KR102014353B1 - Smart farm livestock management system based on machine learning - Google Patents

Abstract

The present invention relates to a system which photographs the interior of a cattle shed in real time to identify the breed of livestock raised in the cattle shed and the number of individual animals, analyzes an image of each individual animal to estimate the weight, and comprehensively combines individual animal number information and weight information generated by such a process to use the information in management and shipping of livestock. According to the present invention, the smart farm livestock management system based on machine learning comprises: a plurality of photographing units to generate image information of livestock raised in a cattle shed; a database to store comparison information to analyze image information such as the shape and behavioral characteristics of livestock by breed and the weight of livestock by breed and size; an image analysis unit to comprehensively analyze information of a plurality of image collected in real time based on the comparison information to generate individual animal number information of livestock in the cattle shed and weight information of each individual animal; a classifying unit to classify livestock by grade based on the individual animal number information and the weight information and calculate the number of individual animals of shippable livestock; and one or more terminal devices capable of checking the individual animal number information, the weight information, the number of individual animals of livestock classified by grade, and the number of individual animals of shippable livestock.

Description

Smart farm livestock management system based on machine learning

The present invention relates to a livestock management system that grasps the status of livestock inside the livestock house and utilizes it for management and shipment. More specifically, the inside of the livestock is taken in real time to grasp the breed and the number of livestock raised in the livestock house. The present invention relates to a system that analyzes an image of an individual and estimates weight, and aggregates population information and weight information generated through such a process and uses the same for management and release of livestock.

In general, livestock raised in livestock farms frequently increase or decrease the number of people due to childbirth or mortality, which may cause errors in the counting status due to various reasons such as delays or omissions of reports. It is very cumbersome and difficult to manage the livestock management history.

The invention related to the management of livestock for solving such a problem is described in "Livestock history information management system and its management method" of the Republic of Korea Patent Publication No. 10-2013-0016628 and "Recognition of livestock individual information" of the Republic of Korea Patent Publication No. 10-1342158 And easy-to-manage management system ”,“ Livestock and livestock management system ”of Korean Patent Publication No. 10-1651771 and“ Livestock delivery system and livestock delivery method ”of Korean Patent Application Publication No. 10-2017-0057916 It has been published.

The "Livestock history information management system and its management method" of the Republic of Korea Patent Publication No. 10-2013-0016628 includes a livestock identification information register to register the authenticated identification information of individual livestock, livestock history management to collect and manage the livestock history information The invention has been proposed a system and method configured to include a server and a livestock shipment management unit for generating information according to the shipment and movement of livestock to systematically and collectively manage the history of livestock, the Republic of Korea Patent Publication No. 10 -“House management system that makes it easier to recognize and manage livestock object information” in -1342158, which includes livestock information display boards installed at the entrances of livestock houses, livestock information input / output units for inputting and outputting individual identification information and feed intake of each animal in the house. To include a control unit for calculating and transmitting or outputting the feed amount of the animal feed. The present invention has been proposed to provide a system for easily recognizing and managing individual information of individual livestock in a livestock house, easily conveying the livestock operation status to outsiders, and easily identifying the amount of feed required. .

In addition, the "Livestock and Livestock Management Device" of the Republic of Korea Patent Publication No. 10-1651771 has a collector, lineage management information and identification information for collecting the identification information and individual management information of the livestock in the barn from the terminal installed in the livestock house It is configured to include a pedigree management unit for generating pedigree pedigree based on the pedigree, and a subject management unit for providing shipment information based on the individual management information, and collects livestock status and individual management information, pedigree information and livestock management information, etc. An invention relating to an apparatus for performing systematic management of a livestock house has been proposed, and the "livestock delivery system and livestock shipment method" of the Republic of Korea Patent Publication No. 10-2017-0057916 is to measure the weight of the livestock ingesting the feed And a sorter that sends the livestock out of the house according to the weight of the livestock measured in the feed It is proposed that the invention is directed to a system and method that is configured to include and enable convenient and rapid determination of whether individual livestock is shipped.

However, some of the prior arts, but it is possible to determine the number of livestock raised in the barn, the weight of each livestock problem can not be accurately estimated, accordingly the problem that the weight of the livestock shipped may not meet the standard There was a problem that the other part of the method of measuring the weight of the livestock is limited, so it is not possible to quickly measure the weight of all livestock raised in the barn.

Therefore, in order to solve such a problem, there is a need for an invention regarding a technology that can be utilized for the management and shipment of livestock by accurately estimating the number of livestock and the weight of each individual raised in a livestock house.

Republic of Korea Patent Publication No. 10-2013-0016628 (2017. 05. 26) Republic of Korea Patent Publication No. 10-1342158 (Dec. 10, 2013) Republic of Korea Patent Publication No. 10-1651771 (2016. 08. 22) Republic of Korea Patent Publication No. 10-2017-0057916 (May 26, 2017)

Machine farming-based smart farm livestock management system according to the present invention is a technology proposed to solve the problems of the preceding inventions,

The above-mentioned inventions have been able to grasp the number of livestock raised in the barn, but the weight of each livestock could not be accurately estimated or quickly measured,

As a result, there has been a problem that the weight of livestock to be shipped is less than the standard, the purpose is to propose a solution.

Machine learning based smart farm livestock management system according to the present invention to achieve the above object,

A plurality of photographing units generating image information of the livestock raised in the barn; A database storing comparative information for analyzing image information such as shape, behavioral characteristics of livestock according to breed, weight of livestock according to breed and size; An image analyzer configured to comprehensively analyze a plurality of image information collected in real time based on the comparison information to generate population information and weight information of each animal in the barn; A classification unit for classifying livestock by grade based on the population information and the weight information, and calculating a population of the domestic animals that can be shipped; At least one terminal device capable of confirming the population number, the weight information, the number of livestock classified into classes and the number of shippable livestock; It proposes a machine learning based smart farm livestock management system, characterized in that configured to include.

Machine learning based smart farm livestock management system according to the present invention,

By comparing the image information of the livestock generated by the filming unit and the comparison information stored in the database, the number of livestock raised in the barn can be calculated, and the weight of each individual can be estimated quickly.

Using the estimated weight of the livestock, the effect of easily identifying the number of livestock livestock has occurred.

Based on the error value between the estimated weight of the livestock and the measured weight of the livestock, the effect of accurately estimating the weight of the livestock occurred.

1 is an overall system configuration according to an embodiment of a machine learning based smart farm livestock management system according to the present invention.
2 is a flowchart illustrating a process of generating information using image information and comparison information of an image analysis unit of a machine learning based smart farm livestock management system according to the present invention.
Figure 3 is a detailed configuration of the image analysis unit of the machine learning based smart farm livestock management system according to the present invention.
Figure 4 is a flow chart showing a process of determining the breeding unit of the animal learning based smart farm livestock management system according to the present invention.
Figure 5 is a further detailed configuration of the image analysis unit of the machine learning based smart farm livestock management system according to the present invention.
Figure 6 is an exemplary view showing a space inside the barn is repeatedly photographed by a plurality of photographing unit of the machine learning based smart farm livestock management system according to the present invention.

The present invention relates to a livestock management system that grasps the status of livestock inside the livestock house and utilizes it for management and shipment.

As shown in Figure 1, a plurality of photographing unit 100 for generating image information of the livestock raised in the barn; A database 110 in which comparison information for analyzing image information such as shape, behavioral characteristics, and weight of livestock according to breed and size according to the breed is stored; An image analysis unit 120 generating comprehensive population analysis of a plurality of image information collected in real time based on the comparison information to generate population information and weight information of each animal in the barn; A classification unit 130 for classifying livestock by grade based on the population information and the weight information, and calculating the population of the livestock; At least one terminal device 140 capable of confirming the population information, the weight information, the population of livestock classified by grade, and the number of shippable livestock; It relates to a machine learning based smart farm livestock management system comprising a.

Hereinafter, with reference to the accompanying drawings will be described embodiments of the present invention;

The plurality of photographing units 100 may include a known camera capable of capturing an image, and may further include a digital video recorder (DVR) for recording the photographed image. It is preferable to be spaced apart from each other so as to shoot at the same time.

The image information generated by the photographing unit 100 is generated for the purpose of distinguishing the breeds of livestock breeding in the barn, calculating the total number of individuals and the number of breeds, and estimating the weight of each of the livestock. The type and behavior characteristics of the livestock according to the variety of the plurality of information constituting the comparison information stored in the database 110 is used for the classification of the breed and the number of individual breeds, and the comparison information is used to estimate the weight of each livestock. The weight of the livestock according to the breed and size is used among the plurality of pieces of information.

That is, as shown in FIG. 2, the image analyzer 120 uses the plurality of image information simultaneously generated by each of the plurality of photographing units 100 and the consonant using comparison information stored in the database 110. Categorize the breeds of livestock in the farm, generate population information on the number of livestock by breed, and generate weight information on the weight of each individual.

In this case, the population information may include not only the number of livestock of each breed, but also the number of all livestock breeding in the barn, and the weight information may include the weight of each individual of each breed as well as the weight of the entire individual of the breed. The weight of all livestock may be included.

The terminal device 140 receives the number information and weight information generated by the image analysis unit 120 and transmitted in a wired or wireless communication manner and displays the information on the screen so that the user can determine the livestock status in the livestock house. Accordingly, the input and output of the information may be configured as any one of a known computer, a tablet PC, a smart phone, and the like, which can select and display desired information among the input information, but is not limited to one of these devices. .

In addition, the terminal device 140 may receive and display the image information generated by the photographing unit 100 on the screen, the identification number assigned to each of the livestock in the following identification unit 123 to be displayed on the screen The weight information generated by the image analyzer 120 may be displayed through the identification number or a separate display.

Therefore, the user can easily identify the shipmentable object by checking the weight information displayed on the screen of the terminal device 140 and ship it.

In addition, the population information and weight information generated by the image analysis unit 120 is used to calculate the number of livestock of the domestic animals can be shipped from the livestock raised in the barn, the classification unit 130 is all livestock raised in the barn By classifying, by classifying the livestock classified by the breed by weight according to a predetermined standard, by classifying the livestock classified by weight more than a predetermined weight of the livestock classified as a livestock available to the user through the terminal device 140 In addition, it is possible to understand in detail the status of livestock in the house and to identify the number of livestock.

In addition, the classification unit 130 may classify the livestock of more than a predetermined weight of the livestock classified by weight after being classified by breed as a livestock and at the same time may classify the livestock below a predetermined weight as non-shipping livestock. This can be further categorized into livestock, impending livestock, and non-shipping livestock.

In addition, as illustrated in FIG. 3, the image analyzer 120 includes a determination unit 121 that determines a breed of a livestock based on the shape and behavioral characteristics of the livestock shown in the image information. do.

As shown in FIG. 4, the determination unit 121 compares the shape of the livestock shown in the image information with the shape of the livestock according to the breed included in the comparison information to determine the similarity of each shape, and the determined shape similarity. Determine the breeds of livestock shown in the image information based on the comparison, and compare the behavioral characteristics of the livestock shown in the video information with the behavioral characteristics of the livestock according to the breed included in the comparison information to determine the similarity of each behavioral characteristic, and the determined behavior Based on the characteristic similarity, the breed of the animal shown in the image information can be determined.

That is, the shape of the livestock according to the breed included in the comparison information is a main element for determining the breed of the livestock, and the behavioral characteristics of the livestock according to the breed included in the comparison information are additional elements for determining the breed of the livestock.

In addition, as illustrated in FIG. 3, the image analyzer 120 may include an estimator 122 estimating the weight of the livestock based on the breed and size of the livestock. 122) estimates the weight of the livestock by comparing the size of the livestock shown in the image information with the weight of the livestock according to the breed and size included in the comparison information based on the breed of the livestock determined by the determination unit 121; .

However, since there is a high possibility that an error may occur between the estimated weight of the livestock and the actual weight of the livestock, the machine learning based smart farm livestock management system according to the present invention uses image learning using machine learning technology. We want to improve the accuracy of livestock weight estimation.

As described above, since the weight information generated by the image analyzer 120 may be displayed on the screen of the terminal device 140, the user may correct the error by directly modifying the displayed weight information, and the image analyzer 120 may improve accuracy of image information analysis based on an error value between the generated weight information and the corrected weight information.

That is, the image analyzer 120 may perform machine learning on the analysis of the image information based on an error value between the weight information and the actual livestock weight. In order to improve the accuracy of the livestock weight estimation, it is preferable to perform the modification of the weight information using the terminal device 140 a plurality of times.

In addition, the estimator 122 may measure the ratio of the head size and the body size of the livestock shown in the image information and estimate the weight of the livestock based on the ratio of the body size to the average head size of each breed. Since the ratio of the body size to the head size can be measured differently according to the angle at which the image is taken, the ratio of the body size to the head size is determined more than 10 times a plurality of times. Preferably, the ratio value is selected.

At this time, the ratio of the body size to the average head size of each variety is stored in the database 110 is utilized.

In addition, as shown in FIG. 5, the image analyzer 120 includes an identification unit 123 which assigns an identification number to each of the livestock appearing in the plurality of image information. The identification number assigned to each of the livestock by the unit 123 may be composed of only numbers, and may be composed of a combination of numbers and letters, and some of the numbers or letters constituting the identification numbers easily identify the breed of the livestock. It may be intended to, and to easily identify the weight of the livestock.

In addition, as illustrated in FIG. 5, the image analyzer 120 may include a detector 124 for detecting live animals overlapping with a plurality of image information collected in real time. 124 analyzes any one or more of breeds, postures, movements or movement times of all livestock appearing in the plurality of image information generated by the plurality of photographing units 100 to always match the breeds, postures, movements or movement times. Multiple livestock are considered the same livestock.

That is, as shown in FIG. 6, in order to photograph the inside of the barn without the blind spot, the plurality of photographing units 100 inevitably overlap the spaces generated by the two or more photographing units 100. ) Detects livestock located in overlapped spaces and can be used to calculate accurate livestock populations.

At this time, the posture of the livestock, which is a standard for detecting the livestock appearing in the plurality of image information, refers to a sitting posture, standing posture, posture conversion into a standing posture or posture conversion into a standing posture, The movement of the livestock, which is another standard, refers to the stopping or moving of the livestock, and the moving time of the livestock, which is another standard, means the time from when the livestock is stopped to moving and stopping again. 100) In consideration of each shooting direction, the moving direction of the livestock is preferably excluded from the determination object.

In addition, as shown in Figure 5, the image analysis unit 120 is characterized in that it comprises a correction unit 125 for correcting the identification number assigned to the livestock, the correction unit 125 is The detection unit 124 corrects such that only one identification number is assigned to one livestock by removing any one or more of the identification numbers duplicated on the livestock detected as duplicate appearing in the plurality of image information.

Therefore, the identification number assigned to the livestock continuously appearing only in the image information generated by any one of the plurality of photographing units 100 may be maintained, and two of the plurality of photographing units 100 may be maintained. Only one of the identification numbers duplicated in the livestock appearing in all the image information generated by the image pickup unit 100 can be maintained as it is, appearing in only one of the image information, but moved to appear only in the other image information Starting livestock may be deleted with a previously assigned identification number and assigned a new identification number or with a previously assigned identification number.

In addition, as shown in FIG. 5, the image analyzer 120 is based on at least one identification number assigned to each of the livestock, and based on the estimated weight of each livestock, the population information and the number of livestock in the barn. And a generation unit 126 for generating weight information of the individual.

First, the generation unit 126 collects the identification number assigned to each of the livestock in the identification unit 123, but calculates the number of all livestock breeding in the barn reflecting the identification number removed from the correction unit 125 In this case, the population information may be generated. As described above, the population information may include not only the number of livestock by breed but also the number of all livestock raised in the barn.

In addition, the generation unit 126 collects the weight of each of the livestock estimated by the estimation unit 122, and generates weight information on this, as described above, the weight information, as well as the weight of each individual by breed It may include the weight of the entire individual by breed and may include the weight of all livestock raised in the house.

The population information and the weight information are displayed on the terminal device 140 so that the user can easily grasp the current status of the livestock in the barn, and is also used for calculating the number of livestocks that can be shipped from the livestock raised in the barn. do.

Therefore, the generation unit 126 generates the number information and weight information, but generates a classification table that further specifies the two information so that the user can easily and specifically grasp the current status of livestock in the barn through the terminal device 140 In addition, the classification table for the implementation of these effects is the total number of livestock raised in the barn, the increase and decrease of the livestock population on the designated date, the production population on the designated date, the number of dead population on the designated date and the number of shipments on the designated date It can be configured to include, etc., the total number of individuals by each breed, the status of increase and decrease of the number of individuals by varieties by the specified date, the current state of production population by the variety of the specified date, the status of our population by the variety of the specified date and the number of shipments by varieties by the specified date It may be configured to include the current state.

However, since only the configuration of the present invention described above, it is impossible to grasp the status of the dead population based on the designated date and the number of dead population by the breed based on the designated date, so that the detection unit 124 may include the plurality of photographing units 100. It is characterized by detecting any one or more of the posture, movement or movement time of all livestock appearing in the plurality of image information generated in the diseased livestock or dead livestock.

That is, the detection unit 124 does not stand up for a predetermined time from the image information generated by the imaging unit 100 in order to detect a livestock disease that is diseased or maintain a long sitting state excessively compared to the standing time Can detect a livestock with a significantly lower activity or a significantly reduced activity compared to other individuals, and can lie down for a predetermined time from the image information to detect dead livestock. It is possible to detect livestock that does not detect movement.

The population of livestock and diseased livestock detected by the detection unit 124 may be reflected when the generation unit 126 generates the classification table. In particular, the population of dead livestock may be calculated as the total population. It is preferable to be excluded from.

In addition, since only the configuration of the present invention described above, it is impossible to grasp the current status of the number of shipments based on the designated date and the number of shipments by the variety of date based on the designated date, so that the generation unit 126 uses the classification unit 130. If the calculated number of livestock can be reduced and the identification number assigned to the livestock in the identification unit 123 is extinguished, this phenomenon is determined as the shipment of the livestock, and the number of shipped livestock is reflected in the classification table. do.

However, it is preferable that the generation unit 126 reflects the decrease in the number of livestock animals and the number of dead livestock with the death of the identification number.

In addition, the generation unit 126 may determine the number of livestock produced in consideration of both the increase and decrease of the population, the disappearance of the identification number according to the disease or mortality, and the number of shipped livestock, which is produced based on the designated date of the classification table. It can be reflected in the current status of population and production population by variety based on the designated date.

In addition, the image analysis unit 120 is characterized in that it comprises a determination unit for determining the status of the livestock and the release time of the livestock based on the change history of the weight information included in the identification number assigned to each livestock. .

Specifically, when the weight information included in the identification number assigned to each livestock indicates a decreasing trend for a predetermined period of time, the determination unit determines that the condition is more than the condition of the livestock such as a disease occurrence, and the weight information included in the identification number is In the case of an increase in a predetermined period of time, it is determined that the livestock is in a normal state, and the estimated time to release the livestock and the undeliverable livestock can be calculated by reflecting the increase in the weight of the livestock during the predetermined period. .

Therefore, the user can determine the number of livestock determined to be abnormal by using the terminal device, check the identification number of the livestock determined to be abnormal, and directly check the status of the livestock through video information to isolate It is possible to take measures such as treatment, treatment, etc., and to use the expected time of livestock release management.

In addition, the machine learning based smart farm livestock management system according to the present invention is characterized in that it comprises a plurality of laser pointers provided in each of the plurality of photographing unit 100, the laser to the cattle classified as shippable It characterized in that it comprises a control unit for controlling the laser pointer to launch.

As described above, some of the numbers or letters constituting the identification number assigned to each of the livestock by the identification unit 123 may be for easily identifying the breed of the livestock, and to easily identify the weight of the livestock. Since the identification number is based on the image information generated by the photographing unit 100, the controller controls the laser pointer based on the image information and the identification number to the livestock classified as being shippable. Let the laser fire.

Accordingly, the user can recognize and ship the livestock specified by the laser and can ship the livestock, and since the laser must not be simultaneously fired toward all the livestock, the controller controls the laser pointer to sequentially ship the livestock. So that the laser can be fired.

The above-described embodiments are provided by way of example so that the technical spirit of the present invention can be sufficiently delivered to those skilled in the art to which the present invention pertains. It is not limited and may be embodied in other forms.

Parts not related to the description are omitted in the drawings in order to clearly describe the present invention, and like reference numerals denote like elements throughout the specification.

100: shooting unit
110: database
120: image analysis unit
121: judgment unit 122: estimation unit
123: identification unit 124: detection unit
125: correction unit 126: generation unit
130: classification unit
140: terminal device

Claims (7)

A plurality of photographing units 100 generating image information of livestock raised in a barn;
A database 110 in which comparison information for analyzing image information such as shape, behavioral characteristics, and weight of livestock according to breed and size according to the breed is stored;
Comprehensive analysis of a plurality of image information collected in real time based on the comparison information to generate the population information of the livestock in the barn and the weight information of each individual, based on the shape and behavior characteristics of the livestock shown in the image information An image analyzer 120 including a determination unit 121 for determining a breed of a livestock;
A classification unit 130 for classifying livestock by grade based on the population information and the weight information, and calculating the population of the livestock; And,
At least one terminal device 140 capable of confirming the population information, the weight information, the population of livestock classified by grade, and the number of shippable livestock; Characterized in that comprises a,
The image analyzer 120,
An estimator 122 estimating the weight of the livestock based on the breed and size of the livestock shown in the image information;
An identification unit (123) for assigning identification numbers to each of the livestock appearing in the plurality of image information;
A detector 124 for detecting a livestock duplicated in a plurality of image information collected in real time based on at least one of a breed, posture, movement, or movement time of the livestock;
A correction unit 125 which removes any one or more of the identification numbers duplicately assigned to the livestock and corrects only one identification number to one livestock;
A generation unit 126 based on at least one identification number assigned to each of the livestock, and based on the estimated weight of each of the livestock, for generating population information and live weight information of the livestock in the barn; Machine learning based smart farm livestock management system, characterized in that comprises a.
delete delete delete delete The method of claim 1,
On the screen of the terminal device 140,
Weight information generated by the image analyzer 120 is matched and displayed on each of the livestock displayed in the image information,
Machine learning based smart farm livestock management system, characterized in that the weight information displayed on the screen can be modified.
The method according to claim 1 or 6,
The image analyzer 120,
Based on the error value between the weight information and the actual livestock weight,
Machine learning based smart farm livestock management system characterized in that the machine learning (Machine Learning) analysis of the image information.

Images