the other kennels other than the kennel and the production of the moving object, and learning a recommended environment estimation algorithm according to at least one of a date, a region, and a place;

The environmental parameter estimator,

And a kennel environment management terminal learning the recommendation environment estimation algorithm using the second training dataset.