Livestock weight estimation system

The livestock weight estimation system uses image analysis and deep learning to accurately determine the weight of pigs, addressing the inefficiencies of existing technologies by providing a compact, cost-effective solution that reduces animal stress and improves shipment quality.

WO2024228595A1PCT designated stage expired Publication Date: 2024-11-07CJ FEED&CARE CORP
View PDF 5 Cites 0 Cited by

Patent Information

Application Number
PCT/KR2024/006051
Authority / Receiving Office
WO · WO
Patent Type
Applications
Current Assignee / Owner
Priority Date
2023-12-26
Filing Date
2024-05-03
Publication Date
2024-11-07

AI Technical Summary

Technical Problem

Existing livestock weight estimation technologies lack accuracy and are cumbersome, leading to inefficiencies in selecting and shipping pigs of appropriate weight, with existing sorting devices being large, costly, and stressful for the animals.

Method used

A system utilizing image information from a vision inspection device with deep learning algorithms to estimate livestock weight, including a passage device with a vision inspection system that captures images and calculates weight, and a sorting device that uses a biological appearance estimation module to determine suitable shipment weight, reducing stress and installation costs.

Benefits of technology

The system achieves high accuracy in weight estimation and selection, improving shipment quality and reducing stress on animals and installation costs, while being compact enough for efficient use in limited spaces.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure KR2024006051_07112024_PF_FP_ABST
    Figure KR2024006051_07112024_PF_FP_ABST
Patent Text Reader

Abstract

The present invention relates to a livestock weight estimation system, and more particularly, to a system capable of estimating the weight of livestock in a barn with high accuracy by using image information about the livestock. According to the livestock weight estimation system according to an embodiment of the present invention, individual livestock in a barn can be accurately recognized and the weight of the individual livestock can be accurately predicted. In addition, image information about livestock in a breeding space can be acquired effectively and efficiently via the configuration of an image acquisition module.
Need to check novelty before this filing date? Find Prior Art

Description

Livestock weight estimation system

[0001] The present invention relates to a livestock weight estimation system, and more specifically, to a system capable of estimating the weight of livestock with high accuracy using image information of livestock within a livestock barn.

[0002] The biometric monitoring and management systems market is growing at a rate of approximately 25% annually. For example, the "Animal health monitoring and comfort" market is projected to grow approximately 140% from 2016 to 2021.

[0003] These livestock monitoring technologies are being developed to predict animal weight, track and manage animals, and manage populations through behavioral and audio data, individual animal tracking, and detection of behavioral patterns and specific behaviors. Accordingly, the trend is shifting from traditional field experience-based management to objective management through data analysis.

[0004] For example, when shipping pigs raised in pig pens, it is very important to select and ship pigs with an appropriate weight, but existing technologies have the problem of insufficient accuracy in selecting and shipping these pigs.

[0005] In addition, the weight and volume of the sorting equipment for sorting out shipped pigs were large, so installation and removal costs were high and there was also a problem of spatial disadvantage.

[0006] The present invention has been devised to solve the above-mentioned problem, and relates to a system capable of estimating the weight of livestock with high accuracy using image information of livestock in a livestock barn.

[0007] According to one embodiment of the present invention, a pig sorting device for shipping includes: a passage device; wherein the passage device includes: a passage frame including a passage space on the inside, and a passage inlet and a passage outlet that open the passage space to the outside; a vision inspection device installed in the passage frame and configured to measure the weight of an object that has entered the passage space and generate a signal; and a sorting display device installed in the passage frame and operated by a signal transmitted from the vision inspection device; wherein the vision inspection device generates an operating signal for operating the sorting display device when the weight of the object that has entered the passage space satisfies a predetermined standard.

[0008] In one embodiment, the vision inspection device includes a vertical bar connected to the passage frame, and an optical device connected to the vertical bar and visually capturing the interior of the passage space.

[0009] In one embodiment, the vision inspection device includes a computing device that receives a signal provided from the optical device, and the computing device is equipped with an algorithm that calculates the weight of an object using visual information provided from the optical device.

[0010] In one embodiment, the passage device further includes a tag device that recognizes a tag provided on an object.

[0011] In one embodiment, the reference weight is a weight at which an object that has entered the passage space is judged to be suitable for shipment, and the object's weight is 110 to 130 kg.

[0012] In one embodiment, the passage frame has a plurality of iron bars forming the side portions and has an open bottom structure.

[0013] According to one embodiment, the apparatus further comprises a classification device connected to the passage device, wherein the classification device comprises a classification frame including a classification space on the inside, and a classification inlet and a classification outlet that open the classification space to the outside; an opening / closing door that opens / closes the classification outlet; and a door operating device that operates the opening / closing door; wherein the classification inlet is connected to the passage exit.

[0014] In one embodiment, the classification outlet includes a first classification outlet and a second classification outlet, each provided on at least one side of the classification frame, and the opening / closing door selectively opens and closes the first classification outlet and the second classification outlet.

[0015] In one embodiment, the opening / closing door is a rotating door rotatable about a door axis, the opening / closing door is located within the sorting space, the first sorting exit and the second sorting exit are respectively disposed on one side and the other side of the sorting frame, and when the opening / closing door rotates in the direction of the first sorting exit, the first sorting exit is closed and the second sorting exit and the sorting inlet are communicated, and when the opening / closing door rotates in the direction of the second sorting exit, the second sorting exit is closed and the first sorting exit and the sorting inlet are communicated.

[0016] In one embodiment, the vision inspection device provides an operating signal for operating the door operating device, and the door operating device is operated by the operating signal provided from the vision inspection device.

[0017] In one embodiment, the classification frame has a plurality of iron bars forming a side portion and has an open bottom structure.

[0018] According to one embodiment, the present invention comprises a livestock appearance estimation module for deriving appearance information of livestock using image information acquired from the vision inspection device; and a weight estimation module for deriving the weight of the livestock using the livestock appearance information derived from the livestock appearance estimation module; wherein the livestock appearance estimation module comprises: an individual recognition unit for recognizing an individual livestock; a first key point detection unit for deriving a specific body point of the individual recognized by the individual recognition unit; a second key point detection unit for deriving an appearance characteristic of the individual according to a posture using the first key point information derived from the first key point detection unit; an integrated 3D model construction unit for deriving a 3D model by merging information derived from the first key point detection unit and the second key point detection unit with an RGB image, depth information, a near-infrared image, a short-wave infrared image, and a thermal image of the livestock acquired from the movable sensor; and a live appearance estimation unit for deriving the appearance information of the individual from the 3D model and the individual appearance characteristics of the individual according to the posture; and the weight estimation module derives the weight of the individual based on the appearance of the individual derived from the livestock appearance estimation unit.

[0019] In one embodiment, the object recognition unit recognizes an object by comparing a learning data set consisting of RGB images with image information.

[0020]

[0021] According to the livestock weight estimation system according to an embodiment of the present invention, livestock individuals within a livestock shed can be accurately recognized and the individual weights can be accurately predicted.

[0022] In addition, through the configuration of the image acquisition module, image information of livestock within the breeding space can be effectively and efficiently acquired.

[0023] In addition, it includes an object recognition unit, first and second keypoint detection units, an integrated 3D model construction unit, a living body appearance estimation unit, and a weight estimation module, each of which includes deep learning and 3D image analysis algorithms, thereby increasing the accuracy of weight derivation of the object. Accordingly, for example, when selecting pigs for shipment, weight can be derived with high accuracy, thereby improving the quality of shipment.

[0024] In addition, since it is possible to track and monitor the growth of the individual, abnormal growth and signs of each individual can be detected, and early measures can be taken accordingly.

[0025] Figures 1 and 2 are drawings showing the structure of a livestock weight estimation system according to an embodiment of the present invention.

[0026] Figure 3 is a drawing showing the structure and operation example of the image acquisition module.

[0027] Figure 4 is a conceptual diagram showing the structure of a weight estimation module and an image acquisition module.

[0028] Figure 5 is an example diagram showing an individual livestock being recognized by an individual recognition unit.

[0029] Figure 6 shows an example of the object recognition process of the object recognition unit.

[0030] Figures 7 and 8 show that key points of a specific livestock object are detected by a key point detection unit.

[0031] Figures 9 and 10 are exemplary diagrams showing that key points, body appearance, and body weight of a specific livestock individual are derived through a key point detection unit, a body appearance estimation unit, and a body weight estimation unit.

[0032] Fig. 11 is a drawing showing a pig sorting device according to one embodiment of the present invention.

[0033] Figure 12 is a drawing showing the shipment money sorting device of Figure 11 as seen from above.

[0034] Figure 13 is a drawing showing the pig sorting device of Figure 11 being installed and used in a pigsty.

[0035] Fig. 14 is a drawing showing a pig sorting device according to one embodiment of the present invention.

[0036] Figure 15 is a drawing showing the shipment money sorting device of Figure 14 as seen from above.

[0037] Figures 16 to 20 are drawings showing the usage form of the shipping money sorting device of Figure 4 in sequence.

[0038] Figure 21 is a drawing showing the pig sorting device of Figure 1 installed and used in a pigsty.

[0039] FIG. 22 is a block diagram of an estimation device and a vision inspection device according to embodiments of the present disclosure.

[0040] The configuration and operation of the present invention will be described in detail with reference to embodiments of the present invention illustrated in the attached drawings below.

[0041] The present invention is capable of various modifications and embodiments. Specific embodiments are illustrated in the drawings and described in detail in the detailed description. The effects and features of the present invention, as well as the methods for achieving them, will become clearer with reference to the embodiments described in detail below, along with the drawings. However, the present invention is not limited to the embodiments disclosed below and can be implemented in various forms.

[0042] Hereinafter, embodiments of the present invention will be described in detail with reference to the attached drawings. When describing with reference to the drawings, identical or corresponding components are given the same drawing reference numerals, and redundant descriptions thereof will be omitted.

[0043] Hereinafter, the term “upper” or “upper” may include not only things that are directly above in contact, but also things that are above in a non-contact manner.

[0044] In the following examples, the terms first, second, etc. are not used in a limiting sense, but are used for the purpose of distinguishing one component from another.

[0045] In the examples below, singular expressions include plural expressions unless the context clearly indicates otherwise.

[0046] In the following examples, terms such as “include” or “have” mean that a feature or component described in the specification is present, and do not preclude the possibility that one or more other features or components may be added.

[0047] For convenience of explanation, the sizes of components in the drawings may be exaggerated or reduced. For example, the sizes and thicknesses of each component shown in the drawings are arbitrarily indicated for convenience of explanation, and thus the present invention is not necessarily limited to what is shown.

[0048] Additionally, terms such as “…part”, “…area”, etc. described in this specification may mean a unit that processes at least one function or operation.

[0049] In an embodiment of the present disclosure, the object may be a domesticated livestock such as a cow, a horse, a sheep, a goat, a pig, a deer, a chicken, a duck, a goose, etc. Preferably, the object may be a pig.

[0050] Figures 1 and 2 are drawings showing the structure of a livestock weight estimation system according to an embodiment of the present invention.

[0051] A livestock weight estimation system according to an embodiment of the present invention may include an image acquisition module (200) installed in a livestock shed (100) and acquiring image information within the livestock shed (100); and an estimation device (300) that calculates a weight value of the livestock using the acquired image information.

[0052] <Congratulatory Speech (100)>

[0053] First, the livestock shed (100) in which livestock are raised and an image acquisition module (200) is installed may be described as follows.

[0054] The livestock shed (100) may be, for example, a pigsty where pigs are raised, but is not necessarily limited thereto.

[0055] The livestock shed (100) includes a plurality of breeding spaces (110). The breeding spaces (110) may have an area of ​​a size suitable for the growth of livestock (K) growing within the livestock shed (100). The breeding spaces (110) may be partitioned by fences, railings, etc.

[0056] The above-described plurality of breeding spaces (110) may have a predetermined arrangement. According to an embodiment, the above-described plurality of breeding spaces (110) may be arranged in parallel along the first direction (L).

[0057] In addition, according to an embodiment, the livestock house (100) may include a first breeding line (A) composed of a plurality of breeding spaces (110) and a second breeding line (B). The first breeding line (A) is a set of predetermined breeding spaces (110) in which the plurality of breeding spaces (110) are arranged along the first direction (L). The second breeding line (B) is also a set of predetermined breeding spaces (110) in which the plurality of breeding spaces (110) are arranged along the first direction (L).

[0058] The first breeding line (A) and the second breeding line (B) are separated by a space line (120). The space line (120) may be a designated passage through which workers and others can pass and work.

[0059] <Image acquisition module (200)>

[0060] Figure 3 is a drawing showing the structure and operation example of the image acquisition module (200).

[0061] The image acquisition module (200) is a device installed in the livestock house (100) to acquire image information within the livestock house (100). The image acquisition module (200) may be a predetermined piece of hardware. The image acquisition module (200) is a device including one or more processors, a computer-readable recording medium, and a memory, and may include a guide rail (210) and a movable sensor (220).

[0062] The above image acquisition module (200) may include a guide rail (210) installed in a livestock house (100) and crossing the livestock house (100) in at least one direction, and a movable sensor (220) that moves along the guide rail (210). The image acquisition module (200) may control the guide rail (210) and the movable sensor (220) by having a processor execute commands stored in a recording medium.

[0063] The guide rail (210) may be a member such as a guide bar that extends in a predetermined direction to guide the displacement of the movable sensor (220). The specific configuration of the guide rail (210) is not limited. The guide rail (210) may be installed on the ceiling of the livestock house (100). However, this is not limited thereto.

[0064] At least a portion of the above guide rail (210) may extend in a direction parallel to the first direction (L). That is, the guide rail (210) may extend in a direction parallel to the direction in which a plurality of breeding spaces (110) are arranged.

[0065] The above-mentioned movable sensor (220) can execute a function capable of acquiring image information.

[0066] The above-described movable sensor (220) is connected to the guide rail (210) and can be displaced along the guide rail (210). Accordingly, the movable sensor (220) may include a sensor body and a guide body that connects the sensor body to the guide rail (210).

[0067] The mobile sensor (220) may include at least one of a LiDAR (Light Detection And Ranging) sensor, a depth sensor, a general image (RGB) sensor, a near-infrared-shortwave infrared spectral sensor, and a thermal imaging sensor.

[0068] Meanwhile, the mobile sensor (220) may have a configuration in which heterogeneous sensors are combined. However, this is not limited thereto. That is, the mobile sensor (220) may apply any number of multiple sensors depending on the on-site environment, research purpose, and business direction of the livestock farm (100).

[0069] The above-mentioned mobile sensor (220) can collect information such as the length and depth of the livestock to obtain three-dimensional image information and / or temperature information.

[0070] The above-mentioned mobile sensor (220) can detect at least one of the distance, direction, speed, temperature, material distribution and / or concentration characteristics to the target by shining a laser light source on the target livestock.

[0071] The above-described movable sensor (220) can be connected to the estimation device (300) wirelessly or by wire and can be executed. Accordingly, data including 3D image information and / or temperature information acquired from the movable sensor (220) can be transmitted to the estimation device (300). Meanwhile, the movable sensor (220) can be turned ON / OFF or displaced along the guide rail (210) by receiving an operation signal from the estimation device (300), a predetermined control module, or a predetermined input module. Accordingly, the image acquisition module (200) may further include a signal exchange device or a communication interface that transmits the acquired information to the estimation device (300).

[0072] The above three-dimensional image information may include two-dimensional images and depth information. In addition, the three-dimensional image information may include three-dimensional image information obtained by photographing at least one of the front, left side, right side, back, or top of the livestock, but is not limited thereto and may be three-dimensional image information obtained from various directions.

[0073] By providing a mobile sensor (220) with the above configuration, various digital biometric data suitable for the field environment can be detected, sensed, and acquired. Furthermore, even when the mobile sensor (220) has a configuration in which heterogeneous sensors are combined, the acquired information can be identical 3D biometric image information. However, this is not limited thereto.

[0074] In addition, the movable sensor (220) can be displaced in the direction in which the breeding space (110) is arranged along the direction in which the guide rail (210) extends, thereby detecting image information of the livestock (K) in the breeding space (110) from the front or diagonally upward direction. The movable sensor (220) can be moved along the guide rail (210) according to a control signal. The movable sensor (220) can be changed in the direction in which the breeding space (110) is located according to the control signal. The movable sensor (220) can perform a focus adjustment function according to the distance from the breeding space (110) according to the control signal. The focus adjustment function can be automatically executed to sense and detect one or more objects in the breeding space (110), respectively. The control signal may be a signal generated by the estimation device (300).

[0075] According to one embodiment, as illustrated in FIG. 1, the guide rail (210) is installed on the space line (120), and the breeding space (110) includes a first breeding line (A) and a second breeding line (B) that are partitioned by the space line (120), so that image information of livestock in the first breeding line (A) and livestock in the second breeding line (B) can be acquired using one movable sensor (220). The image acquisition module (200) can acquire one or more image information of one or more objects in the breeding space (110) by using the movable sensor (220) that moves the guide rail (210).

[0076] Meanwhile, according to one embodiment, as illustrated in FIG. 2, at least a portion of the guide rail (210) is positioned on the breeding space (110) and may extend in a direction crossing the plurality of breeding spaces (110). Therefore, according to one embodiment, the movable sensor (220) may acquire image information over the plurality of breeding spaces (110).

[0077] As illustrated in FIG. 4, the image acquisition module (200) may be a device including a processor (v), a communication interface (200b), and a memory (200c). The image acquisition module (200) may be implemented with more components than those illustrated in FIG. 4, or with fewer components than those illustrated in FIG. 4. The image acquisition module (200) may be a user terminal such as a laptop, desktop, smartphone, PDA, or the like, may be a server, may be a network system, or may be a separate device.

[0078] According to one embodiment of the present disclosure, the processor (200a) typically controls the overall operation of the image acquisition module (200). For example, the processor (200a) may control the overall components included in the image acquisition module (200) by executing a program stored in the image acquisition module (200). The processor (200a) may capture images, videos, etc. by a control signal from the estimation device (300) and transmit the captured images, videos, etc. to the estimation device (300). The processor (200a) may communicate with the estimation device through a communication interface (200b). The processor (200a) may store captured image information in a memory (200c).

[0079] The processor (200a) is a configuration for controlling the overall operation of the image acquisition module (200). Specifically, the processor (200a) controls the overall operation of the image acquisition module (200) using various programs stored in the storage medium of the image acquisition module (200). For example, the processor (200a) may include a CPU, RAM, ROM, and a system bus. Here, the ROM is a configuration in which a command set for system booting is stored, and the CPU copies the stored operating system of the image acquisition module (200) to RAM according to the commands stored in the ROM and executes the O / S to boot the system. When the system booting is complete, the CPU can copy various stored applications to RAM and execute them to perform various operations. Although the image acquisition module (200) has been described as including only one CPU in the above, it may be implemented with multiple CPUs (or DSPs, SoCs, etc.) during implementation.

[0080] According to one embodiment of the present invention, the processor (200a) may be implemented as a digital signal processor (DSP) for processing digital signals, a microprocessor, or a time controller (TCON). However, the present invention is not limited thereto, and may include one or more of a central processing unit (CPU), a micro controller unit (MCU), a micro processing unit (MPU), a controller, an application processor (AP), a communication processor (CP), or an ARM processor, or may be defined by the corresponding terms. In addition, the processor (200a) may be implemented as a system on chip (SoC) having a built-in processing algorithm, a large scale integration (LSI), or may be implemented in the form of a field programmable gate array (FPGA).

[0081] According to one embodiment of the present disclosure, the communication interface (200b) creates a channel for communicating with an external device and enables data transmission and reception with the external device. The communication interface (200b) may include at least one of a Bluetooth communication interface, a BLE (Bluetooth Low Energy) communication interface, a near field communication interface, a WLAN (Wi-Fi) communication interface, a Zigbee communication interface, an infrared (IrDA, infrared Data Association) communication interface, a WFD (Wi-Fi Direct) communication interface, an UWB (ultra wideband) communication interface, an Ant+ communication interface, and a mobile communication network.

[0082] The estimation device (300) may be a device that outputs a result value for estimating the appearance of livestock using acquired image information. The image information may be acquired through an image acquisition module (200) or a vision inspection device (15). The estimation device (300) may be a computing device including one or more processors (310), a computer-readable recording medium, a memory, and a communication interface (320). The estimation device (300) may further include an input device, an output device, and the like. The estimation device (300) according to some embodiments may be a device including a processor (310) and a communication interface (320). The estimation device (300) may be implemented with more components than those illustrated in FIG. 4, or with fewer components than those illustrated in FIG. 4. The estimation device (300) may be a user terminal, a server, a network system, or a separate device.

[0083] According to one embodiment of the present disclosure, the processor (310) typically controls the overall operation of the estimation device (300). For example, the processor (310) may control the components included in the estimation device (300) by executing a program stored in the estimation device (300).

[0084] The processor (310) is a component for controlling the overall operation of the estimation device (300). Specifically, the processor (310) controls the overall operation of the estimation device (300) using various programs stored in the storage medium of the estimation device (300). For example, the processor (310) may include a CPU, RAM, ROM, and a system bus. Here, the ROM is a component in which a set of instructions for system booting is stored, and the CPU copies the stored operating system of the estimation device (300) to RAM according to the instructions stored in the ROM and executes the O / S to boot the system. When the system booting is complete, the CPU can copy various stored applications to RAM and execute them to perform various operations. Although the estimation device (300) has been described as including only one CPU, it may be implemented with multiple CPUs (or DSPs, SoCs, etc.) during implementation.

[0085] According to one embodiment of the present invention, the processor (310) may be implemented as a digital signal processor (DSP), a microprocessor, or a time controller (TCON) that processes digital signals. However, the present invention is not limited thereto, and may include one or more of a central processing unit (CPU), a micro controller unit (MCU), a micro processing unit (MPU), a controller, an application processor (AP), a communication processor (CP), or an ARM processor, or may be defined by the corresponding terms. In addition, the processor (110) may be implemented as a system on chip (SoC) having a built-in processing algorithm, a large scale integration (LSI), or may be implemented in the form of a field programmable gate array (FPGA).

[0086] According to one embodiment of the present disclosure, the communication interface (320) creates a channel for communicating with an external device and allows data to be transmitted and received with the external device. The communication interface (320) may include at least one of a short-range communication interface such as a Bluetooth communication interface, a BLE (Bluetooth Low Energy) communication interface, a near field communication interface, a WLAN (Wi-Fi) communication interface, a Zigbee communication interface, an infrared (IrDA, infrared Data Association) communication interface, a WFD (Wi-Fi Direct) communication interface, an UWB (ultra wideband) communication interface, an Ant+ communication interface, and a mobile communication network.

[0087] The estimation device (300) may include a biometric appearance estimation module (330) and / or a weight estimation module (340).

[0088] These biometric appearance estimation modules (330) and / or weight estimation modules (340) may be software embedded in hardware such as a CPU as a predetermined algorithm. Image information may be acquired from an image acquisition module (200).

[0089] The biometric appearance estimation module (330) may include an object recognition unit (331), a first key point detection unit (332), a second key point detection unit (333), an integrated 3D model construction unit (334), and a biometric appearance estimation unit (335).

[0090] The estimation device (300) can receive image information by communicating with the image acquisition module (200). The image acquisition module (200) can be installed within a breeding space and generate image information by capturing one or more objects. The image acquisition module (200) can capture image information by including a camera, an image sensor, etc.

[0091] Figure 5 is an example drawing showing that an individual livestock (T) is recognized by an individual recognition unit (331).

[0092] The object recognition unit (331) may be a predetermined algorithm that can recognize an object of livestock using image information acquired by the image acquisition module (200). The object recognition unit (331) may receive image information including visual data such as images and videos. The object recognition unit (331) may detect an object from the image information. The object recognition unit (331) may detect an object from the image information using feature data about an object to be detected. The object recognition unit (331) may input the image information into a model learned using deep learning or machine learning and output meta information about the object included in the image information. The learned model may include one or more models. The learned model may infer one or more feature values ​​about an object to be detected and detect meta information about the object based on the inferred feature values. The meta information about the object is data including at least one of location information, identification information, and weight information about the detected object, and may include coordinate values ​​in two-dimensional or three-dimensional image information. Meta information about an object may have the form of, for example, [x1, y1, x2, y2, cls]. The learned model may be learned using a training data set including image information acquired through a mobile sensor and meta information (location information, identification information, weight information, etc.) about the object for each image information. The image information used for learning may include image data such as RGB or Depth Map. The training data set may include image information acquired over a predetermined period of time, for example, more than 5,000 data. The learned model may be a model designed including a deep learning model development library, and may include, for example, at least one of Pytorch, Tensorflow, and ONNX.

[0093] This object recognition unit (331) can analyze a three-dimensional image using deep learning and recognize livestock individuals. That is, it can distinguish the environment within the livestock barn (100) and the image of the livestock using deep learning, remove noise, and recognize objects such as livestock. The object recognition unit (331) can detect an object that satisfies a predetermined condition. The predetermined condition may include at least one of weight, length, type of livestock, age, sex, and breed. The object recognition unit (331) can detect a pig weighing between 30 kg and 120 kg, for example. Here, the image information acquired by the image acquisition module (200) may be three-dimensional image information acquired by the depth camera as described above. Of course, it may also include other forms of image information such as temperature distribution depending on the specifications of the movable sensor (220).

[0094] Such object recognition may be based on a predetermined learning data set. For example, as illustrated in Fig. 6, by comparing image information with a learning data set consisting of RGB images, an object can be recognized, key points detected, and the appearance of a living body can be estimated.

[0095] As a result of verifying the object recognition unit (331) according to the embodiment, detection was successful for 225 out of 235 verification images, confirming that the object recognition accuracy was 95.74%.

[0096] Figures 7 and 8 are example drawings representing key points of a specific livestock object detected by a key point detection unit (332).

[0097] According to embodiments of the present disclosure, the first key point detection unit (332) can detect a specific body point of an object from image information. As illustrated in FIG. 7, information on a shoulder point of a pig can be detected as (207, 208, shoulder), information on a neck point of the pig can be detected as (150, 177, neck), information on a center point of the pig can be detected as (301, 242, center), information on a hip point of the pig can be detected as (453, 264, hip), and information on a tail point of the pig can be detected as (303, 276, tail).

[0098] The first key point detection unit (332) can detect specific body points of an object, such as shoulders, neck, center, hips, and tail, using a detection model. The first key point detection unit (332) can generate indications for the detected points and information on coordinates and names for the points and display the information through an output device. Here, the detection model may be learned using machine learning or deep learning. The detection model may be learned using a training data set including a region including an object of image information and a specific body point within the region. The detection model may be a model designed and learned using a deep learning model development library. The detection model may be designed to infer key point values ​​for a plurality of body points. The detection model can detect a plurality of body points based on the key point values. The detection model can detect a plurality of body points, and as the number of points to be detected increases, the required CPU capacity may increase. The detection model can adjust the number of points to be detected based on the capacity of the computing device running, such as the CPU or GPU, and can be designed to detect up to five points. As the number of points to be detected increases, the output waiting time of the detection model may also increase. The detection model can correct the key point values ​​detected using image analysis technology during the process of detecting multiple body points. The detection model can select and remove unwanted key point values ​​from among the key point values ​​detected using image analysis technology, or can modify some of the key point values ​​among the detected values. The image analysis technology can output key point values ​​related to specific body points of the object by additionally detecting coordinate values ​​for the object using interpolation.More specifically, the image analysis technique may be implemented to calculate a second or higher regression equation based on two or more coordinate values ​​associated with an object, and use the calculated regression equation to calculate additional detection coordinate values ​​for the object. The additional detection coordinate values ​​may be key point values.

[0099] The regression equation may be an equation calculated for each individual. For example, the relationship between points such as the head, legs, and tail of a pig may be calculated as a regression equation. The relationship between points such as the head, legs, and tail of a cow may be calculated as a regression equation. The detection model may adjust the number of key point values ​​to be detected by considering the resolution of the image information. For a first resolution, the detection model may detect a first number of key point values, and may detect a specific body point of the object with the first number of key point values. The higher the resolution, the more detectable key point values ​​may be. If a large number of key point values ​​are calculated compared to the resolution of the image information, the first and / or second key point detection units (332, 333) may output an error message. The number of key point values ​​may be determined based on the acquired image information, and may be determined by a table including information on the number for each preset resolution or a mathematical equation for the number for the preset resolution. The higher the resolution and pixel count of the acquired image information, the greater the number of key point values. The resolution of the image information can be determined by the frame rate of the image information, the number of pixels per frame, etc.

[0100] Here, the specific body point can be one of the head (neck), shoulder, or buttocks (hip) of the object to be detected.

[0101] The first key point detection unit (332) can output a specific body point based on the image information of the object recognized by the object recognition unit (331). The method for deriving the specific body point of the object is not limited. The information derived from the first key point detection unit (332) is displayed as illustrated in FIG. 7 and may be referred to as first key point information. Additionally, the first key point detection unit (332) can detect and display the detected points and the lines (A81, A82) connecting the points for each object, as illustrated in FIG. 8.

[0102] The second key point detection unit (333) can output the appearance characteristic values ​​of the object according to its pose using the first key point information derived from the first key point detection unit (332). For example, based on the depth information and the change in the depth information according to the pose of the object, the detection of a specific body point of the object can be performed using a detection model. As another example, based on the temperature information of the body part of the object, the detection of a specific body point of the object can be performed. In this case, the image information of the object may further include infrared information as well as information by visible light. The information derived by the second key point detection unit (333) as described above may be referred to as second key point information. In addition, the first key point information and the second key point information may be collectively referred to as key point information.

[0103] According to embodiments of the present disclosure, the detection model can have an automatic detection accuracy of 96% for key point information among 174 verification images. The automatic detection accuracy can be calculated based on whether the result values ​​of the first and second key point detection units (332, 333) according to the embodiment match the result values ​​of visual detection.

[0104] The integrated 3D model building unit (334) can output 3D rendering data by merging the key point information, the appearance characteristic values ​​for each posture, and the RGB image, depth information, near-infrared image, short-wave infrared image, and thermal imaging biometric image acquired from the movable sensor (220). The integrated 3D model building unit (334) can output 3D rendering data using a learned model. The learned model may be a model learned using machine learning or deep learning.

[0105] FIG. 9 and FIG. 10 are exemplary diagrams showing that key points, body appearance, and body weight of a specific livestock individual are derived through the first and second key point detection units (332, 333), body appearance estimation unit (335), and body weight estimation unit (340). FIG. 9 is a diagram showing a normal individual, and FIG. 10 is a diagram showing an abnormal individual.

[0106] The biological appearance estimation unit (335) can output the appearance information of an object using the information derived from the first and second key point detection units (332, 333) and the information derived from the integrated 3D model construction unit (334).

[0107] As illustrated in FIG. 9, the appearance information of an object derived from the biometric appearance estimation unit (335) may include a length value, an area value, a width value, a height value, etc. of the object. For example, the biometric appearance estimation unit (335) may output appearance information for a shoulder point by using the shoulder information (-12, -6, 90, shoulder) calculated by the first key point detection unit (332). The biometric appearance estimation unit (335) may output appearance information for a neck point by using the neck information (-58, 92, -2, neck) calculated by the first key point detection unit (332). The biometric appearance estimation unit (335) may output appearance information for a center point of an object by using the center point information (102, -1, 36, center) calculated by the first key point detection unit (332). The living body appearance estimation unit (335) can calculate appearance information for the hip point of the object as (-30, 88, -1, hip). The living body appearance estimation unit (335) can output appearance information for the tail point of the object using the tail point information (-4, 23, 96, tail) calculated by the first key point detection unit (332). The living body appearance estimation unit (335) can input the values ​​calculated by the first key point detection unit (332) into an appearance estimation model to output appearance information for each point of the object. The living body appearance estimation unit (335) can generate and display appearance information for each point of the object. The appearance estimation model can use deep learning and / or 3D image analysis. Here, at least some of the derived appearance information may be, for example, height values ​​of key points of the object derived by the first and second key point detection units (332, 333), distance values ​​between key points, etc.

[0108] The length value of an object can be the distance from the neck to the tail. The length value of an object can take into account height differences, similar to the method of measuring length using a tape measure.

[0109] The height difference of an object can be measured when each point has a different height based on a plan view of the object's upper part. The length value of an object can be calculated by considering the height difference of the object. If the height difference is not considered, the length value of the object can be calculated as the shortest length value. If the height difference is not considered, the length value of the object may be shorter than the actual length value of the object.

[0110] According to embodiments of the present disclosure, the height difference of an object can be calculated using the distance information between cameras, which is a coordinate value acquired by the first key point detection unit (320). The area value of an object can refer to the surface area of ​​the object based on the plan view of the object. The area value of an object can be the surface area of ​​the object excluding a specific body point, for example, the head of the object. The area value of an object can have different values ​​even for the same object due to movement of the object. In order to accurately calculate the area value of an object, the area value of the object can be calculated by limiting only a part of the object at a certain height, for example, 50 centimeters, based on the floor surface. The height value of an object is the shortest distance from the floor surface to a specific body point of the object, i.e., the shortest distance in the direction of a vertical vector from the floor surface. The floor surface of an object can be specified using depth map data acquired from a 'movement sensor (220)'. A predetermined 3D data-based floor surface detection technology can be applied to obtain a plane vector of the floor surface of an object.

[0111] According to embodiments of the present disclosure, it is possible to compare and contrast previously acquired information with currently acquired information to determine whether a currently detected object is the same object as a previously detected object. According to embodiments of the present disclosure, the growth and / or weight of a specific livestock individual can be monitored or tracked using appearance information about the same object.

[0112] The weight estimation module (340) may be a software embedded in hardware such as a CPU as a predetermined algorithm.

[0113] The weight estimation module (340) can output the weight value of the object based on the point-by-point appearance information of the object derived from the living body appearance estimation unit (335), as illustrated in FIG. 9. The weight estimation module (340) can infer and output the weight value of the object by inputting the point-by-point appearance information of the object into a weight estimation model. The weight estimation module (340) can calculate and display the length value as 98.68 cm and the weight value as 118.88 kg based on the appearance information of (-12,-6, 90, shoulder), (-58, 92, -2, neck), (-30, 88, -1, hip), (102,-1, 36, center), and (-4,23, 96, tail). The weight estimation module (340) can infer the weight value of an object based on the appearance information of (14, -5, 87, shoulder), (-52, 95, 8, neck), (-34, 90, 0, hip), and (97, 7, 9, center). If the weight value of the object is outside the preset reference range, the weight value of the object may not be displayed. An alarm message may be generated for such an object. The weight output of the object may be calculated based on the appearance information of the object, such as the length, area, width, and height, output from the biological appearance estimation unit (335). Here, the weight estimation model may be a model learned by a machine learning or deep learning method. The weight estimation model can output the weight value using a predetermined formula. The weight estimation model can output the weight value of the object by inputting input variables, such as the length value, area, width, and height value of the object. The weight estimation model can output the weight value using the following mathematical expression 1.

[0114] [Mathematical Formula 1]

[0115] Object's weight value = K1 × chest circumference + K2 × body length

[0116] Here, K1 may be a chest phase constant, and K2 may be a body length constant. K1 and / or K2 may be values ​​determined through a detection model. The weight estimation model is not limited to the above mathematical expression 1, and may be trained to output a weight value based on at least two input variables among the length value, area value, width value, and height value of the object. The weight estimation model may be a model trained using a multivariate statistical model such as MLR, PLSR, SVM, or a deep learning model such as CNN.

[0117] The weight value of an object can be output by inputting information derived from the first and second key point detection units (332, 333) and information derived from the body appearance estimation unit (335).

[0118] The weight estimation module (340) can determine whether an object is abnormal or normal based on its weight value. The weight estimation module (340) can determine a reference value that serves as a criterion for determining whether an object is abnormal or normal through a weight estimation model. The weight estimation module (340) can process a region of image information to display an alarm message indicating whether the object is normal or abnormal. In this case, the object may be an object detected in the displayed image information.

[0119] As shown in Fig. 10, in the case of an object that does not meet a predetermined criterion, a predetermined error message may be generated.

[0120] It was confirmed that the weight value of the object output through the weight estimation module (340) has an accuracy of approximately 92% with respect to the weight value of the actual object.

[0121] The image acquisition module (200) according to some embodiments may be a device including a processor (310) and a communication interface (320). The estimation device (300) may be implemented with more components than those illustrated in FIG. 4, or with fewer components than those illustrated in FIG. 4. The estimation device (300) may be a user terminal, a server, a network system, or a separate device.

[0122] According to one embodiment of the present disclosure, the processor (310) typically controls the overall operation of the estimation device (300). For example, the processor (110) may control the components included in the estimation device (300) by executing a program stored in the estimation device (300).

[0123] The processor (310) is a component for controlling the overall operation of the estimation device (300). Specifically, the processor (310) controls the overall operation of the estimation device (300) using various programs stored in the storage medium of the estimation device (300). The processor (310) can execute the biological appearance estimation module (330) to cause the device (300) to detect an object and generate information (coordinate values, appearance information, weight value, length value, etc.) about the detected object. For example, the processor (310) may include a CPU, RAM, ROM, and a system bus. Here, the ROM is a component in which a command set for system booting is stored, and the CPU copies the stored operating system of the estimation device (300) to the RAM according to the command stored in the ROM and executes the O / S to boot the system. When the system booting is complete, the CPU can copy various stored applications to the RAM and execute them to perform various operations. In the above, the estimation device (300) is described as including only one CPU, but it may be implemented with multiple CPUs (or DSP, SoC, etc.).

[0124] According to one embodiment of the present invention, the processor (310) may be implemented as a digital signal processor (DSP), a microprocessor, or a time controller (TCON) that processes digital signals. However, the present invention is not limited thereto, and may include one or more of a central processing unit (CPU), a micro controller unit (MCU), a micro processing unit (MPU), a controller, an application processor (AP), a communication processor (CP), or an ARM processor, or may be defined by the corresponding terms. In addition, the processor (110) may be implemented as a system on chip (SoC) having a built-in processing algorithm, a large scale integration (LSI), or may be implemented in the form of a field programmable gate array (FPGA).

[0125] According to one embodiment of the present disclosure, the communication interface (320) creates a channel for communicating with an external device and allows data to be transmitted and received with the external device. The communication interface (320) may include at least one of a short-range communication interface such as a Bluetooth communication interface, a BLE (Bluetooth Low Energy) communication interface, a near field communication interface, a WLAN (Wi-Fi) communication interface, a Zigbee communication interface, an infrared (IrDA, infrared Data Association) communication interface, a WFD (Wi-Fi Direct) communication interface, an UWB (ultra wideband) communication interface, an Ant+ communication interface, and a mobile communication network.

[0126] <Effects of the present invention>

[0127] According to the livestock weight estimation system according to the embodiment of the present invention, livestock individuals within a livestock shed (100) can be accurately recognized and the weight of the individuals can be accurately predicted.

[0128] First, since the image acquisition module (200) is installed in the livestock house (100) in the same form as the embodiment, the movable sensor (220) can be displaced in the direction in which the guide rail (210) extends and in the direction in which the breeding space (110) is arranged, thereby effectively and efficiently acquiring image information of livestock in the breeding space (110).

[0129] In addition, the estimation device (300) includes a processor (310), a communication interface (320), a bio-appearance estimation module (330), and a weight estimation module (340). The weight estimation module (400) estimates the weight of an individual using information derived from the bio-appearance estimation module (330). Each part includes a deep learning and 3D image analysis algorithm, so that the accuracy of deriving the weight of the individual can be increased. Accordingly, for example, when selecting pigs for shipment, the weight can be derived with high accuracy, so that the quality of shipment can be improved.

[0130] In addition, since it is possible to track and monitor the growth of an individual, abnormal growth and signs of each individual can be detected, and early measures can be taken accordingly.

[0131] Fig. 11 is a drawing showing a pig sorting device (1) according to one embodiment of the present invention. Fig. 12 is a drawing showing the pig sorting device (1) of Fig. 11 as seen from above.

[0132] A shipping money sorting device (1) according to one embodiment of the present invention includes a passage device (10).

[0133] A passage device (10) may include a passage frame (11) having a passage space (12) on the inside, a passage entrance (13) located on one side, and a passage exit (14) located on the other side, a vision inspection device (15) installed on the passage frame (11), and a selection display device (16).

[0134] The passage frame (11) may be a frame with a steel frame structure formed by multiple steel bars. Therefore, it may be suitable for viewing the interior from the outside. Since the passage frame (11) has a steel frame structure and the interior and exterior are not visually obscured, an object shipped can enter the passage frame (11) without any discomfort. In addition, a user can easily check the interior space of the passage frame (11). In addition, the passage frame (11) may have an appropriate weight for a user to lift or move.

[0135] According to an embodiment, the passage frame (11) may not have a floor structure forming the bottom surface. That is, the passage frame (11) may have a plurality of iron bars forming the side surfaces, but may have an open structure without iron bars on the bottom surface. Accordingly, the weight of the entire passage frame (11) may be reduced, and the installation and removal of the shipping pig sorting device (1) may be simplified.

[0136] Although not shown, the passage frame (11) may include a base forming the bottom surface, or may be connected to the base. If necessary, the passage frame (11) and the base may be detachable from each other. Accordingly, the passage frame (11) may be stably supported on the ground. In addition, the passage frame (11) and the passage device (10) may be prevented from unintentionally moving out of position or falling over due to external force. In addition, if the passage frame (11) and the base are detachable from each other, the user may easily lift or move the passage device (10).

[0137] The passage frame (11) may have a passage space (12) on the inside. The size and shape of the passage space (12) are not limited. According to an embodiment, the passage space (12) may have a size and shape suitable for one adult pig to be located and pass through. That is, the passage space (12) may have a rectangular shape having a height, length, and width greater than the height, length, and width of one adult pig. For example, the passage space (12) may have a size of 1,400 mm in length and 600 mm in width.

[0138] The passage frame (11) may include a passage inlet (13) and a passage outlet (14). The space communicates with the external space through the passage inlet (13) and the passage outlet (14). That is, after entering the passage space (12) through the passage inlet (13), one can exit to the outside through the passage outlet (14). The size and shape of the passage inlet (13) and the passage outlet (14) are not limited. According to an embodiment, the passage inlet (13) and the passage outlet (14) may have a size and shape suitable for the passage of one adult piglet. That is, they may have a rectangular shape having a length and width greater than the height and width of one adult piglet.

[0139] The vision inspection device (15) is a device that can measure the weight and body weight value of a pig to be shipped that has entered the passage space (12) of the passage frame (11). The vision inspection device (15) can measure the weight of a pig to be shipped by visually capturing the body information of the pig to be shipped located in the passage space (12). The vision inspection device (15) can transmit image information captured by an estimation device and receive the weight value of the pig to be shipped from the estimation device through the image information. The body information is not limited, and may include, for example, the length of the pig to be shipped, the height of each body part (shoulder, back, hip, etc.), the width of each body part (head, shoulder, back, hip, etc.), etc.

[0140] Specifically, the vision inspection device (15) may include an optical device, and a predetermined computing device capable of exchanging signals with the optical device, and may include a communication device that mediates signal exchange between the optical device and the computing device. For example, the optical device may be a predetermined camera module. The camera module may acquire all visual and optical information, such as depth information and RGB information, and its type and operating method are not limited. The optical device may capture image information within a passage space. The optical device may be designed to capture an image when an object is detected within the passage space. The optical device may transmit the captured image information to the vision inspection device (15).

[0141] Meanwhile, according to an example, the calculation device may be a predetermined device located outside the shipping pig sorting device (1). The calculation device may include a predetermined algorithm that calculates the weight of the shipping pig object using visual information of the shipping pig object captured by the optical device. The algorithm may be, for example, an algorithm that learns using AI and derives the weight through data, but is not limited thereto.

[0142] For example, as an implementation example of the above-mentioned computing device, a predetermined biological appearance estimation module and a weight estimation module that exchange signals with the vision inspection device (15) may be provided. This will be described later.

[0143] The above vision inspection device (15) may be connected to the passage frame (11). Specifically, according to one embodiment, the vision inspection device (15) may have a vertical bar extending to the upper portion of the passage frame (11), and may have a configuration in which an optical device is connected to the vertical bar. The optical device can visually view the interior of the space from above and below. Therefore, visual information of an object located within the space can be accurately captured.

[0144] In one embodiment, the vision inspection device (15) may be detachably attached to the passage frame (11). Therefore, when using the passage device (10), the process of installing it in the pigsty (PH) or retrieving it from the pigsty (PH) can be easily accomplished.

[0145] The selection display device (16) is a device that can display a predetermined display on a shipping pig located within a passage frame (11). Here, displaying a display may mean spraying a display-related substance. For example, the selection display device (16) may be a spray capable of displaying a color. The location of the spray, the amount of pigment sprayed, the type of pigment, etc. are not limited. The selection display device (16) may be designed so that a display corresponding to the weight information of an object within the passage frame (11) and the passage space is provided to the object. For example, when a standard range of 110 to 130 kg is set for an object, the vision inspection device (15) may be controlled to spray a first display substance to a first object having a weight within the standard range, and to spray a second display substance to a second object having a weight outside the standard range. A control signal for spraying the first display substance may be generated based on the weight value of the first object and transmitted to the selection display device (16). The first and second indicator materials can be visually distinct, allowing for direct identification of the first and second objects. The above description is merely an example, and one of the multiple indicator materials can be sprayed onto an object based on multiple criteria.

[0146] According to an embodiment, the sorting display device (16) is positioned adjacent to the passage exit (14) of the passage device (10), is positioned at a predetermined height from the ground, and may have a predetermined spray angle in a diagonal direction to spray the pigment toward the passage entrance (13). Therefore, when the pigment filled in the sorting display device (16) is sprayed, the pigment may be sprayed from the front of the individual pigs being shipped located in the passage space (12) toward the body of the individual pigs being shipped. Therefore, the pigment may be sprayed from the front of the individual pigs being shipped toward the body of the individual pigs being shipped. Therefore, the stress on the individual pigs being shipped due to the spraying sound may be relatively reduced compared to when the pigment is sprayed from the rear of the individual pigs being shipped. In addition, the stress on the individual pigs being shipped due to the pigment spray may be relatively reduced compared to when the pigment is sprayed toward the head of the individual pigs being shipped.

[0147] According to one embodiment, the passage device (10) may further include a tag device (17). The tag device (17) may be a predetermined tag recognition device that recognizes a tag displayed on the body of a pig to be shipped. The tag may be, for example, an RFID tag, and the type thereof is not limited. The passage device (10) further includes the tag device (17) as described above, and the tag device (17) recognizes the tag displayed on the pig to be shipped, thereby recognizing the pig to be shipped that has passed through the passage device (10), determining whether the pig to be shipped has passed and been selected, and converting the data into information. The vision inspection device (15) identifies objects through the tag device (17), and statistically calculates the number of pigs to be shipped that satisfy predetermined passing criteria and the number of pigs to be shipped that satisfy predetermined selection criteria by considering the identified objects and the weight values ​​of the objects. The tag device (17) may recognize a tag displayed on each object, a tag inserted into an object, or a tag attached to an object. The vision inspection device (15) can classify and store information about an object sprayed with a first marking material and an object sprayed with a second marking material. The vision inspection device (15) can identify an object through a tag device (17) and detect whether the object is within a reference weight range or outside the reference weight range based on a weight value previously stored for the object. The vision inspection device (15) can recalculate the weight value of the object based on newly captured image information and determine again whether the object is included within the reference weight range. The reference weight range can be set by an administrator. The reference weight range can be changed.

[0148] A pig sorting device (1) according to an embodiment of the present invention can be easily installed and removed within a pigsty. The pig sorting device (1) can operate in conjunction with an estimation device. The vision inspection device (15) can capture image information of an object and receive appearance information, weight values, etc. of the captured object while communicating with the estimation device (300). Therefore, weight measurement and sorting of pigs can be easily performed. The pig sorting device according to the embodiment has a relatively small weight, can be simply installed within a pigsty, and can select and mark pigs simply by having pigs pass through the pig sorting device. Therefore, installation and removal costs are reduced, and user convenience can be improved. In addition, since weight measurement and sorting are performed while pigs simply pass through a passage, stress on the pigs can be reduced, and the accuracy of weight measurement can be improved.

[0149] Since the pig sorting device according to the embodiment of the present invention can sort individual pigs by installing them within a growing space in a pigsty, it can overcome spatial limitations.

[0150] According to an embodiment of the present invention, a pig sorting device for shipping can accurately measure the weight of an individual within a pigsty. In particular, the device includes an individual recognition unit, first and second key-point detection units, an integrated three-dimensional model construction unit, a living body appearance estimation unit, and a weight estimation module, each of which includes deep learning and three-dimensional image analysis algorithms, thereby increasing the accuracy of deriving the weight of the individual. Accordingly, for example, when sorting pigs for shipping, the weight can be derived with high accuracy, thereby improving the quality of shipping.

[0151] According to one embodiment of the present invention, a pig sorting device for shipping includes: a passage device; wherein the passage device includes: a passage frame including a passage space on the inside, and a passage inlet and a passage outlet that open the passage space to the outside; a vision inspection device installed in the passage frame and configured to measure the weight of an object that has entered the passage space and generate a signal; and a sorting display device installed in the passage frame and operated by a signal transmitted from the vision inspection device; wherein the vision inspection device generates an operating signal for operating the sorting display device when the weight of the object that has entered the passage space satisfies a predetermined standard.

[0152] In one embodiment, the vision inspection device includes a vertical bar connected to the passage frame, and an optical device connected to the vertical bar and visually capturing the interior of the passage space.

[0153] In one embodiment, the vision inspection device includes a computing device that receives a signal provided from the optical device, and the computing device is equipped with an algorithm that calculates the weight of an object using visual information provided from the optical device.

[0154] In one embodiment, the passage device further includes a tag device that recognizes a tag provided on an object.

[0155] In one embodiment, the reference weight is a weight at which an object that has entered the passage space is judged to be suitable for shipment, and the object's weight is 110 to 130 kg.

[0156] In one embodiment, the passage frame has a plurality of iron bars forming the side portions and has an open bottom structure.

[0157] According to one embodiment, the apparatus further comprises a classification device connected to the passage device, wherein the classification device comprises a classification frame including a classification space on the inside, and a classification inlet and a classification outlet that open the classification space to the outside; an opening / closing door that opens / closes the classification outlet; and a door operating device that operates the opening / closing door; wherein the classification inlet is connected to the passage exit.

[0158] In one embodiment, the classification outlet includes a first classification outlet and a second classification outlet, each provided on at least one side of the classification frame, and the opening / closing door selectively opens and closes the first classification outlet and the second classification outlet.

[0159] In one embodiment, the opening / closing door is a rotating door rotatable about a door axis, the opening / closing door is located within the sorting space, the first sorting exit and the second sorting exit are respectively disposed on one side and the other side of the sorting frame, and when the opening / closing door rotates in the direction of the first sorting exit, the first sorting exit is closed and the second sorting exit and the sorting inlet are communicated, and when the opening / closing door rotates in the direction of the second sorting exit, the second sorting exit is closed and the first sorting exit and the sorting inlet are communicated.

[0160] In one embodiment, the vision inspection device provides an operating signal for operating the door operating device, and the door operating device is operated by the operating signal provided from the vision inspection device.

[0161] In one embodiment, the classification frame has a plurality of iron bars forming a side portion and has an open bottom structure.

[0162] According to one embodiment, a livestock appearance estimation module for deriving livestock appearance information using image information acquired from the vision inspection device; and a weight estimation module for deriving livestock weight using the livestock appearance information derived from the livestock appearance estimation module. and the bio-appearance estimation module includes an object recognition unit that recognizes an object of livestock, a first key-point detection unit that derives a specific body point of the object recognized by the object recognition unit, a second key-point detection unit that derives an appearance characteristic of the object according to its posture using the first key-point information derived from the first key-point detection unit, an integrated 3D model construction unit that derives a 3D model by merging the information derived from the first key-point detection unit and the second key-point detection unit with an RGB image, depth information, a near-infrared image, a short-wave infrared image, and a thermal image bio-image acquired from the movable sensor, and a bio-appearance estimation unit that derives appearance information of the object from the 3D model and the appearance characteristic of the object according to its posture, and the weight estimation module derives the weight of the object based on the appearance of the object derived from the bio-appearance estimation unit.

[0163] In one embodiment, the object recognition unit recognizes an object by comparing a learning data set consisting of RGB images with image information.

[0164] Fig. 13 is a drawing showing the pig sorting device (1) of Fig. 11 being installed and used in a pigsty (PH).

[0165] Below, the use and effect of a sorting device (1) for sorting out shipped pigs including a passage device according to an embodiment of the present invention will be described.

[0166] The pig sorting device (1) according to the embodiment can be installed in a pigsty (PH) where piglets (P1) are raised. As illustrated in FIG. 13, when a piglet (P1) raised in the pigsty (PH) enters the passage device (10) through the passage entrance (13) of the passage device (10) of the piglet sorting device (1), the vision inspection device (15) can visually capture the body information of the piglet (P1). At this time, a predetermined feeding means can be provided in the passage device (10) so as to induce the entry of the piglet (P1).

[0167] If the vision inspection device (15) determines that the shipping pig individual (P1) located in the passage space (12) within the passage frame (11) satisfies a predetermined standard, the selection display device (16) can be operated. For example, if the shipping pig individual (P1) is determined to have a weight suitable for shipment, the selection display device (16) can be operated, and a selection display (S) can be made on the shipping pig individual (P1) within the passage space (12). That is, what is measured by the vision inspection device (15) is the weight of the shipping pig individual (P1), and whether or not the selection display device (16) is operated can be determined based on whether or not the shipping pig individual (P1) has a weight suitable for shipment. Here, whether or not the weight is suitable for shipment can be determined differently depending on the embodiment, and for example, the weight of the individual can be in the range of 110 to 130 kg.

[0168] If the vision inspection device (15) determines that the shipping pig object (P1) located in the space within the passage frame (11) has a weight that is not suitable for shipping, the selection display device (16) does not operate. Then, when the shipping pig object (P1) located in the passage space (12) within the passage frame (11) passes through the passage exit (14) of the passage frame (11), another shipping pig object (P1) can enter the passage entrance (13) within the passage frame (11), and the above process can be repeated.

[0169] By repeating the above process, when the selection of all pigs (P1) to be shipped within the pigsty (PH) is completed, or when it is confirmed that a certain number of pigs to be shipped are within the pigsty (PH), the pigs (P2) with the selection mark (S) are guided into a designated shipping space, and then the pigs (P2) located within the shipping space can be shipped. The formation of the shipping space and the shipping of the pigs to be shipped can be performed automatically or manually. For example, as illustrated in Fig. 3, a method may be possible in which a space within the pigsty (PH) is partitioned using a predetermined sunshade (SS), and then the partitioned space is used as a shipping space.

[0170] The above weight measurement, judgment, and selection marking processes can be performed on each pig in the pigsty (PH). At this time, a predetermined guide means can be provided to facilitate the pig in the pigsty (PH) to enter the passageway inlet (13) of the passageway frame (11) and then exit through the passageway exit (14). The guide means can be, for example, a predetermined feeding means for providing feed, and the feeding means can be provided in a form in which the passageway space (12) in the passageway frame (11) is provided. However, the present invention is not limited thereto.

[0171] The pig sorting device (1) according to the embodiment can be easily installed and removed within a pigsty (PH). Therefore, weight measurement and sorting of pigs can be easily performed.

[0172] For example, conventional pig sorting methods for shipping pigs involve a weight measuring device equipped with a load cell, which results in a very heavy sorting device, which incurs high installation and removal costs and is difficult to handle. Furthermore, because pigs must pass over the load cell, guiding the pigs is difficult, putting a lot of stress on the pigs, and reducing the accuracy of weight measurement due to resistance from the pigs on the load cell.

[0173] However, the pig sorting device (1) according to the embodiment has a relatively small weight, is simply installed in a pigsty (PH), and can sort and mark pigs for shipment simply by passing them through the pig sorting device (1) in the pigsty (PH). Therefore, installation and removal costs are reduced, and user convenience can be improved. In addition, since weight measurement and sorting are performed while the pigs for shipment simply pass through the passage device (10), stress applied to the pigs for shipment can be reduced, and the accuracy of weight measurement can be improved.

[0174] In particular, it can be more advantageous in the case of a pigsty (PH) where the area is not sufficiently large. That is, in the case of a narrow pigsty, it may not be easy to install a pig sorting device (1) for slaughtering pigs, but the pig sorting device (1) according to the embodiment of the present invention can sort pigs by installing it within the growing space of the pigsty (PH), thereby overcoming spatial limitations.

[0175] In addition, according to the embodiment, the passage device (1) can have a size and shape suitable for one shipping pig to pass through, so that the posture of the pig to be weighed is maintained constant, so that weight measurement using the vision inspection device (15) can be easily performed.

[0176] Fig. 14 is a drawing showing a pig sorting device (1) according to one embodiment of the present invention. Fig. 15 is a drawing showing the pig sorting device (1) of Fig. 14 as seen from above.

[0177] A shipping money sorting device (1) according to one embodiment of the present invention includes a passage device (10); and a classification device (20).

[0178] Since the passage device (10) is as described in the above embodiment, its description is omitted.

[0179] The classification device (20) may include a classification frame (25a), an opening / closing door (26) installed on the classification frame (25b), and a door operating device (27) that operates the opening / closing door (26). The classification device (20) may be connected to the passage device (10).

[0180] The sorting frame (25) may be a frame with a steel frame structure formed by multiple steel bars. Therefore, it may be suitable for viewing the interior from the outside. Since the sorting frame (25) has a steel frame structure and the inside and outside are not visually obscured, the shipped objects can pass through the inside of the sorting frame (25) without any discomfort. In addition, the user can easily check the space inside the sorting frame (25). In addition, the sorting frame (25) may have an appropriate weight for the user to lift or move.

[0181] According to an embodiment, the sorting frame (25) may not have a floor structure forming the bottom surface. That is, the sorting frame (25) may have a plurality of iron bars forming the side surfaces, but may have an open structure without iron bars on the bottom surface. Accordingly, the weight of the entire sorting frame (25) may be reduced, and the installation and removal of the sorting device (2) for sorting the dried radish may be simplified.

[0182] Although not shown, the sorting frame (25) may include a base forming the bottom surface, or may be connected to the base. If necessary, the sorting frame (25) and the base may be detachable from each other. Accordingly, the sorting frame (25) may be stably supported on the ground. In addition, the sorting frame (25) and the sorting device (20) may be prevented from unintentionally moving out of position or falling over due to external force. In addition, if the sorting frame (25) and the base are detachable from each other, the user may easily lift or move the sorting device (20).

[0183] The classification frame (25) may have a classification space (28) on the inside. The size and shape of the classification space (28) are not limited.

[0184] The sorting frame (25) may include a sorting inlet (21) and a sorting outlet. The sorting space (28) communicates with the external space through the sorting inlet (21) and the sorting outlet. That is, after entering the sorting space (28) through the sorting inlet (21), one can exit to the outside through the sorting outlet. The size and shape of the sorting inlet (21) and the sorting outlet are not limited. According to an embodiment, the sorting inlet (21) and the sorting outlet may have a size and shape suitable for one adult pig to pass through.

[0185] The sorting inlet (21) may be located at one end in the longitudinal direction of the sorting frame (25). The sorting inlet (21) may be connected to the passage outlet (14) of the passage frame (11). Therefore, the shipped pigs that have exited the passage space (12) through the passage outlet (14) of the passage frame (11) may enter the sorting space (28) of the sorting frame (25) through the sorting inlet (21) of the sorting frame (25).

[0186] The sorting outlet may include a first sorting outlet (22) and a second sorting outlet (23). The first sorting outlet (22) may be located on one side of the sorting frame (25), and the second sorting outlet (23) may be located on the other side of the sorting frame (25). For example, the sorting inlet (21) may be located on one longitudinal end of the sorting frame (25), and the sorting outlet may be located on the other longitudinal end of the sorting frame (25), but the first sorting outlet (22) and the second sorting outlet (23) may be located on both sides.

[0187] According to an embodiment, the classification frame (25) may include a connecting portion (25a) and a classification portion (25b). The connecting portion (25a) has a connecting space (212), and the classification portion (25b) has a classification space (28). The connecting portion (25a) may be a portion having a shape similar to the passage frame (11). The connecting portion (25a) may have a passage-shaped space extending in one direction. The classification inlet (21) may be provided at an end of the connecting portion (25a). The classification portion (25b) may be a portion having a wider width in the horizontal direction than the connecting portion (25a). Accordingly, the classification portion (25b) may have a space having a wider width in the lateral direction than the connecting portion (25a). The first classification outlet (22) and the second classification outlet (23) may be provided on both sides of the classification portion (25b).

[0188] As described above, the classification frame (25) may have an overall hammer shape when viewed from above by including a connecting portion (25a) and a classification portion (25b). For example, the internal space of the connecting portion (25a) may have a size of 1,000 mm in length and 600 mm in width. For example, the internal space of the classification portion (25b) may have a size of 1,000 mm in length and 1,300 mm in width.

[0189] The opening / closing door (26) can be positioned within the classification space (28) of the classification frame (25). Specifically, the opening / closing door (26) can be positioned within the classification section (25b). The opening / closing door (26) has a predetermined door axis and can rotate around the door axis.

[0190] The door operating device (27) is a device that can operate the opening / closing door (26). For example, as illustrated in FIG. 15, the door axis of the opening / closing door (26) is connected to the door operating device (27), and the door operating device (27) can rotate the opening / closing door (26) around the door axis. The door operating device (27) can operate the opening / closing door (26) by receiving a signal provided from the outside. For example, an operating signal can be transmitted from the vision inspection device (15) to the door operating device (27).

[0191] The opening / closing door (26) and the door operating device (27) can be positioned within the classification space (28) of the classification frame (25). In one example, the door operating device (27) is configured in the form of an erected bulkhead and can be positioned within the space of the classification section (25b). In addition, the opening / closing door (26) connected to the door operating device (27) has one end that rotates around the door axis as indicated by an arrow, and the other end that can come into contact with one side or the opposite side of the connecting portion (25a) depending on the rotation of the opening / closing door (26).

[0192] According to the operation of the opening / closing door (26) as described above, communication and blocking between the classification inlet (21) and the first classification outlet (22) of the classification frame (25), and communication and blocking between the classification inlet (21) and the second classification outlet (23) can be selectively performed.

[0193] That is, when the opening / closing door (26) rotates in the direction of the second sorting exit (23) of the arrow, the sorting inlet (21) of the sorting frame (25) and the first sorting exit (22) are connected, and the sorting inlet (21) and the second sorting exit (23) are blocked. Therefore, the pigs that have passed through the sorting inlet (21) and entered the sorting frame (25) can exit in the direction of the first sorting exit (22).

[0194] Meanwhile, when the opening / closing door (26) rotates in the direction of the first sorting exit (22) of the arrow, the sorting inlet (21) and the second sorting exit (23) of the sorting frame (25) are connected, and the sorting inlet (21) and the first sorting exit (22) are blocked. Therefore, the pigs that have passed through the sorting inlet (21) and entered the sorting frame (25) can exit in the direction of the second sorting exit (23).

[0195] That is, to explain this differently, the first classification outlet (22) and the second classification outlet (23) are respectively arranged on one side and the other side of the classification frame (25), and when the opening / closing door (26) rotates toward the first classification outlet (22), the first classification outlet (22) is closed and the second classification outlet (23) is opened, and when the opening / closing door (26) rotates toward the second classification outlet (23), the second classification outlet (23) is closed and the first classification outlet (22) is opened. Here, the opening of the first classification outlet (22) or the second classification outlet (23) can be understood as the first classification outlet (22) and the classification inlet (21) being connected, and the second classification outlet (23) and the classification inlet (21) being connected, respectively.

[0196] Figures 16 to 20 are drawings sequentially showing the usage form of the shipping money sorting device (1) of Figure 14.

[0197] Hereinafter, the use and effect of a pig sorting device (1) including a passage device and a sorting device (20) according to an embodiment of the present invention will be described.

[0198] The pig sorting device (1) according to the embodiment can be installed in a pigsty (PH).

[0199] As illustrated in FIG. 16, when a piglet (P) growing in a pigsty (PH) enters the entrance of a passage frame (11) of a piglet sorting device (1), a vision inspection device (15) can visually capture the physical information of the piglet (P). At this time, a predetermined feeding means (F) may be provided in the passage device (10) to induce the entry of the piglet (P). Meanwhile, as described above, the piglet (P) may have a predetermined tag (T), and a tag device (17) capable of recognizing the tag (T) may be provided.

[0200] As illustrated in Fig. 17, if the vision inspection device (15) determines that the shipping pig object (P) located in the passage space (12) within the passage frame (11) has a weight suitable for shipment, the selection display device (16) is operated, and a selection display (S) can be made on the shipping pig object (P) within the space. If the vision inspection device (15) determines that the shipping pig object (P) located in the space within the passage frame (11) has a weight unsuitable for shipment, the selection display device (16) is not operated.

[0201] Next, as illustrated in Fig. 18, the shipping pig object (P) that has passed through the passage device (10) can enter the classification device (20). At this time, the door operating device (27) can be operated to operate the opening / closing door (26) according to the weight of the shipping pig object (P) captured by the vision inspection device (15). That is, an operating signal can be transmitted from the vision inspection device (15) to the door operating device (27).

[0202] For example, if it is determined that the piglet (P) located in the space within the passage frame (11) has a weight suitable for shipment, as illustrated in FIG. 19, the door operating device (27) can rotate the opening / closing door (26) in the direction of the arrow first sorting exit (22). Accordingly, the second sorting exit (23) can be connected to the sorting inlet (21). Accordingly, the piglet (P) can go out through the second sorting exit (23). At this time, according to an example, as illustrated in FIG. 21, the second sorting exit (23) can be connected to a compartment space within the pigsty (PH). The compartment space can be, for example, a space compartmentalized by a predetermined fence (FC). Therefore, the piglet having a weight suitable for shipment can be naturally guided into the compartment space.

[0203] Meanwhile, if it is determined that the piglets (P) located in the space within the passage frame (11) have a weight that is not suitable for shipment, as illustrated in FIG. 20, the door operating device (27) can rotate the opening / closing door (26) toward the second sorting exit (23). Accordingly, the first sorting exit (22) can be made to communicate with the sorting inlet (21). Accordingly, the piglets (P) can go out through the first sorting exit (22). The first sorting exit (22) can be connected to a space within the pigsty (PH). Accordingly, the piglets determined to be unsuitable for shipment can be guided back into the pigsty (PH).

[0204] By repeating the above process, when the selection of all pigs (P) to be shipped within the pigsty (PH) is completed, or when it is confirmed that a certain number of pigs to be shipped are within the pigsty (PH), the pigs marked for selection are guided into a designated shipping space, and the pigs located within the shipping space can be shipped. Meanwhile, as described above, when a separate compartment is provided within the pigsty (PH), pigs (P2) to be shipped that are determined to be suitable for shipment are automatically guided into the compartment, so that selection and shipment can be easily performed.

[0205] At this time, a predetermined guide means may be provided to facilitate the piglets (P) being shipped from the pigsty (PH) to enter the entrance of the passage frame (11) and then exit through the sorting exit of the sorting frame (25). The guide means may be, for example, a predetermined feeding means for providing feed, and the feeding means may be provided in a space within the passage frame (11) or the sorting frame (25). However, the present invention is not limited thereto.

[0206] The pig sorting device (1) according to the embodiment can be easily installed and removed within a pigsty (PH). Therefore, weight measurement and sorting of pigs can be easily performed.

[0207] The pig sorting device (1) according to the embodiment has a relatively small weight, is simply installed in a pigsty (PH), and can sort and mark pigs for shipment simply by passing them through the pig sorting device (1) in the pigsty (PH). Therefore, installation and removal costs are reduced, and user convenience can be improved. In addition, since weight measurement and sorting are performed while the pigs for shipment simply pass through a passage device (10), stress applied to the pigs for shipment can be reduced, and the accuracy of weight measurement can be improved.

[0208] In addition, by including a classification device (20) in addition to the passage device (10), objects suitable for shipment can be naturally guided into the compartment space, thereby improving user convenience. This can be particularly advantageous in the case of a pigsty (PH) with a sufficiently large area and suitable for providing a compartment space.

[0209] According to one embodiment, the shipping money sorting device (1, 2) may further include a biological appearance estimation module (300) and a weight estimation module (400).

[0210] The livestock appearance estimation module (300) is a module that estimates the livestock appearance using acquired image information. The image information can be acquired from a vision inspection device (15). This livestock appearance estimation module (300) may be software embedded in hardware such as a CPU as a predetermined algorithm. Accordingly, a predetermined device equipped with the livestock appearance estimation module (300) and the weight estimation module (400) and a predetermined communication device used for signal exchange may be further provided.

[0211] Fig. 21 is a block diagram of a vision inspection device (15).

[0212] The vision inspection device (15) is a device that can measure the weight of an object located in a predetermined area. The vision inspection device (15) can additionally be equipped with a camera or an image sensor to obtain an image or video of the object. The vision inspection device (15) can transmit the captured image to an estimation device (300) to receive a weight value for the object. The vision inspection device (15) may be a device including a processor (15a), a communication interface (15b), and a memory (15c). The vision inspection device (15) may be implemented with more components than those illustrated in FIG. 21, or with fewer components than those illustrated in FIG. 21. The vision inspection device (15) may be a user terminal such as a laptop, a desktop, a smartphone, a PDA, or the like, may be a server, a network system, or a separate device.

[0213] According to one embodiment of the present disclosure, the processor (15a) typically controls the overall operation of the vision inspection device (15). For example, the processor (15a) can control the components included in the vision inspection device (15) by executing a program stored in the vision inspection device (15).

[0214] The processor (15a) is configured to control the overall operation of the vision inspection device (15). Specifically, the processor (15a) controls the overall operation of the vision inspection device (15) using various programs stored in the storage medium of the vision inspection device (15).

[0215] The processor (15a) can capture images, videos, etc. by a control signal of the estimation device (300) and transmit the captured images, videos, etc. to the estimation device (300). The processor (15a) can generate a signal to be transmitted to at least one of a selection display device, an optical device, a passage device, a classification device, and a tag device. The processor (15a) can receive a weight value and / or appearance information of an object from the estimation device (300). The processor (15a) can generate and transmit a signal to spray a marking material on an object using the selection display device. The processor (15a) can generate and transmit a signal to capture an object using an optical device, and receive captured image information from the optical device. The processor (15a) can generate and transmit a signal to operate an opening / closing door of the classification device, and transmit the signal to a door operating device. The processor (15a) can communicate with the estimation device (300) via a communication interface (15b). The processor (15a) can store captured image information in a memory (200c).

[0216] For example, the processor (15a) may include a CPU, RAM, ROM, and a system bus. Here, the ROM is a configuration in which a set of instructions for system booting is stored, and the CPU copies the stored operating system of the vision inspection device (15) to RAM according to the instructions stored in the ROM and executes the O / S to boot the system. When the system booting is complete, the CPU can copy various stored applications to RAM and execute them to perform various operations. Although the vision inspection device (15) has been described as including only one CPU in the above, it may be implemented with multiple CPUs (or DSPs, SoCs, etc.) during implementation.

[0217] According to one embodiment of the present invention, the processor (15a) may be implemented as a digital signal processor (DSP), a microprocessor, or a time controller (TCON) that processes digital signals. However, the present invention is not limited thereto, and may include one or more of a central processing unit (CPU), a micro controller unit (MCU), a micro processing unit (MPU), a controller, an application processor (AP), a communication processor (CP), or an ARM processor, or may be defined by the corresponding terms. In addition, the processor (15a) may be implemented as a system on chip (SoC) having a built-in processing algorithm, a large scale integration (LSI), or may be implemented in the form of a field programmable gate array (FPGA).

[0218] According to one embodiment of the present disclosure, the communication interface (15b) creates a channel for communicating with an external device such as an estimation device (300) and allows data to be transmitted and received with the external device. The communication interface (15b) may include at least one of a Bluetooth communication interface, a BLE (Bluetooth Low Energy) communication interface, a near field communication interface, a WLAN (Wi-Fi) communication interface, a Zigbee communication interface, an infrared (IrDA, infrared Data Association) communication interface, a WFD (Wi-Fi Direct) communication interface, an UWB (ultra wideband) communication interface, an Ant+ communication interface, and a short-range communication interface, such as a mobile communication network.

[0219] Although the preferred embodiments have been illustrated and described above, the present invention is not limited to the specific embodiments described above, and various modifications may be made by a person skilled in the art to which the invention pertains without departing from the gist of the invention as claimed in the claims. Furthermore, such modifications should not be understood individually from the technical idea or prospect of the present invention.

[0220] 1: Shipment pig sorting device 2: Shipment pig sorting device

[0221] PH: Pig pen SS: Sunshade

[0222] FC: Fence 10: Passage Device

[0223] 11: Passage frame 12: Passage space

[0224] 13: Passage entrance 14: Passage exit

[0225] 15: Vision inspection device 16: Selection display device

[0226] 17: Tag device 20: Classification device

[0227] 25: Classification Frame 21: Classification Entrance

[0228] 22: Exit 1 of the first classification 23: Exit 2 of the second classification

[0229] 24: Connection space 25a: Connection part

[0230] 25b: Classification section 28: Classification space

[0231] 26: Opening / Closing Door 27: Door Operating Device

[0232] 300: Estimation device 330: Biometric appearance estimation module

[0233] 331: Object recognition unit 332: First key point detection unit

[0234] 333: Second Key Point Detection Unit 334: Integrated 3D Model Construction Unit

[0235] 335: Biometric Appearance Estimation Unit 340: Weight Estimation Module

Claims

1. As a livestock weight estimation system, A estimation device comprising one or more processors, a computer-readable recording medium, a memory, and a communication interface, The one or more processors read the recording medium, Obtain image information about the first object in the barn, Using image information about the first object, the appearance information of the first object is output, A livestock weight estimation system that outputs a weight value of the first object using the appearance information of the first object, determines whether the first object is abnormal or normal based on whether the weight value of the first object falls within a preset reference range, and outputs an alarm message including whether the first object is abnormal or normal in one area of ​​image information for the first object.

2. In paragraph 1, The image information for the first object is measured by the image acquisition module, The above image acquisition module, Guide rails that cross the barn in at least one direction, A livestock weight estimation system comprising a movable sensor that is capable of displacing along the above guide rail and obtaining RGB images, depth information, near-infrared images, short-wave infrared images, and thermal imaging biometric images.

3. In paragraph 2, One or more processors of the above estimation device, By inputting the image information into the learned model, the first object is detected and information about the first object is output, A livestock weight estimation system that inputs information about the first object into a detection model to detect a specific body point of the first object, outputs a key point value for the specific body point, and outputs an appearance characteristic value of the first object using the key point value.

4. In paragraph 3, One or more processors of the above estimation device, Merge the RGB image, the depth information, the near-infrared image, the short-wave infrared image, and the thermal imaging biometric image, and output 3D rendering data using the key point value and the appearance characteristic value for each pose of the first object, A livestock weight estimation system that outputs appearance information of a first object from the above three-dimensional rendering data and the appearance characteristic values ​​for each posture of the first object.

5. In paragraph 4, One or more processors of the above estimation device, A livestock weight estimation system that inputs appearance information of the first object using a weight estimation model and outputs a weight value of the first object.

6. In paragraph 5, A livestock weight estimation system, wherein at least one of the learned model, the detection model, and the weight estimation model is a model learned by a machine learning or deep learning method.

7. In paragraph 5, The above weight estimation model is, A livestock weight estimation system that outputs the weight value of the first object using a mathematical formula that uses two input variables among the appearance information of the first object.

8. In paragraph 2, The above congratulatory message is, It comprises a plurality of breeding spaces arranged in parallel along the first direction, The above guide rail extends in a direction parallel to the first direction at least in part, A livestock weight estimation system in which the above movable sensor is displaced along the above guide rail.

9. In paragraph 1, The image information for the above first object is measured by a vision inspection device installed in a shipping money sorting device, The above vision inspection device, A livestock weight estimation system, which is installed in a passage frame including a passage space on the inside and a passage entrance and a passage exit that open the passage space to the outside, generates a signal for measuring the weight of an object that has entered the passage space, captures image information about the first object, and transmits the image information to the estimation device.

10. In paragraph 9, The above vision inspection device, A livestock weight estimation system that receives the weight value of the first object from the estimation device.

11. In paragraph 10, A livestock weight estimation system further comprising a selection marking device installed in the passage frame and spraying a marking material corresponding to the weight value of the first object onto the first object by a signal transmitted from the vision inspection device.

12. In paragraph 10, The above vision inspection device, A livestock weight estimation system that determines whether the weight value of the first object is within a reference weight range, and generates a signal to spray a first indicator substance to the first object if it is within the reference weight range, and transmits the signal to the selection display device to spray a second indicator substance to the first object if it is outside the reference weight range.

13. In paragraph 10, Further comprising a tag device for recognizing a tag displayed on the body of the first object, The above vision inspection device, A livestock weight estimation system that converts the weight value of the first object and information about the tag of the first object into structured data and stores it.

14. In paragraph 12, Further comprising a tag device for recognizing a tag displayed on the body of the first object, The above vision inspection device, A livestock weight estimation system that classifies and stores information about an object to which the first indicator substance has been sprayed and information about an object to which the second indicator substance has been sprayed.

15. In a livestock weight estimation method of a livestock weight estimation system using image information captured through an image acquisition module installed in a livestock barn, A step in which the livestock weight estimation system acquires image information about a first object within the livestock shed using the image acquisition module; A step in which the livestock weight estimation system outputs appearance information of the first object using image information about the first object; A livestock weight estimation method, comprising the steps of: outputting a weight value of a first object using the appearance information of the first object, determining whether the first object is abnormal or normal based on whether the weight value of the first object falls within a preset reference range, and outputting an alarm message including whether the first object is abnormal or normal in one area of ​​image information for the first object.

Citation Information

Patent Citations

  • Keeping system for spread of disease for poultry farm

    KR101918762B1

  • Livestock weight estimation system and method using livestock image

    KR102264281B1

  • Stationary Livestock weight estimation system based on 3D images and Livestock weight estimation method using the same

    KR102404137B1

  • Livestock sorter

    KR102446704B1

  • KR20230039473A