KR102180077B1 - Livestock weight estimation system using livestock weight estimation application based on 3D images and Livestock weight estimation method using the same - Google Patents

Abstract

The present invention relates to a livestock weight estimation system using a livestock weight estimation application based on three-dimensional images and a livestock weight estimation method using the same to achieve excellent weight estimation accuracy and reliability. According to the present invention, the livestock weight estimation system using a livestock weight estimation application based on three-dimensional images comprises a terminal including a three-dimensional sensor to photograph livestock and a livestock weight estimation application to estimate the weight of livestock by using a livestock frame extracted through the three-dimensional sensor. The livestock weight estimation application includes a sensor photographing unit which operates the three-dimensional sensor to photograph livestock and extract one or more livestock frames and a weight estimation unit which derives an estimated weight by using the extracted livestock frames. Also, the livestock weight estimation method using the livestock weight estimation system using the livestock weight estimation application based on three-dimensional images comprises: (a) a photographing step of photographing livestock through the terminal including the three-dimensional sensor to photograph livestock and the livestock weight estimation application to estimate the weight of livestock by using livestock frames extracted through the three-dimensional sensor and extracting one or more livestock frames; and (b) a weight estimation step of using the extracted livestock frames to derive an estimated weight.

Description

Livestock weight estimation system using livestock weight estimation application based on 3D images and Livestock weight estimation method using the same}

The present invention relates to a livestock weight estimation system using a livestock weight estimation application based on a 3D image, and a livestock weight estimation method using the same. More specifically, a candidate list method using a plurality of frames obtained from a 3D sensor and By estimating the weight of livestock through a multiple regression method, a livestock weight estimation system using a livestock weight estimation application based on a 3D image with excellent weight estimation accuracy and reliability, and a livestock weight estimation method using the same can be provided.

In the case of livestock industry, regular weight management is required for feeding management of individual breeding animals.

In particular, in the case of pig farms, the standard standard is very important at the time of shipment, and according to the shipping amount of pigs that meet the standard standard, the income of the farm has a very large difference. Pigs are graded according to the quantitative standard according to the weight and fat thickness, the qualitative standard according to the fat distribution and meat color of the pork. Generally, pigs weighing 115kg to 120kg are called standard pigs.

It is very important to select shipping pigs by accurately measuring their weight because if they satisfy the weight of standard pigs, they can receive a higher grade.

For this, the need for periodic weighing or monitoring is required in the field.

Currently, the weight of pigs is measured by thoracic positioning method and pig pen.

The thoracic positioning method converts the weight by applying the value obtained by measuring the chest of a pig with a tape measure to a weight calculation formula. It is used as an advantage that no special facilities need to be installed, but the measurement error is very large.

In addition, the pig machine measures the weight of pigs in an enclosed space after installing an accessory device on the scale that measures the weight of pigs. To measure the weight of pigs, all pigs are guided to the pig machine and then There is a problem of having to hold down a certain amount of time within the prison period, and in this process, it takes about 10 minutes or more for a single worker to measure the weight of one animal, and there is a problem that electronic scales are frequently used due to pig manure. There were breakdowns and difficulties in maintenance.

In addition, due to the decrease in the population of farm households and an aging population, there is a shortage of manpower, which requires a countermeasure.

Therefore, by simply and accurately measuring the weight of the pig, it is possible to continuously manage the weight of the pig and reduce the labor of the farmer, and through this, a technology for accurately predicting the weight of the pig at the time of shipment is required.

Recently, a system or method for predicting the volume and weight of livestock through images by photographing livestock has been developed, but errors due to obstacles (sunlight, movement of livestock, etc.) and other objects (ground, structures, etc.) Therefore, there was a problem of poor accuracy.

In order to solve the above problem, the present invention estimates the weight of livestock through a candidate list method and a multiple regression method using a plurality of frames acquired from a 3D sensor, based on a 3D image with excellent weight estimation accuracy and reliability. An object of the present invention is to provide a livestock weight estimation system using a livestock weight estimation application and a livestock weight estimation method using the same.

In order to solve the above problems, in the livestock weight estimation system using the livestock weight estimation application based on a 3D image according to an embodiment of the present invention, a 3D sensor for photographing livestock and the extracted through the 3D sensor Including a terminal including a livestock weight estimation application for estimating the weight of livestock using the livestock frame, wherein the livestock weight estimation application is a sensor photographing unit for extracting one or more livestock frames by photographing livestock by operating the 3D sensor and It is possible to provide a livestock weight estimation system including a weight estimation unit that derives the estimated weight using the extracted livestock frame.

In addition, the weight estimating unit may include a conversion unit for converting the extracted livestock frame into 3D coordinate data; A pre-processing unit for pre-processing the converted 3D coordinate data to generate 3D livestock data, and a weight extraction unit for deriving an estimated weight from the 3D livestock data.

In addition, the preprocessor may include a factor extractor for extracting a factor from the 3D coordinate data.

In addition, the preprocessor may further include a validity verification unit determining validity of the 3D coordinate data and a noise removing unit removing noise of the 3D coordinate data.

In addition, the factor includes at least one of head direction, floor angle, focal length, height, length 1, chest circumference, median, posterior circumference, average trunk circumference, total circumference, and number of cloud points of 3D livestock data. can do.

(Here, length 1 is the distance from the front leg to the buttocks with the head removed, and length 2 is the distance between the front and hind legs.)

In addition, the weight extraction unit generates a candidate list by selecting one or more of the three-dimensional livestock data according to a factor, and determines the number of cases of conditional satisfaction of the generated candidate list to derive the first weight of the three-dimensional livestock data. A candidate list unit; An algorithm deriving unit for deriving a second weight by applying a multiple regression formula to the 3D livestock data, and a final deriving unit for deriving an estimated weight using the first and second weights.

In addition, the candidate list unit selects one or more three-dimensional livestock data as candidates according to a factor, and determines the number of cases of conditional satisfaction of the list generation unit and the generated candidate list to generate a candidate list. It may include a list processing unit for deriving the first weight of the dimensional livestock data.

In addition, the list generation unit generates a first candidate list by comparing at least one of the head direction, the floor angle, the focal length, the length 1, and the total body circumference of the 3D livestock data with livestock standard data, and generates a first candidate list, and the 3D livestock data The second candidate list is created by comparing one or more of the head direction, total body circumference, height, length 1 and median of the animal with the livestock standard data, and among the head direction, length 1, chest, median and posterior of the 3D livestock data. At least one is compared with livestock standard data to generate a third candidate list.

In addition, when generating the first, second, and third candidate lists, the list generator may also generate first, second, and third candidate backup lists.

In addition, the algorithm derivation unit derives the second weight through Equation 1 below using the height, length 2, the average of the body circumference, and the number of cloud points of the 3D livestock data among the factors of the 3D livestock data. It features.

[Equation 1]

2nd weight = m ₁ + (h ₁ × height) + (2ㅣ ₁ × length2) + (g ₁ × average body circumference) + (p ₁ × number of cloud points in 3D livestock data)

(Where m ₁ is the weight constant, h ₁ is the height constant, 2ㅣ ₁ is the length 2 constant, g ₁ is the circumferential average constant, and p ₁ is the cloud point constant of 3D livestock data)

In addition, the final derivation unit is characterized in that the estimated weight is derived according to an absolute value of a value obtained by subtracting the first weight from the second weight.

In addition, in the livestock weight estimation method using a livestock weight estimation system using a livestock weight estimation application based on a 3D image according to an embodiment of the present invention, (a) through a 3D sensor for photographing livestock and the 3D sensor A photographing step of extracting one or more livestock frames by photographing livestock through a terminal including a livestock weight estimation application that estimates the weight of livestock using the extracted livestock frames, and (b) calculating the estimated weight using the extracted livestock frames. It is possible to provide a method for estimating livestock weight including the deriving weight estimation step.

Here, the step (b) includes a transformation step of converting the extracted livestock frame into 3D coordinate data; It may include a pre-processing step of pre-processing the converted 3D coordinate data to generate 3D livestock data, and a weight derivation step of deriving an estimated weight from the 3D livestock data.

In addition, the preprocessing step may include a factor extraction step of extracting a factor from the 3D coordinate data.

In addition, before the factor extraction step, a validity verification step of determining validity of the 3D coordinate data and a noise removal step of removing noise of the 3D coordinate data may be further included.

In addition, in the weight derivation step, a candidate list is generated by selecting one or more of the three-dimensional livestock data according to a factor, and the first weight of the three-dimensional livestock data is determined by determining the number of cases in which the conditions of the generated candidate list are satisfied. A first weight extraction step of deriving; A second weight derivation step of deriving a second weight by applying a multiple regression formula to the 3D livestock data, and an estimated weight derivation step of deriving an estimated weight using the first weight and the second weight.

In addition, in the first weight extraction step, a list generation step of generating a candidate list by selecting one or more three-dimensional livestock data as candidates according to a factor, and determining the number of cases in which the conditions of the generated candidate list are satisfied, Accordingly, a list processing step of deriving the first weight of the 3D livestock data may be included.

In addition, the list generation step is a first candidate step of generating a first candidate list by comparing at least one of the total of the head direction, the floor angle, the focal length, the length 1 and the body circumference of each of the three-dimensional livestock data with livestock standard data. ; A second candidate step of generating a second candidate list by comparing at least one of the head direction, total body circumference, height, length 1 and median of each of the 3D livestock data with livestock standard data, and the head of each of the 3D livestock data It may include a third candidate step of generating a third candidate list by comparing at least one of the direction, length 1, chest, median and rear with livestock standard data.

In addition, in the first, second, and third candidate steps, when generating the first, second, and third candidate lists, respectively, the first, second, and third candidate backup lists are also generated.

In addition, in the list processing step, when the first, second, and third candidate lists all have at least one 3D livestock data, a first number of determining whether there is one 3D livestock data in the first candidate list Judgment step; If there is one 3D livestock data of the first candidate list, the weight of the 3D livestock data of the first candidate list is selected as a predicted weight, and then the predicted weight of the 3D livestock data of the second and third candidate lists is selected. A first detection step of finding whether there is data having the same weight as and a first derivation step of deriving a first weight according to whether both the second and third candidate lists have 3D livestock data having the same weight as the predicted weight. It may include.

In addition, in the list processing step, if there is not one 3D livestock data of the first candidate list in the first repair plate step, the length 2 of the 3D livestock data to be derived from the estimated weight is a length determination step of determining whether or not it is within the range of the length 2 of the nth 3D livestock data; If it is within the range of the length 2 of the nth 3D livestock data of the first candidate list, after selecting the weight of the nth 3D livestock data of the first candidate list as a predicted weight, the second and third candidate lists A second detection step of finding whether there is data having the same weight as the predicted weight among the three-dimensional livestock data of; Depending on whether both the second and third candidate lists have 3D livestock data having the same weight as the predicted weight, a temporary list is generated to add the nth 3D livestock data of the first candidate list to a first temporary list Step and a second derivation step of deriving a first weight according to the number of three-dimensional livestock data in the generated first temporary list, wherein the length determination step, the second detection step, and the temporary list generation step are repeated n times. It characterized in that it is made.

(Where, n is 2 or more, which is the number of 3D livestock data in the first candidate list)

In addition, in the list processing step, when only the first and second candidate lists have one or more 3D livestock data, the height of the 3D livestock data to be estimated weight is the first 3D livestock data of the first candidate list. It may include a third derivation step of deriving the first weight according to whether it is greater than the height of the dimensional livestock data.

In addition, in the list processing step, when only the second and third candidate lists have one or more three-dimensional livestock data, a second number determination step of determining whether there are more than one three-dimensional livestock data in the third candidate list, and If there are more than one 3D livestock data in the third candidate list, a 4-1 derivation step of deriving the average weight of the 3D livestock data of the second candidate list as the first weight.

In addition, in the list processing step, when the number of 3D livestock data of the third candidate list does not exceed one in the second number determination step, the weight of the kth 3D livestock data of the second candidate list is 3 Calculation step of classifying and calculating the 3D livestock data of the second candidate list according to whether or not the weight of the 3D livestock data of the candidate list is greater, and the height of the 3D livestock data to derive the estimated weight is the third candidate A 4-2 derivation step of deriving the first weight according to whether or not the height of the 3D livestock data of the list is higher, wherein the calculation step is repeated k times.

(Here, k is 1 or more, and is the number of 3D livestock data in the second candidate list)

In addition, in the list processing step, when only the first and third candidate lists have one or more 3D livestock data, the height of the 3D livestock data to be estimated weight is the first 3rd of the third candidate list. It may include a fifth derivation step of deriving the first weight according to whether the height of the dimensional livestock data or more.

In addition, the list processing step includes a sixth derivation step of deriving the weight of the first 3D livestock data of the first candidate list as the first weight when only the first candidate list has one or more 3D livestock data. Can include.

In addition, the list processing step includes a seventh derivation step of deriving the average weight of the 3D livestock data of the second candidate list as the first weight when only the second candidate list has one or more 3D livestock data. can do.

In addition, in the list processing step, when only the third candidate list has one or more 3D livestock data, the first weight is derived for deriving the first weight according to whether there is one 3D livestock data in the third candidate list. It may include an eighth derivation step.

The livestock weight estimation system using the livestock weight estimation application based on the 3D image and the livestock weight estimation method using the same according to an embodiment of the present invention as described above is a candidate list method using a plurality of frames obtained from the 3D sensor By estimating the weight of livestock through a multiple regression method, the accuracy and reliability of weight estimation can be excellent.

Therefore, additional equipment is not required to measure the weight of livestock, and breeding costs can be reduced through feed control through continuous weight management of livestock, and the time of shipment can be accurately predicted, thereby increasing the profits of farmers.

In addition, there is no hassle of inducing livestock to be congested for a certain period of time in order to measure body weight, so it is possible to solve problems caused by a lack of manpower of farmers, an aging manpower, and an increase in scale.

In addition, it can be applied not only to pigs, but also to various livestock such as chickens and cows, so its utility is expected to expand.

1 is a block diagram showing a livestock weight estimation system using a livestock weight estimation application based on a 3D image according to an embodiment of the present invention.
2 is an exemplary view showing a preview screen output from a terminal of a livestock weight estimation system using a livestock weight estimation application based on a 3D image according to an embodiment of the present invention.
Figure 3 is a block diagram showing the configuration of the weight estimation unit of Figure 1;
4 is an exemplary diagram showing a factor.
Figure 5 is a block diagram showing the configuration of the weight extraction unit of Figure 3;
6 is a flowchart schematically illustrating a method of estimating livestock weight using a livestock weight estimation system using a livestock weight estimation application based on a 3D image according to an embodiment of the present invention.
7 is a flowchart sequentially showing step S2 of FIG. 6;
8 is a flow chart showing step S22 of FIG. 7.
9 is a flowchart sequentially showing step S220 of FIG. 8.
10 is a flowchart illustrating step S2201 when the first, second, and third candidate lists satisfy conditions.
11 is a flowchart showing a process in which step S22012a of FIG. 10 is performed.
12 is a flowchart illustrating a process in which step S22016a of FIG. 10 is performed.
13 is a flowchart illustrating a process in which step S22010b is performed from step S2201 when the first and second candidate lists are satisfied with conditions.
14 is a flowchart showing step S2201 when the second and third candidate lists satisfy conditions.
15 is a flowchart showing a process in which step S22013c of FIG. 14 is performed.
16 is a flowchart illustrating a process in which step S22010d is performed in step S2201 when the first and third candidate lists are satisfied with conditions.
17 is a flowchart illustrating a process in which step S22010e is performed in step S2201 when only a third candidate list is satisfied with a condition.
18 is a flowchart illustrating a process in which step S22010f is performed in step S2201 when there is no list of candidates for satisfying conditions.

Hereinafter, the description of the present invention with reference to the drawings is not limited to a specific embodiment, and various transformations may be applied and various embodiments may be provided. In addition, the content described below should be understood to include all conversions, equivalents, and substitutes included in the spirit and scope of the present invention.

In the following description, terms such as first and second are terms used to describe various elements, and their meanings are not limited thereto, and are used only for the purpose of distinguishing one element from other elements.

The same reference numbers used throughout this specification denote the same elements.

Singular expressions used in the present invention include plural expressions unless the context clearly indicates otherwise. In addition, terms such as "include", "include" or "have" described below are intended to designate the presence of features, numbers, steps, actions, components, parts, or combinations thereof described in the specification. It is to be construed and not to preclude the possibility of the presence or addition of one or more other features, numbers, steps, actions, components, parts, or combinations thereof.

Unless otherwise defined, all terms used herein, including technical or scientific terms, have the same meaning as commonly understood by one of ordinary skill in the art to which the present invention belongs. Terms as defined in a commonly used dictionary should be interpreted as having a meaning consistent with the meaning in the context of the related technology, and should not be interpreted as an ideal or excessively formal meaning unless explicitly defined in this application. Does not.

In addition, in the description with reference to the accompanying drawings, the same reference numerals are assigned to the same components regardless of the reference numerals, and redundant descriptions thereof will be omitted. In describing the present invention, when it is determined that a detailed description of related known technologies may unnecessarily obscure the subject matter of the present invention, a detailed description thereof will be omitted.

In the present invention, the livestock weight estimation application is preferably provided in a mobile communication terminal including a smart phone that communicates through a wireless communication network, but may also be provided in a computer environment that communicates through a wired communication network. Hereinafter, a description will be given focusing on a mobile communication terminal.

Hereinafter, embodiments of the present invention will be described in detail with reference to FIGS. 1 to 18.

1 is a block diagram showing a livestock weight estimation system using a livestock weight estimation application based on a 3D image according to an embodiment of the present invention, Figure 2 is a livestock based on a 3D image according to an embodiment of the present invention An exemplary diagram showing a preview screen output from a terminal of a livestock weight estimation system using a weight estimation application, FIG. 3 is a block diagram showing the configuration of the weight estimation unit of FIG. 1, and FIG. 4 is an exemplary diagram showing a factor, 5 is a block diagram showing the configuration of the weight extraction unit of FIG. 3.

1, a livestock weight estimation system using a livestock weight estimation application based on a 3D image according to an embodiment of the present invention includes a terminal 1, the terminal 1 is a 3D sensor 10 and Livestock weight estimation application 20 may be included.

The terminal 1 may be a mobile terminal of a user, such as a breeder or manager of a farm, and may be easily applied to a PC tablet other than a mobile terminal.

As described above, the terminal 1 is implemented as a mobile terminal to facilitate mobility and convenience, so that an administrator can easily measure the weight of livestock.

The 3D sensor 10 photographs livestock to secure a three-dimensional livestock frame. To this end, the 3D sensor 10 may be provided with various sensors capable of capturing a 3D image such as a lidar.

This 3D sensor 10 may be operated by the sensor photographing unit 22 of the livestock weight estimation application 20.

The livestock weight estimation application 20 is capable of estimating the weight of livestock using livestock frames extracted through the 3D sensor 10, and refers to a general application based on Android and iOS, but the terminal 1 or wired/wireless service Depending on the type, it may be provided as a general application or a web service-based application. As a method of providing, the server may be accessed and downloaded through the terminal 1, or may be downloaded and installed through an online application market (eg, Android market, Apple store, online market of a telecommunication company, etc.).

Specifically, the livestock weight estimation application 20 may include a data storage unit 21, a sensor photographing unit 22, a weight estimation unit 23, and a management unit 24.

The data storage unit 21 may store livestock standard data generated by using the livestock data collected in advance, and store biometric information and estimated weight data of livestock input from a user.

Here, the livestock standard data is data standardized by months or weights by analyzing livestock data collected in advance by measuring livestock in livestock farms and taking 3D photographs, and may include three-dimensional livestock data by month or by weight, The livestock data collected in advance may include livestock biometric information, 3D model data, and weight.

In addition, the biometric information of livestock is the biometric information of livestock for which the weight is to be measured, and may include at least one of a livestock classification number, species, sex, and number of months, and preferably includes all.

The sensor photographing unit 22 may extract one or more livestock frames by photographing livestock by operating the 3D sensor 10.

First, when the livestock weight estimation application is executed, the sensor photographing unit 22 operates the 3D sensor 10 to detect livestock. When livestock is detected, the 3D sensor 10 is initialized to photograph livestock. In this case, a preview screen as shown in FIG. 2 may be provided on the screen of the terminal 1, and capture may be performed as the user presses the photographing button by activating the photographing button.

That is, when the user presses the photographing button to input photographing start information, the photographing button may be deactivated and one or more livestock frames may be extracted through capture.

It is preferable that the capture consists of continuous capture so that a plurality of livestock frames can be extracted, and continuous capture may be performed according to the value of the number of consecutive captures. The continuous capture count value may be 10 to 12, but is not limited thereto.

The livestock frame obtained here may include 2D information and depth information.

By extracting livestock frames through the capture in this way, it is possible to minimize performance degradation.

On the other hand, when the livestock is not detected, the sensor photographing unit 22 may stop the operation and proceed again.

The weight estimation unit 23 may derive the estimated weight using the extracted one or more livestock frames.

Referring to FIG. 3, the weight estimation unit 23 may include a conversion unit 230, a preprocessor 231, and a weight extraction unit 232.

After arranging the livestock frames, the conversion unit 230 may extract points and convert them into 3D coordinate data. Such 3D coordinate data may be stored in the data storage unit 21.

The preprocessor 231 may pre-process the converted 3D coordinate data to generate 3D livestock data. To this end, the preprocessor 231 may include a validity verification unit, a noise removal unit, and a factor extraction unit.

The validity verification unit determines the validity of the 3D coordinate data, and because it can obtain livestock frames that are not suitable for use due to obstruction factors (sunlight, movement of livestock, etc.), the validity is verified by checking the number of extracted points. I can judge.

The validity determination can be determined according to whether the number of extracted points is greater than or equal to a set number of points, and when the number of points is less than the predetermined number of points, the corresponding 3D coordinate data may be eliminated so that it is not used to derive the estimated weight.

The noise removal unit removes noise from the 3D coordinate data, and because objects other than livestock (ground, frames of buildings, etc.) can also be photographed, points corresponding to objects other than livestock are selected among the extracted points. Can be removed.

The factor extracting unit may generate 3D livestock data by extracting a factor from the 3D coordinate data. In this case, the 3D livestock data may include 3D coordinate data and a factor.

At this time, the factors are head direction, floor angle, focal length, height, length 1 (length01), girth_first, girth_mid, girth_end, average circumference of the body, total circumference of the body, and point cloud ( cloud point) may be one or more of the number, but is not limited thereto, and various factors such as bending or not may be extracted.

As shown in FIG. 4, length 1 is the distance from the front leg to the buttocks with the head removed, and length 2 is the distance between the forelimbs and the hind legs.

The weight extraction unit 232 may derive the estimated weight from 3D livestock data.

Here, when there are two or more three-dimensional livestock data, the weight extraction unit 232 repeats the process of deriving the estimated weight according to the number of three-dimensional livestock data to obtain an estimated weight according to each three-dimensional livestock data. Estimated weight data corresponding to the number can be obtained. In other words, the estimated weight is calculated for each 3D livestock data.

The estimated weight data may include the derived estimated weight and 3D livestock data, but is not limited thereto and may further include other information.

Referring to FIG. 5, the weight derivation unit 232 may include a candidate list unit 2320, an algorithm derivation unit 2321, and a final derivation unit 2322.

The candidate list unit 2320 generates a candidate list by selecting one or more 3D livestock data according to a factor, and determines the number of cases of condition satisfaction of the generated candidate list to derive the first weight of the 3D livestock data. I can.

To this end, the candidate list unit 2320 may include a list generation unit and a list processing unit.

The list generator may generate a candidate list by selecting one or more 3D livestock data as candidates according to factors.

Specifically, the list generation unit compares one or more three-dimensional livestock data through factors based on livestock standard data, and if it belongs to livestock standard data, selects candidate 1, candidate 2, and candidate 3 as first, second and second 3 The final first, second, and third candidate lists may be generated by adding them to the candidate list.

In this case, when the 3D livestock data is selected as a candidate, the weight according to the matched livestock standard data may be included as the weight of the 3D livestock data.

The list generator may generate a first candidate list by comparing at least one of the head direction, the floor angle, the focal length, the length 1 and the total body circumference of the 3D livestock data with the livestock standard data.

In addition, the list generator may generate a second candidate list by comparing at least one of the head direction, the total body circumference, height, length 1 and median of the 3D livestock data with the livestock standard data.

In addition, the list generator may generate a third candidate list by comparing one or more of the head direction, length 1, chest, median and rear of the 3D livestock data with livestock standard data.

Meanwhile, the list generator may also generate first, second and third candidate backup lists when generating each of the first, second and third candidate lists.

The above process will be described in more detail in the following livestock weight estimation method.

The list processing unit may determine the number of cases of condition satisfaction of the generated candidate list (first, second, and third candidate lists) and derive a first weight of the 3D livestock data according to the case.

Conditional satisfaction is whether there is one or more 3D livestock data, and as the first, second, and third candidate lists are generated, the number of cases is 3 satisfactory (first to third candidate list), and 2 satisfactory ( First and second candidate lists, second and third candidate lists, first and third candidate lists), one satisfaction (first, second, third candidate list), and zero satisfaction may occur.

The process of deriving the first weight in each of these cases will be described in more detail in the following livestock weight estimation method.

The algorithm derivation unit 2321 may derive the second weight by applying a multiple regression formula to the 3D livestock data.

Specifically, the algorithm derivation unit 2321 uses the height, length 2, the average of the trunk circumference, and the number of cloud points of the 3D livestock data, among the factors of the 3D livestock data, and uses the second weight through Equation 1 below. Can be derived.

[Equation 1]

2nd weight = m ₁ + (h ₁ × height) + (2ㅣ ₁ × length2) + (g ₁ × average body circumference) + (p ₁ × number of cloud points in 3D livestock data)

Here, m ₁ is the weight constant, h ₁ is the height constant, 2 l ₁ is the length 2 constant, g ₁ is the circumference average constant, and p ₁ is the cloud point constant of 3D livestock data.

In addition, the algorithm derivation unit 2321 may derive the second weight using Equations 2 to 4, respectively.

More specifically, the algorithm derivation unit 2321 may derive the second weight through Equation 2 using the height, length 1, the average of the trunk circumference, and the number of cloud points of the 3D livestock data among the factors. .

[Equation 2]

2nd weight = m ₂ + (h ₂ × height) + (1ㅣ ₂ × length1) + (g ₂ × average body circumference) + (p ₂ × number of cloud points in 3D livestock data)

Here, m ₂ is the weight constant, h ₂ is the height constant, 1ㅣ ₂ is the length 1 constant, g ₂ is the circumferential average constant, and p ₂ is the cloud point constant of 3D livestock data.

In addition, the algorithm derivation unit 2321 may derive the second weight through Equation 3 using the height, the average of the body circumference, and the number of cloud points of 3D livestock data among the factors.

[Equation 3]

Second weight = m ₃ + (h ₃ × height) + (g ₃ × average body circumference) + (p ₃ × number of cloud points in 3D livestock data)

Here, m ₃ is the weight constant, h ₃ is the height constant, g ₃ is the circumference average constant, and p ₃ is the cloud point constant of 3D livestock data.

In addition, the algorithm derivation unit 2321 may derive the second weight through Equation 4 using the average of the body circumference and the number of cloud points of 3D livestock data among the factors.

[Equation 4]

Second weight = m ₄ + (g ₄ × average body circumference) + (p ₄ × number of cloud points in 3D livestock data)

Here, m ₄ is the weight constant, g ₄ is the circumferential average constant, and p ₄ is the cloud point constant of 3D livestock data.

The final derivation unit 2322 may derive the estimated weight using the first weight and the second weight, and may derive the estimated weight according to an absolute value of a value obtained by subtracting the first weight from the second weight. When the absolute value is less than the predetermined weight value, the second weight is derived as the estimated weight, and when the absolute value is greater than the predetermined weight value, the estimated weight can be derived as 0 kg.

The final derivation unit 2322 may derive the estimated weight through the above process, generate the estimated weight data, provide it to the user, and store it in the data storage unit 21.

At this time, if 0kg is derived as the estimated weight, it may be regarded as a'weight measurement failure'. Accordingly, the final derivation unit 2322 notifies the user of a'weight measurement failure' so that the livestock can be retaken and reestimated or automatically By operating the 3D sensor 10, it is possible to return to the beginning and proceed again.

The management unit 24 may provide the estimated weight data stored in the data storage unit 21 in a form organized by livestock, date, and weight.

A method for estimating livestock weight using the livestock weight estimation system as described above will be described in detail below.

6 is a flowchart schematically showing a livestock weight estimation method using a livestock weight estimation system using a livestock weight estimation application based on a 3D image according to an embodiment of the present invention, and FIG. 7 is a sequential step S2 of FIG. 8 is a flowchart showing step S22 of FIG. 7, and FIG. 9 is a flowchart sequentially showing step S220 of FIG. 8, and FIG. 10 is a date when the first, second, and third candidate lists are satisfied. FIG. 11 is a flowchart showing a process in which step S22012a of FIG. 10 is performed, FIG. 12 is a flowchart showing a process in which step S22016a of FIG. 10 is performed, and FIG. 13 is a flow chart showing step S2201 of the case, and FIG. FIG. 14 is a flowchart showing a process in which step S22010b is performed in step S2201 when the list is satisfied with the condition, FIG. 14 is a flowchart showing step S2201 when the second and third candidate lists are satisfied, and FIG. 15 is S22013c in FIG. Fig. 16 is a flowchart showing a process in which step S22010d is performed in step S2201 when the first and third candidate lists are satisfied with the condition, and FIG. 17 is a flowchart showing the process of performing the step S22010d in the case where only the third candidate list is satisfied with the condition. It is a flowchart illustrating a process of performing step S22010e in step S2201 of, and FIG. 18 is a flowchart illustrating a process of performing step S22010f in step S2201 when there is no condition satisfaction candidate list.

6, a livestock weight estimation method using a livestock weight estimation system using a livestock weight estimation application based on a 3D image according to an embodiment of the present invention includes a photographing step (S1) and a weight estimation step (S2). Can include.

First, in the photographing step (S1), by photographing livestock through the terminal 1, one or more livestock frames may be extracted.

Here, the terminal 1 includes a 3D sensor 10 and a livestock weight estimation application 20.

Accordingly, in step S1, when the livestock weight estimation application 20 is executed, the livestock is photographed through the 3D sensor 10, and the screen is captured through the livestock weight estimation application 20 according to the shooting start information to extract one or more livestock frames. I can. This description has been described in detail in the above system and will be omitted.

In the weight estimation step (S2), the livestock weight estimation application 20 may derive the estimated weight using the extracted livestock frame.

Referring to FIG. 7, step S2 may include a conversion step S20, a pretreatment step S21, and a weight extraction step S22.

In the conversion step (S20), after arranging one or more livestock frames, points are extracted and converted into 3D coordinate data.

In the pre-processing step S21, the converted 3D coordinate data may be preprocessed to generate 3D livestock data.

To this end, step S22 may include a validation step, a noise removal step, and a factor extraction step.

In the validation step, the validity of each 3D coordinate data is determined, and 3D coordinate data that is not valid may be eliminated. Detailed descriptions have been made in the above system and thus will be omitted.

The noise removal step may remove noise from 3D coordinate data. Detailed descriptions have been made in the above system and thus will be omitted.

In the factor extraction step, a factor may be extracted from the 3D coordinate data to generate 3D livestock data. Detailed descriptions have been made in the above system and thus will be omitted.

The weight derivation step (S22) can derive the estimated weight from the 3D livestock data, and this step S22 It is repeated according to the number of 3D livestock data, and the estimated weight according to each 3D livestock data can be derived one by one. However, the present invention is not limited thereto and may be performed at once.

Referring to FIG. 8, step S22 may include a first weight extraction step S220, a second weight extraction step S221, and an estimated weight extraction step S222.

In the first weight extraction step (S220), a candidate list is generated by selecting one or more of the three-dimensional livestock data according to a factor, and determining the number of cases in which the conditions of the generated candidate list are satisfied, You can derive the weight.

Referring to FIG. 9, step S220 may include a list generation step (S2200) and a list processing step (S2201 ).

In the list generation step S2200, a candidate list may be generated by selecting one or more of the 3D livestock data as candidates according to a factor.

That is, in step S2200, one or more three-dimensional livestock data are compared through a factor based on livestock standard data, and if it belongs to livestock standard data, candidate 1, candidate 2, and candidate 3 are selected and the first, second and third Final first, second, and third candidate lists may be generated by adding them to the candidate list.

To this end, step S2200 may include a first candidate step, a second candidate step, and a third candidate step.

In the first candidate step, a first candidate list may be generated by comparing at least one of a head direction, a floor angle, a focal length, a length 1, and a total body circumference of each 3D livestock data with the livestock standard data. Compared with the livestock standard data, if the 3D livestock data belongs thereto, the corresponding 3D livestock data can be selected as candidate 1 and added to the first candidate list. Through this process, a first candidate list may be generated.

More specifically, the first candidate stage is whether the three-dimensional livestock data is'the same as the head direction' of the livestock standard data,'is it within the range of the floor angle','is it within the range of focal length','is it within the range of length 1', ' The comparison is performed in the order of whether it is within the total range of the trunk circumference', and only the 3D livestock data that is YES in all cases can be added to the first candidate list by proceeding to the following order only when YES (correct).

It is not limited to the above order, and may be easily changed and applied.

In addition, the first candidate step can also create a first candidate backup list. The first candidate backup list is added by adding NO (no) 3D livestock data to the first candidate backup list only in the last order, whether it is within the total body circumference range. You can create a list of candidate backups.

When added to the list, the 3D livestock data may include the weight according to the corresponding livestock standard data as the weight of the 3D livestock data.

The second candidate step is to compare one or more of the head direction, total body circumference, height, length 1, and median of each 3D livestock data with livestock standard data, and if the 3D livestock data belongs thereto, the corresponding 3D livestock data is A second candidate list may be generated by selecting as candidate 2 and adding it to the second candidate list.

More specifically, the second candidate stage is whether the 3D livestock data is'the same as the head direction','is it within the total range of the body circumference','is it within the height range','is it within the length 1 range', and ' The comparison is conducted in the order of'Is it within the median range?', and only when YES (correct) is passed to the following order, only 3D livestock data that is YES in all cases can be added to the second candidate list. It is not limited to the above order, and may be easily changed and applied.

In addition, the second candidate step can also create a second candidate backup list, and the second candidate backup by adding NO (no) 3D livestock data to the second candidate backup list only in the last order, whether it is within the median range. You can create a list.

The third candidate step is to compare one or more of the head direction, length 1, chest, median and rear of each of the 3D livestock data with livestock standard data, and if the 3D livestock data belongs thereto, the corresponding 3D livestock data is candidate. A third candidate list may be generated by selecting 3 and adding it to the third candidate list.

More specifically, the third candidate stage is whether the 3D livestock data is the same as the head direction of the livestock standard data, whether it is within the length 1 range, ``is within the chest range,'' ``is within the median range,'' and ``back range''. The comparison is performed in the order of'I am mine', and only the three-dimensional livestock data that is YES in all cases can be added to the third candidate list by proceeding to the following order only when YES (correct). It is not limited to the above order, and may be easily changed and applied.

In addition, the third candidate step can also create a third candidate backup list. The third candidate backup by adding NO (no) 3D livestock data to the third candidate backup list only in the last order,'Is it within the rear range?' You can create a list.

In the list processing step S2201, the number of cases of condition satisfaction of the generated candidate list (first, second, and third candidate lists) may be determined, and a first weight of the 3D livestock data may be derived depending on the case. .

Conditional satisfaction is whether there is one or more 3D livestock data, and as the first, second, and third candidate lists are generated, the number of cases is 3 satisfactory (first to third candidate list), and 2 satisfactory ( First and second candidate lists, second and third candidate lists, first and third candidate lists), one satisfaction (first, second, and third candidate list), and zero satisfaction may occur.

Each of the processes in which step S2201 proceeds according to the number of cases in which these conditions are satisfied will be described in detail.

Referring to FIG. 10, step S2201 includes a first number determination step (S22010a) when all of the first, second, and third candidate lists have one or more three-dimensional livestock data (if three are satisfied). I can.

In the first number determination step S22010a, it may be determined whether there is one 3D livestock data in the first candidate list. Depending on whether there is one or not, step S2201 includes a first detection step (S22011a) and a first derivation step (S22012a) or a length determination step (S22013a), a second detection step (S22014a), a temporary list generation step (S22015a), and It may include a 2 derivation step (S220216a).

First, a case where there is one 3D livestock data in the first candidate list in step S22010a will be described.

In the first detection step (S22011a), if there is one 3D livestock data in the first candidate list in step S22010a, the weight of the 3D livestock data in the first candidate list is selected as the predicted weight, and then the second and third candidate lists are Among the 3D livestock data, it is possible to find whether there are 3D livestock data having the same weight as the predicted weight.

The first derivation step (S22012a) can derive the first weight according to the result of step S22011a, that is, the first weight according to whether there is 3D livestock data having the same weight as the predicted weight in both the second and third candidate lists. Can be derived.

In step S22012a, as shown in FIG. 11, when there is 3D livestock data having the same weight as the predicted weight in both the second and third candidate lists, the first weight may be derived from Equations 5 and 6 below.

[Equation 5]

Average range = average body circumference-average body circumference of 3D livestock data in the first candidate list

Here, the average body circumference is the average body circumference of the 3D livestock data for which the estimated weight is to be derived.

[Equation 6]

1st weight = predicted weight + (average range Х 1.5)

Further, in step S22012a, when there is no 3D livestock data in which at least one of the second and third candidate lists has the same weight as the predicted weight, the predicted weight may be derived as the first weight.

Next, a case in which there is not one 3D livestock data in the first candidate list in step S22010a will be described.

In the length determination step (S22013a), if there is not one 3D livestock data in the first candidate list in step S22010a, the length 2 of the 3D livestock data for which the estimated weight is to be derived is the nth 3D livestock data of the first candidate list. It can be determined whether it is within the range of length 2.

Here, n is 2 or more, and is the number of 3D livestock data in the first candidate list, and the length determination step (S22013a), the second detection step (S22014a), and the temporary list generation step (S22015a) are repeated n times.

In step S22013a, if the length 2 of the 3D livestock data is within the range of the length 2 of the nth 3D livestock data in the first candidate list, the process proceeds to the second detection step (S22014a), and the length 2 of the 3D livestock data is determined. If it is not within the range of the length 2 of the n-th 3D livestock data of the 1 candidate list, step S22013a may be performed again through checking the number of repetitions, or step S220216a may be performed.

In the second detection step (S22014a), if it is within the range of the length 2 of the nth 3D livestock data of the first candidate list, the weight of the nth 3D livestock data of the first candidate list is selected as the predicted weight, and then the second And whether there is data having the same weight as the predicted weight among 3D livestock data in the third candidate list.

In the temporary list generation step (S22015a), according to whether both the second and third candidate lists have 3D livestock data having the same weight as the predicted weight through the result of step S22014a, n of the first candidate list is added to the first temporary list. A third 3D livestock data can be added.

In step S22015a, when there is 3D livestock data having the same weight as the predicted weight in both the second and third candidate lists, the nth 3D livestock data of the first candidate list may be added to the first temporary list.

On the other hand, if at least one of the second and third candidate lists does not have 3D livestock data having the same weight as the predicted weight, the nth 3D livestock data of the first candidate list is not added to the first temporary lits.

When the number of repetitions becomes equal to n by repeating the above step, the second derivation step S220216a may proceed.

In the second derivation step S220216a, the first weight may be derived according to the number of 3D livestock data in the generated first temporary list.

In step S220216a, as shown in FIG. 12, it is determined whether the number of 3D livestock data in the generated first temporary list is one, and if not more than one, the first weight may be derived as 0 kg.

Further, in step S220216a, when the number of 3D livestock data in the first temporary list is one or more, it may be determined whether the number of 3D livestock data in the generated first temporary list is one.

Accordingly, in step S220216a, when the number of 3D livestock data in the first temporary list is 1, the weight of the 3D livestock data in the first temporary list may be derived as the first weight, and if not 1 (means two or more) ) The average weight of the 3D livestock data in the first temporary list can be derived as the first weight.

Referring to FIG. 13, in step S2201, when both the first and second candidate lists have one or more 3D livestock data (in the case of the first and second candidate lists of two satisfactions), a third derivation step (S22010b) It may include.

In the third derivation step S22010b, the first weight may be derived according to whether the height of the 3D livestock data for which the estimated weight is to be derived is greater than the height of the first 3D livestock data of the first candidate list.

In detail, step S22010b determines whether the height of the 3D livestock data is greater than the height of the first 3D livestock data of the first candidate list, and if it is greater than the height of the first 3D livestock data of the first candidate list, the following math The first weight can be derived from Equation 7.

[Equation 7]

1st weight = list01.data(0).weight + (list01.data(0).weight × 0.03)

Here, list01.data(0).weight is the weight of the first 3D livestock data in the first candidate list.

Further, in step S22010b, when the height of the 3D livestock data is less than the height of the first 3D livestock data of the first candidate list, whether the height of the 3D livestock data is less than the height of the last 3D livestock data of the first candidate list I can judge.

Accordingly, in step S22010b, when the height is less than or equal to the height of the last 3D livestock data of the first candidate list, the first weight may be derived from Equation 8 below.

[Equation 8]

1st weight = list01.data(last).weight + (list01.data(last).weight × 0.03)

Here, list01.data (last).weight is the weight of the last 3D livestock data in the first candidate list.

Also, in step S22010b, when the height of the last 3D livestock data of the first candidate list is not less than the height, the average weight of the 3D livestock data of the first candidate list may be derived as the first weight.

Referring to FIG. 14, in step S2201, when both the second and third candidate lists have one or more 3D livestock data (in the case of the second and third candidate lists of two satisfactions), the second number determination step (S22010c) ) Can be included.

The second number determination step S22010c may determine whether there are more than one 3D livestock data in the third candidate list. Here, depending on whether there is more than one, step S2201 may include a 4-1 derivation step (S22011c) or a calculation step (S22012c) and a 4-2 derivation step (S22013c).

First, a case where the number of 3D livestock data in the third candidate list exceeds one in step S22010c will be described.

In the 4-1 derivation step (S22011c), if there are more than one 3D livestock data in the third candidate list in step S22010c, the average weight of the 3D livestock data in the second candidate list may be derived as the first weight.

Next, a case in which the 3D livestock data in the third candidate list does not exceed one (meaning one) in step S22010c will be described.

In the calculation step S22012c, the 3D livestock data of the second candidate list may be classified and calculated according to whether the weight of the k-th 3D livestock data of the second candidate list is equal to or greater than the weight of the 3D livestock data of the third candidate list. .

Here, k is 1 or more, and is the number of 3D livestock data in the second candidate list, and the calculation step S22012c may be repeated k times.

Step S22012c is to convert the 3D livestock data of the second candidate list to have a weight equal to or greater than the weight of the 3D livestock data of the third candidate list (three-dimensional livestock data greater than or equal to the weight) and have a weight less than the weight (three-dimensional livestock less than the weight Data), and the total weight (sum01, sum02), number (cnt01, cnt02), and weight average (avr01, avr02) of each of the 3D livestock data above the weight and the 3D livestock data below the weight can be obtained. .

When the number of repetitions becomes the same as k by repeating the above step, the 4-2 derivation step (S22013c) may proceed.

In the 4-2 derivation step (S22013c), the first weight may be derived according to whether the height of the 3D livestock data for which the estimated weight is to be derived is equal to or greater than the height of the 3D livestock data of the third candidate list.

Specifically, in step S22013c, when the height of the 3D livestock data is greater than or equal to the height of the 3D livestock data of the third candidate list, it may be determined whether sum01 is greater than 0.

This is to prevent a case in which the first weight cannot be obtained by checking for an error before deriving the first weight because there may be a case where sum01 is 0 because there is no 3D livestock data that is more than the weight.

Accordingly, in step S22013c, when sum01 is greater than 0, the first weight may be derived through Equation 9 below.

[Equation 9]

1st weight = sum01/cnt01

Here, sum01 is the total weight of 3D livestock data that is more than the weight, and cnt01 is the number of 3D livestock data that is more than the weight.

In addition, when sum01 is not greater than 0, the first weight may be derived through Equation 10 below.

[Equation 10]

1st weight = sum02/cnt02

Here, sum02 is the total weight of 3D livestock data less than the weight, and cnt02 is the number of 3D livestock data less than the weight.

On the other hand, in step S22013c, when the height of the 3D livestock data is less than the height of the 3D livestock data of the third candidate list, it may be determined whether sum02 is greater than 0.

As above, there may be cases where sum02 is 0 because there is no 3D livestock data less than the weight, so the error is checked before deriving the first weight.

Accordingly, in step S22013c, when sum02 is greater than 0, the first weight is derived through Equation 10, and when sum02 is not greater than 0, the first weight may be derived through Equation 8.

Through this, it is possible to improve estimation efficiency by preventing 0 from appearing as the value of the first weight.

Referring to FIG. 16, in step S2201, when both the first and third candidate lists have one or more 3D livestock data (for the first and third candidate lists of two satisfactions), the fifth derivation step (S22010d) It may include.

In the fifth derivation step (S22010d), the first weight may be derived according to whether the height of the 3D livestock data for which the estimated weight is to be derived is equal to or greater than the height of the first 3D livestock data of the third candidate list.

In detail, in step S22010d, it is determined whether the height of the 3D livestock data is greater than or equal to the height of the first 3D livestock data of the third candidate list, and if the height is greater than or equal to the height of the first 3D livestock data of the third candidate list, the following equation The first weight can be derived from 11.

[Equation 11]

1st weight = list03.data(0).weight + (list03.data(0).weight × 0.03)

Here, list03.data(0).weight is the weight of the first 3D livestock data in the third candidate list.

Further, in step S22010d, when the height of the 3D livestock data is less than the height of the first 3D livestock data of the third candidate list, whether the height of the 3D livestock data is less than the height of the last 3D livestock data of the third candidate list I can judge.

Accordingly, in step S22010d, when the height is less than or equal to the height of the last 3D livestock data of the third candidate list, the first weight may be derived from Equation 12 below.

[Equation 12]

1st weight = list03.data(last).weight + (list03.data(last).weight × 0.03)

Here, list03.data (last).weight is the weight of the last 3D livestock data in the third candidate list.

In addition, in step S22010d, when the height of the last 3D livestock data of the third candidate list is not less than the height, the average weight of the 3D livestock data of the third candidate list may be derived as the first weight.

Step S2201 may include a sixth derivation step when only the first candidate list has one or more 3D livestock data (the first candidate list having one satisfaction).

In the sixth derivation step, the weight of the first 3D livestock data in the first candidate list may be derived as the first weight.

Step S2201 may include a seventh derivation step when only the second candidate list has one or more 3D livestock data (a second candidate list having one satisfaction).

In the seventh derivation step, the average weight of the 3D livestock data in the second candidate list may be derived as the first weight.

Referring to FIG. 17, step S2201 may include an eighth derivation step (S22010e) when only the third candidate list has one or more 3D livestock data (a third candidate list having one satisfaction).

In the eighth derivation step (S22010e), the first weight for deriving the first weight may be derived according to whether there is one 3D livestock data in the third candidate list.

First, in step S22010e, it is determined whether there is one 3D livestock data of the third candidate list, and if there is one 3D livestock data of the third candidate list, the 3D livestock data of the third candidate list is derived as the first weight. I can.

On the other hand, in step S22010e, when there is not one 3D livestock data in the third candidate list, the first weight may be derived by determining whether the height of the 3D livestock data is greater than or equal to the height of the first 3D livestock data in the third candidate list. .

Further, in step S22010e, when the height of the 3D livestock data is greater than or equal to the height of the first 3D livestock data in the third candidate list, the first weight may be derived from Equation 11.

In step S22010e, if the height of the first 3D livestock data in the third candidate list is less than the height of the first 3D livestock data, then it is determined whether the height of the 3D livestock data is less than the height of the last 3D livestock data in the third candidate list to derive the first weight. can do.

In step S22010e, when the height of the dimensional livestock data is less than the height of the last 3D livestock data of the third candidate list, the first weight may be derived from Equation 12, and the height of the last 3D livestock data of the third candidate list If it is exceeded, the average weight of the 3D livestock data in the third candidate list may be derived as the first weight.

Referring to FIG. 18, step S2201 may include a ninth derivation step (S22010f) when none of the first to third candidate lists have 3D livestock data (0 is satisfied).

In a ninth derivation step (S22010f), it may be determined whether the number of 3D livestock data in the second candidate backup list is greater than 0.

If the number of 3D livestock data in the second candidate backup list is greater than 0, S22010f is determined according to whether the average of the body circumference of the 3D livestock data is within the average range of the body circumference of the j-th 3D livestock data of the second candidate backup list. The 3D livestock data of the second candidate backup list may be classified and calculated.

Here, j is 1 or more, and is the number of 3D livestock data in the second candidate backup list.

In the above process, S22010f converts the 3D livestock data of the second candidate backup list to the body circumference average of the 3D livestock data within the range (3D livestock data within the range) and the body circumference average not within the range (outside range) 3D livestock data), and the total weight (sum03, sum04), number (cnt03, cnt04) of each can be calculated.

Then, S22010f can determine whether sum03 is greater than 0, and when sum03 is greater than 0, the first weight is derived by Equation 13 below, and when sum03 is less than 0, the first weight can be derived by Equation 14 below. have.

[Equation 13]

1st weight = sum03/cnt03

Here, sum03 is the total weight of 3D livestock data within the range, and cnt03 is the number of 3D livestock data within the range.

[Equation 14]

1st weight = sum04/cnt04

Here, sum04 is the total weight of 3D livestock data outside the range, and cnt04 is the number of 3D livestock data outside the range.

On the other hand, S22010f may derive the first weight as 100kg when the number of 3D livestock data in the second candidate backup list is less than 0. This may mean'weight measurement failure'.

The second weight extraction step (S221) may derive the second weight by applying one of Equations 1 to 4, which are multiple regression formulas, to the dimensional livestock data, and it is preferable to apply Equation 1, but is not limited thereto. Does not. Detailed descriptions have been made in the description of the system, and thus will be omitted.

In the step of deriving the estimated weight (S222), the estimated weight may be derived using the derived first weight and the second weight, and the estimated weight data may be generated and stored. Detailed descriptions have been made in the description of the system, and thus will be omitted.

In addition, the method for estimating livestock weight according to an embodiment of the present invention may further include a biometric information input step, and the biometric information input step may receive and store biometric information of livestock from a user through the livestock weight estimation application 20.

The biometric information input step may be performed before step S1, but is not limited, such as performed after step S2, and may be located in various orders.

As described above, a livestock weight estimation system using a livestock weight estimation application based on a 3D image according to an embodiment of the present invention and a livestock weight estimation method using the same are performed using a plurality of frames obtained from a 3D sensor. By estimating the weight of livestock through the candidate list method and the multiple regression method, the accuracy and reliability of weight estimation can be excellent.

Therefore, additional equipment is not required to measure the weight of livestock, and breeding costs can be reduced through feed control through continuous weight management of livestock, and the time of shipment can be accurately predicted, thereby increasing the profits of farmers.

In addition, there is no hassle of inducing livestock to be congested for a certain period of time in order to measure body weight, so it is possible to solve problems caused by a lack of manpower of farmers, an aging manpower, and an increase in scale.

In addition, it can be applied not only to pigs, but also to various livestock such as chickens and cows, so its utility is expected to expand.

1: terminal
10: 3D sensor
20: Livestock weight estimation application
21: data storage unit
22: sensor imaging unit
23: weight estimation
230: conversion unit
231: pretreatment unit
232: weight extraction unit
2320: candidate list unit
2321: algorithm derivation unit
2322: final derivation
24: management

Claims (21)

In a livestock weight estimation system using a livestock weight estimation application based on 3D images,
Including a terminal including a livestock weight estimation application for estimating the weight of livestock using a 3D sensor for photographing livestock and livestock frame extracted through the 3D sensor,
The livestock weight estimation application,
A sensor photographing unit for extracting one or more livestock frames by photographing livestock by operating the 3D sensor, and
It includes a weight estimation unit that derives the estimated weight using the extracted livestock frame,
The weight estimation unit,
A conversion unit converting the extracted livestock frame into 3D coordinate data;
A preprocessing unit that preprocesses the converted 3D coordinate data to generate 3D livestock data, and
Including a weight extraction unit for deriving the estimated weight from the three-dimensional livestock data,
The weight extraction unit,
A candidate list unit for generating a candidate list by selecting one or more of the 3D livestock data according to a factor, determining the number of cases of condition satisfaction of the generated candidate list, and derives a first weight of the 3D livestock data;
An algorithm derivation unit for deriving a second weight by applying a multiple regression formula to the 3D livestock data, and
Including a final derivation unit for deriving an estimated weight using the first weight and the second weight,
The candidate list unit,
A list generator for generating a candidate list by selecting one or more of the 3D livestock data as candidates according to a factor; and
And a list processing unit that determines the number of cases of condition satisfaction of the generated candidate list, and derives a first weight of the 3D livestock data according to the case,
The list generation unit,
Create a first candidate list by comparing at least one of the total head direction, floor angle, focal length, length 1 and body circumference of 3D livestock data with livestock standard data,
Comparing at least one of the head direction, total body circumference, height, length 1, and median of the 3D livestock data with livestock standard data to create a second candidate list,
Livestock weight estimation system, characterized in that for generating a third candidate list by comparing at least one of the head direction, length 1, chest, median and rear of the 3D livestock data with livestock standard data.
delete The method of claim 1,
The pretreatment unit,
Livestock weight estimation system including a factor extracting unit for extracting a factor from the 3D coordinate data.
The method of claim 3,
The factor is,
Livestock weight estimation system including one or more of head direction, floor angle, focal length, height, length1, chest, median, posterior, average trunk circumference, total trunk circumference, and number of cloud points.
(Here, length 1 is the distance from the front leg to the buttocks with the head removed, and length 2 is the distance between the front and hind legs.)
delete delete The method of claim 1,
The algorithm derivation unit,
A livestock weight estimation system, characterized in that the second weight is derived through Equation 1 below by using a height, length 2, an average of a trunk circumference, and a number of points among the factors of the 3D livestock data.
[Equation 1]
2nd weight = m ₁ + (h ₁ × height) + (2ㅣ ₁ × length2) + (g ₁ × average of trunk circumference) + (p ₁ × number of cloud points in 3D livestock data)
(Where m ₁ is the weight constant, h ₁ is the height constant, 2ㅣ ₁ is the length 2 constant, g ₁ is the circumferential average constant, and p ₁ is the cloud point constant of 3D livestock data)
In a livestock weight estimation method using a livestock weight estimation system using a livestock weight estimation application based on 3D images,
(a) Shooting to extract one or more livestock frames by photographing livestock through a terminal including a 3D sensor for photographing livestock and a livestock weight estimation application that estimates the weight of livestock using livestock frames extracted through the 3D sensor Step and
(b) including a weight estimation step of deriving the estimated weight using the extracted livestock frame,
The step (b),
A transformation step of converting the extracted livestock frame into 3D coordinate data;
Pre-processing step of pre-processing the converted 3D coordinate data to generate 3D livestock data, and
Including a weight derivation step of deriving the estimated weight from the three-dimensional livestock data,
The weight extraction step,
A first weight extraction step of generating a candidate list by selecting one or more three-dimensional livestock data according to a factor, determining the number of cases in which the conditions of the generated candidate list are satisfied, and deriving the first weight of the three-dimensional livestock data ;
A second weight derivation step of deriving a second weight by applying a multiple regression formula to the 3D livestock data, and
Including an estimated weight derivation step of deriving an estimated weight using the first weight and the second weight,
The first weight extraction step,
A list generation step of generating a candidate list by selecting one or more of the three-dimensional livestock data as candidates according to a factor, and
A list processing step of determining the number of cases of condition satisfaction of the generated candidate list, and deriving a first weight of the 3D livestock data according to the case,
The list generation step,
A first candidate step of generating a first candidate list by comparing at least one of a head direction, a floor angle, a focal length, a length 1, and a total body circumference of each of the three-dimensional livestock data with livestock standard data;
A second candidate step of generating a second candidate list by comparing at least one of the head direction, total body circumference, height, length 1 and median of each of the three-dimensional livestock data with livestock standard data, and
And a third candidate step of generating a third candidate list by comparing one or more of the head direction, length 1, chest, median and rear of each of the three-dimensional livestock data with livestock standard data.
delete delete delete delete The method of claim 8,
The list processing step,
When the first, second and third candidate lists all have one or more 3D livestock data,
A first number determination step of determining whether there is one 3D livestock data in the first candidate list;
If there is one 3D livestock data of the first candidate list, the weight of the 3D livestock data of the first candidate list is selected as a predicted weight, and then the predicted weight of the 3D livestock data of the second and third candidate lists is selected. A first detection step to find whether there is data having the same weight as and
And a first derivation step of deriving a first weight according to whether both the second and third candidate lists have 3D livestock data having the same weight as the predicted weight.
The method of claim 13,
The list processing step,
In the first number determination step, if there is not one 3D livestock data of the first candidate list, the length 2 of the 3D livestock data to be derived from the estimated weight is the nth 3D livestock data of the first candidate list. A length determination step of determining whether it is within the range of length 2;
If it is within the range of the length 2 of the nth 3D livestock data of the first candidate list, after selecting the weight of the nth 3D livestock data of the first candidate list as a predicted weight, the second and third candidate lists A second detection step of finding whether there is data having the same weight as the predicted weight among the three-dimensional livestock data of;
Depending on whether both the second and third candidate lists have 3D livestock data having the same weight as the predicted weight, a temporary list is generated to add the nth 3D livestock data of the first candidate list to a first temporary list Step and
Including a second derivation step of deriving the first weight according to the number of three-dimensional livestock data in the generated first temporary list,
The length determination step, the second detection step, and the temporary list generation step,
Livestock weight estimation method, characterized in that made by repeating n times.
(Where, n is 2 or more, which is the number of 3D livestock data in the first candidate list)
The method of claim 8,
The list processing step,
When only the first and second candidate lists have one or more 3D livestock data,
Livestock weight estimation method comprising a third derivation step of deriving the first weight according to whether the height of the 3D livestock data to be derived from the estimated weight is greater than the height of the first 3D livestock data in the first candidate list .
The method of claim 8,
The list processing step,
When only the second and third candidate lists have more than one 3D livestock data,
A second number determination step of determining whether there are more than one 3D livestock data in the third candidate list, and
And a 4-1 derivation step of deriving an average weight of the 3D livestock data of the second candidate list as a first weight if there are more than one 3D livestock data of the third candidate list.
The method of claim 16,
The list processing step,
When the number of 3D livestock data in the third candidate list does not exceed one in the second number determination step,
A calculation step of classifying and calculating the 3D livestock data of the second candidate list according to whether the weight of the k-th 3D livestock data of the second candidate list is equal to or greater than the weight of the 3D livestock data of the third candidate list; and
A 4-2 derivation step of deriving the first weight according to whether the height of the 3D livestock data to be derived from the estimated weight is greater than or equal to the height of the 3D livestock data of the third candidate list,
The calculation step,
Livestock weight estimation method, characterized in that made by repeating k times.
(Here, k is 1 or more, and is the number of 3D livestock data in the second candidate list)
The method of claim 8,
The list processing step,
When only the first and third candidate lists have more than one 3D livestock data,
Livestock weight estimation method including a fifth derivation step of deriving the first weight according to whether the height of the 3D livestock data to be derived from the estimated weight is equal to or greater than the height of the first 3D livestock data of the third candidate list .
The method of claim 8,
The list processing step,
When only the first candidate list has one or more 3D livestock data,
And a sixth derivation step of deriving the weight of the first 3D livestock data in the first candidate list as a first weight.
The method of claim 8,
The list processing step,
If only the second candidate list has more than one 3D livestock data,
And a seventh derivation step of deriving an average weight of the 3D livestock data of the second candidate list as a first weight.
The method of claim 8,
The list processing step,
If only the third candidate list has more than one 3D livestock data,
And an eighth derivation step of deriving a first weight for deriving a first weight according to whether there is one 3D livestock data in the third candidate list.

Images