Image acquisition device, rank estimation device, carcass cross-sectional image output device, image acquisition method, rank estimation method, carcass cross-sectional image output method, and program

The image acquisition and rank estimating devices, utilizing ultrasound and machine learning, address the challenge of skilled expertise in marbling evaluation, enabling early and accurate meat quality assessment in live cows, optimizing cattle management and reducing costs.

JP7590719B2Active Publication Date: 2024-11-27NAT UNIV CORP HOKKAIDO NAT UNIV ORG +2
View PDF 16 Cites 0 Cited by

Patent Information

Application Number
JP2021006874
Authority / Receiving Office
JP · JP
Patent Type
Patents
Current Assignee / Owner
Filing Date
2021-01-20
Publication Date
2024-11-27
Estimated Expiration
2041-01-20

AI Technical Summary

Technical Problem

Conventional methods for evaluating beef marbling in live cows require skilled techniques, making it difficult for non-experts to assess meat quality accurately and early.

Method used

An image acquisition device that captures ultrasound images of a cow's torso using specific gain settings, combined with a rank estimating device that utilizes machine learning to predict meat quality based on these images, and a carcass cross-sectional image output device that generates predicted carcass images.

Benefits of technology

Enables early and accurate judgment of meat quality in live cows, allowing for more efficient cattle management and cost reduction by identifying suitable shipping times.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure 0007590719000001
    Figure 0007590719000001
  • Figure 0007590719000002
    Figure 0007590719000002
  • Figure 0007590719000003
    Figure 0007590719000003
Patent Text Reader

Abstract

To solve the problem that it has been difficult to determine meat quality of a living body of a cow at an early stage.SOLUTION: It has been made possible to determine meat quality of a living body of a cow at an early stage by a rank estimation device equipped with: a model acquisition part for acquiring a learning model from a model storage part which stores the learning model acquired by learning processing of machine learning to two or more pieces of teacher data having one or more ultrasonic images acquired by using an ultrasonic wave emitted to a predetermined region of each living body of two or more cows and rank information indicating a rank of meat quality of the cow; an image reception part for receiving one or more ultrasonic images acquired by using an ultrasonic wave emitted to a predetermined region of a living body of one cow; a prediction part for executing prediction processing of the machine learning using the one or more ultrasonic images received by the image reception part and the learning model acquired by the model acquisition part, and acquiring the rank information; and an output part for outputting the rank information acquired by the prediction part.SELECTED DRAWING: Figure 3
Need to check novelty before this filing date? Find Prior Art

Description

[Technical field]

[0001] The present invention relates to a method for estimating the rank of meat quality of a cow using ultrasonic images obtained from a living cow. [Background technology]

[0002] Currently, beef marbling is rated on a scale of 1 to 12, which has a significant impact on the unit price of carcass. Understanding the marbling condition of the live animal will lead to more efficient production from then on. For example, if the marbling condition of a 20-month-old animal is below a certain level, it is expected that the resulting carcass will be below average. Since it costs 500 yen per day to feed Japanese Black cattle, shipping such cattle several months earlier than usual will lead to cost savings.

[0003] On the other hand, the assessment of meat quality by reading the state of fat marbling from ultrasound images became widespread when Fujidaira Kogyo launched "Super Eye Meat" in the 1990s, and has since been introduced in various prefectures (Non-Patent Document 1). [Prior art documents] [Non-patent literature]

[0004] [Non-Patent Document 1] Fukushima Prefecture, "Live meat quality diagnosis was carried out by Super Eye Meat!", [online], [Retrieved January 6, 2021], Internet [URL: https: / / www.pref.fukushima.lg.jp / sec / 36242a / kitakata226.html] Summary of the Invention [Problem to be solved by the invention]

[0005] However, in the past, reading the marbling condition from ultrasound images required skilled techniques, and was not something that just anyone could do. In other words, it was not easy to judge the meat quality of live cattle at an early stage. [Means for solving the problem]

[0006] The image acquisition device of the first invention is an image acquisition device comprising an image acquisition unit that acquires one or more ultrasound images with a gain of approximately 60 to approximately 80 using ultrasound emitted to a specified area of ​​a living cow, and an image storage unit that stores the one or more ultrasound images.

[0007] With this configuration, images can be obtained that enable early assessment of the meat quality of live cattle.

[0008] In addition, the image acquisition device of the second invention is an image acquisition device in which, compared to the first invention, the predetermined area is an area of ​​the trunk within a range of a rear threshold of the front legs of a live cow.

[0009] With this configuration, images can be obtained that enable early assessment of the meat quality of live cattle.

[0010] In addition, the rank estimation device of the third invention is a rank estimation device comprising: a model acquisition unit that acquires a learning model from a model storage unit in which a learning model obtained by a machine learning learning process is stored for two or more teacher data having one or more ultrasound images obtained using ultrasound emitted toward a specified area of ​​each of two or more live cows and rank information indicating the rank of the cow's meat quality; an image receiving unit that accepts one or more ultrasound images obtained using ultrasound emitted toward a specified area of ​​a live cow; a prediction unit that performs a machine learning prediction process and acquires rank information using the one or more ultrasound images accepted by the image receiving unit and the learning model acquired by the model acquisition unit; and an output unit that outputs the rank information acquired by the prediction unit.

[0011] With this configuration, rank information for early assessment of meat quality of live cattle can be obtained.

[0012] In addition, compared to the third invention, the rank estimation device of the fourth invention is a rank estimation device in which the teacher data includes an image set consisting of two or more ultrasound images obtained using ultrasound emitted to two or more locations on a living cow, and rank information.

[0013] With this configuration, more accurate ranking information can be obtained for early assessment of meat quality of live cattle.

[0014] Moreover, in the rank estimation device of the fifth invention, in comparison with the fourth invention, a location identifier for identifying a location is associated with the ultrasound image, and the teacher data also has the location identifier.

[0015] With this configuration, more accurate ranking information can be obtained for early assessment of meat quality of live cattle.

[0016] In addition, the rank estimation device of the sixth invention is a rank estimation device according to any one of the third to fifth inventions, in which the teacher data includes an image set consisting of two or more ultrasound images obtained at two or more gains using ultrasound emitted to a specified area of ​​a living cow, and rank information.

[0017] With this configuration, more accurate ranking information can be obtained for early assessment of meat quality of live cattle.

[0018] Moreover, the rank estimation device of the seventh invention is a rank estimation device according to the sixth invention, in which a gain identifier for identifying a gain is associated with the ultrasound image, and the teacher data also has the gain identifier.

[0019] With this configuration, more accurate ranking information can be obtained for early assessment of meat quality of live cattle.

[0020] Furthermore, the rank estimation device of the eighth invention is the rank estimation device of any one of claims 3 to 7, in which, in contrast to the first or second invention, the ultrasound images are ultrasound images stored by an image acquisition device.

[0021] With this configuration, rank information for early assessment of meat quality of live cattle can be obtained.

[0022] In addition, the carcass cross-section image output device of the ninth invention is a carcass cross-section image output device comprising: a rank acquisition unit that acquires rank information acquired by the rank estimation device of any one of the third to eighth inventions, which is rank information of a single cow; a parameter acquisition unit that acquires one or more cow biological parameters of a single cow; a carcass cross-section image acquisition unit 45 that acquires, using a parameter set having the rank information acquired by the rank acquisition unit and the one or more cow biological parameters acquired by the parameter acquisition unit, one or more carcass cross-section images in a carcass cross-section image storage unit in which one or more carcass cross-section images corresponding to a parameter set having the rank information and the one or more cow biological parameters are stored, and which correspond to a parameter set that satisfies an approximation condition; and a carcass cross-section image output unit 46 that outputs the one or more carcass cross-section images acquired by the carcass cross-section image acquisition unit 45.

[0023] Such a configuration provides future predicted carcass cross section images of the cow.

[0024] In addition, the carcass cross-section image output device of the tenth invention is a carcass cross-section image output device in which, compared to the ninth invention, the parameter acquisition unit uses ultrasound emitted to a specified area of ​​a cow to perform image recognition processing on the acquired ultrasound image and acquire one or more cow biological parameters.

[0025] With this configuration, future predicted carcass cross-sectional images of a cow can be easily obtained. Effect of the Invention

[0026] The image acquisition device according to the present invention makes it possible to judge the meat quality of live cattle at an early stage. [Brief description of the drawings]

[0027] [Figure 1] Configuration diagram of cow information processing system A in embodiment 1 [Diagram 2] Block diagram of the cow information processing system A [Diagram 3] Block diagram of the rank estimation device 3 [Figure 4] Block diagram of the carcass cross-sectional image output device 4 [Diagram 5] A flowchart illustrating an example of the operation of the image acquisition device 1. [Figure 6] A flowchart for explaining an example of the operation of the rank estimation device 3. [Figure 7] A flowchart for explaining an example of the operation of the carcass cross-sectional image output device 4 [Figure 8] A flowchart illustrating an example of a rank acquisition process. [Figure 9] A flowchart explaining a first example of the cow's biological parameter acquisition process. [Figure 10] A flowchart explaining a second example of the cow's biological parameter acquisition process. [Figure 11] FIG. 2 is a diagram illustrating an area where the ultrasound image is acquired. [Figure 12] FIG. 13 shows an example of the ultrasound image. [Figure 13] FIG. 13 shows an example of the ultrasound image. [Figure 14] A diagram showing the teacher data management table [Figure 15] A diagram showing the concept of the BMS number prediction process [Figure 16] FIG. 13 shows a carcass cross-sectional image of the carcass cross-sectional image storage unit 41. [Figure 17] FIG. 13 shows an example of the ultrasound image. [Figure 18] A diagram showing an example of the output [Figure 19] Overview of the computer system [Figure 20] Block diagram of the computer system DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0028] Hereinafter, embodiments of an image acquisition device and the like will be described with reference to the drawings. Note that components with the same reference numerals in the embodiments perform similar operations, and therefore repeated description may be omitted.

[0029] (Embodiment 1) In this embodiment, an image acquisition device that acquires and stores ultrasonic images from a live cow with a gain between 60 dB and 80 dB will be described. The area from which ultrasonic images are acquired is preferably the trunk of the live cow within a threshold range behind the front legs.

[0030] In this embodiment, a learning device that uses one or more ultrasound images to construct a learning model used to estimate rank information will be described.

[0031] In addition, in this embodiment, a rank estimation device that estimates rank information using a learning model constructed by learning two or more teacher data having one or more ultrasound images and rank information, and one or more ultrasound images, will be described.

[0032] Furthermore, in this embodiment, a carcass cross-sectional image output device will be described which acquires and outputs one or more carcass cross-sectional images using rank information and one or more cattle biological parameters such as the thickness of the trapezius muscle, the thickness of the subcutaneous fat, and the thickness of the latissimus dorsi muscle.

[0033] Fig. 1 is a configuration diagram of a cattle information processing system A in this embodiment. The cattle information processing system A comprises an image acquisition device 1, a learning device 2, a rank estimation device 3, and a carcass cross-sectional image output device 4. Two or more devices constituting the cattle information processing system A may be integrated. For example, the learning device 2, the rank estimation device 3, and the carcass cross-sectional image output device 4 may be integrated into one device.

[0034] Furthermore, the image acquisition device 1, the learning device 2, the rank estimation device 3, and the carcass cross-section image output device 4 may be capable of communicating with each other via a network such as the Internet.

[0035] Fig. 2 is a block diagram of the cattle information processing system A in this embodiment. Fig. 3 is a block diagram of the rank estimation device 3. Fig. 4 is a block diagram of the carcass cross-section image output device 4.

[0036] The image acquisition device 1 includes an image storage unit 11, a reception unit 12, an image acquisition unit 13, and an image accumulation unit .

[0037] The learning device 2 includes a learning storage unit 21, a learning reception unit 22, and a learning processing unit 23. The learning storage unit 21 includes a teacher data storage unit 211 and a model storage unit 212. The learning processing unit 23 includes a learning unit 231 and a learning accumulation unit 232.

[0038] The rank estimation device 3 includes a storage unit 31, a reception unit 32, a processing unit 33, and an output unit 34. The storage unit 31 includes a model storage unit 212 and an image storage unit 311. The reception unit 32 includes an image reception unit 321. The processing unit 33 includes a model acquisition unit 331 and a prediction unit 332.

[0039] The carcass cross-sectional image output device 4 comprises a carcass cross-sectional image storage section 41, a reception section 42, a rank acquisition section 43, a parameter acquisition section 44, a carcass cross-sectional image acquisition section 45, and a carcass cross-sectional image output section 46.

[0040] One or more ultrasound images are stored in the image storage unit 11 constituting the image acquisition device 1. It is preferable that an ultrasound image is associated with a location identifier that identifies a location to which ultrasound is applied. It is preferable that an ultrasound image is associated with a gain identifier that identifies a gain.

[0041] The reception unit 12 receives various instructions and information. The various instructions and information are an image acquisition instruction, a gain identifier, and a location identifier. The image acquisition instruction is an instruction to acquire an ultrasound image. The image acquisition instruction may include a gain identifier and a location identifier.

[0042] The image acquisition unit 13 acquires one or more ultrasound images using ultrasound emitted to a predetermined area of ​​the living cow. The image acquisition unit 13 is a known technique, and therefore a detailed description thereof will be omitted.

[0043] It is preferable that the image acquisition unit 13 acquires one or more ultrasound images with a gain between about 60 and about 80 using ultrasound emitted to a predetermined region of the living cow.

[0044] It is preferable that the image acquisition unit 13 acquires ultrasound images of two or more locations on the cow. For example, the image acquisition unit 13 acquires ultrasound images of four locations on the cow.

[0045] It is preferable that the image acquiring unit 13 acquires two or more ultrasound images using two or more gains. For example, the image acquiring unit 13 acquires ultrasound images with two levels of gain.

[0046] For example, the image acquisition unit 13 acquires eight ultrasound images for each of four locations on the cow using two levels of gain.

[0047] The predetermined area is preferably an area of ​​the trunk within a threshold range of the rear of the front legs of the live cow. The predetermined area is an area of ​​the trunk within a threshold range of the rear of the shoulder blades of the live cow.

[0048] The image acquisition unit 13 acquires an ultrasound image for each of one or more locations with a gain of a fixed value between about 60 dB and about 80 dB (e.g., 74 dB), and then continuously lowers the gain automatically or in response to a user's instruction, and captures the image when the image becomes clear, thereby acquiring an ultrasound image. Note that the gain here is information indicating the degree of amplification when the maximum sensitivity is expressed as 100 dB for convenience. In other words, the gain here is information indicating the normalized degree of amplification with a maximum of 100. Also, the determination that the image has become clear may be made manually or automatically by image processing.

[0049] The image storage unit 14 stores one or more ultrasound images acquired by the image acquisition unit 13. It is preferable that the image storage unit 14 stores one or more ultrasound images acquired by the image acquisition unit 13 in pairs with a gain identifier that identifies the gain used when acquiring the ultrasound image. It is preferable that the image storage unit 14 stores one or more ultrasound images acquired by the image acquisition unit 13 in pairs with a location identifier that identifies the location to which the ultrasound was applied. The location identifier is usually input by the user.

[0050] The image accumulation unit 14 normally accumulates one or more ultrasound images acquired by the image acquisition unit 13 in the image storage unit 11. However, the image accumulation unit 14 may accumulate one or more ultrasound images in an external device.

[0051] Various types of information are stored in the learning storage unit 21 constituting the learning device 2. The various types of information are, for example, teacher data (to be described later) and a learning model (to be described later).

[0052] The teacher data storage unit 211 stores two or more teacher data. The teacher data has, for example, one or more ultrasound images and rank information. The teacher data may be only one ultrasound image and rank information. The teacher data has, for example, one or more feature amounts acquired from each of one or more ultrasound images and rank information. The teacher data has, for example, one or more feature amounts acquired from each of one or more ultrasound images, one or more ultrasound images, and rank information. The feature amounts are, for example, cow biological parameters described later. The feature amounts may be, for example, known feature amounts of images.

[0053] It is preferable that the ultrasound image of the training data is associated with a location identifier that identifies a location of the target of the ultrasound image. The location is a location of the live cow and is a location from which the ultrasound is emitted. The location is a location of the observation target. The location identifier may be associated with the training data, which may include a location identifier.

[0054] It is preferable that an ultrasound image in the training data is associated with a gain identifier that identifies the gain used to obtain the ultrasound image. It is also preferable that the gain is between about 60 and about 80. The gain is, for example, 74, 71, 67, or 66 (in dB). Associating a gain identifier with an ultrasound image may mean that the training data has a gain identifier.

[0055] The training data may have an image set which is a set of two or more ultrasound images. The two or more ultrasound images constituting the image set are preferably two or more ultrasound images obtained using ultrasound emitted to two or more locations of the living cow. The two or more ultrasound images constituting the image set are preferably two or more ultrasound images obtained at two or more gains using ultrasound emitted to a predetermined area of ​​the living cow.

[0056] The two or more ultrasound images constituting the image set are, for example, ultrasound images of four locations in a predetermined region of the cow, and are ultrasound images acquired at two levels of gain (for example, 74, 66) for each location. In this case, the number of ultrasound images constituting the image set is eight.

[0057] The rank information is information that indicates the rank of the beef quality. The rank information is, for example, a BMS number. However, the rank information may be, for example, a grade. The rank information is information that will be used in the future, and may be information that indicates the rank.

[0058] The model storage unit 212 stores a learning model. The learning model may be called a learning device, a classifier, or the like. The learning model is information used in machine learning prediction processing together with one or more ultrasound images, and is used to output rank information. The learning model is, for example, information acquired by the learning unit 231 described later.

[0059] The learning receiving unit 22 receives various instructions, information, etc. The various instructions, information, etc. are, for example, a learning instruction, teacher data, etc. A learning instruction is an instruction to start learning.

[0060] The means for inputting various instructions and information may be any means, such as a touch panel, a keyboard, a mouse, or a menu screen.

[0061] The learning processing unit 23 performs various types of processing. The various types of processing are, for example, processing performed by a learning unit 231 and a learning accumulation unit 232.

[0062] The learning unit 231 performs a machine learning learning process. The learning unit 231 performs a machine learning learning process on two or more pieces of teacher data in the teacher data storage unit 211, and acquires a learning model. More specifically, the learning unit 231 acquires two or more pieces of teacher data from the teacher data storage unit 211, provides the two or more pieces of teacher data to a module that performs a machine learning learning process, executes the module, and acquires a learning model.

[0063] The learning unit 231 may acquire one or more cow biological parameters from each ultrasound image in the teacher data before the learning process. In this case, the learning unit 231 may perform a machine learning learning process on two or more teacher data having one or more cow biological parameters, an ultrasound image, and rank information, to acquire a learning model. The process of acquiring one or more cow biological parameters from an ultrasound image is a process performed by the parameter acquisition unit 44 described later.

[0064] The learning unit 231 may acquire one or more feature amounts of an image from each ultrasound image included in the teacher data before the learning process. In this case, the learning unit 231 may perform a machine learning learning process on two or more teacher data having one or more feature amounts, an ultrasound image, and rank information, for example, to acquire a learning model. In this case, the learning unit 231 may perform a machine learning learning process on two or more teacher data having one or more feature amounts and rank information, for example, to acquire a learning model.

[0065] Any machine learning algorithm can be used here. Although deep learning is preferable as an algorithm for performing machine learning learning processing, any type of machine learning can be used, for example, decision tree, random forest, SVM, SVR, etc. Also, the learning model is a model for inputting one or more ultrasound images and outputting rank information.

[0066] In addition, modules that perform machine learning learning processes include, for example, various functions in the TensorFlow library, the random forest module of the R language, tinySVM, etc.

[0067] The learning accumulation unit 232 accumulates the learning model acquired by the learning unit 231. The learning accumulation unit 232 usually accumulates the learning model acquired by the learning unit 231 in the model storage unit 212.

[0068] Various types of information are stored in the storage unit 31 constituting the rank estimation device 3. The various types of information are, for example, a learning model and one or more ultrasound images. It is preferable that the learning model here is a model acquired by the learning device 2.

[0069] One or more ultrasound images are stored in the image storage unit 311. Typically, one or more ultrasound images of a single live cow are stored in the image storage unit 311. The one or more ultrasound images are ultrasound images of the target cow for which rank information is to be obtained.

[0070] One or more ultrasound images associated with a biometric identifier may be stored in the image storage unit 311. In other words, the image storage unit 311 may store two or more ultrasound images of each cow.

[0071] The receiving unit 32 receives various instructions and information, such as a prediction instruction and one or more ultrasound images.

[0072] The means for inputting various instructions and information may be any means, such as a touch panel, a keyboard, a mouse, or a menu screen.

[0073] The image receiving unit 321 receives one or more ultrasound images. The ultrasound images are images obtained by using ultrasound emitted to a predetermined area of ​​a living cow. The one or more ultrasound images received by the image receiving unit 321 are, for example, ultrasound images acquired by the image acquisition device 1, and are stored in the image storage unit 311 at least temporarily.

[0074] The processing unit 33 performs various types of processing. The various types of processing are, for example, processing performed by a model acquisition unit 331 and a prediction unit 332.

[0075] The model acquisition unit 331 acquires the learning model from the model storage unit 212. Note that the model storage unit 212 may exist in an external device (not shown) other than the rank estimation device 3.

[0076] The prediction unit 332 performs machine learning prediction processing using one or more ultrasound images accepted by the image accepting unit 321 and the learning model acquired by the model acquiring unit 331, and acquires rank information.

[0077] The prediction unit 332 may acquire one or more cow biological parameters from each of the one or more ultrasound images accepted by the image accepting unit 321. The prediction unit 332 may then perform machine learning prediction processing to acquire rank information, for example, using the one or more ultrasound images accepted by the image accepting unit 321, the one or more cow biological parameters, and the learning model acquired by the model acquiring unit 331. The prediction unit 332 may also perform machine learning prediction processing to acquire rank information, for example, using the acquired one or more cow biological parameters and the learning model acquired by the model acquiring unit 331.

[0078] The prediction unit 332 may acquire one or more feature amounts of an image from each of one or more ultrasound images accepted by the image accepting unit 321. Then, the prediction unit 332 may perform a machine learning prediction process using the acquired one or more feature amounts and the learning model acquired by the model acquiring unit 331, and acquire rank information.

[0079] More specifically, the prediction unit 332 provides one or more ultrasound images accepted by the image accepting unit 321 and the learning model acquired by the model acquiring unit 331 to a module that performs machine learning prediction processing, executes the module, and acquires rank information that is the prediction result.

[0080] As described above, the machine learning algorithm is not limited to this. Although deep learning is preferable as the algorithm for performing machine learning prediction processing, other types of machine learning are also acceptable, such as decision trees, random forests, SVR, and SVM.

[0081] Modules that perform machine learning prediction processing include, for example, various functions in the TensorFlow library, the random forest module in the R language, tinySVM, etc.

[0082] The output unit 34 outputs the rank information acquired by the prediction unit 332. Here, output is a concept including display on a display, projection using a projector, printing on a printer, sound output, transmission to an external device, storage in a recording medium, delivery of processing results to other processing devices or other programs, etc.

[0083] Various information is stored in the carcass cross-sectional image storage unit 41 constituting the carcass cross-sectional image output device 4. The various information is, for example, one or more carcass cross-sectional images. The carcass cross-sectional image is associated with a parameter set having rank information and one or more cattle biological parameters. The carcass cross-sectional image storage unit 41 may be present in an external device other than the carcass cross-sectional image output device 4.

[0084] The carcass cross-section image storage unit 41 may store one or more ultrasound images of a cow from which a predicted carcass cross-section image is to be obtained, rank information of the cow, or one or more cow biological parameters of the cow.

[0085] The reception unit 42 receives various instructions and information. The various instructions and information are, for example, a carcass cross-sectional image output instruction and one or more cattle biological parameters. The carcass cross-sectional image output instruction is an instruction to output a carcass cross-sectional image.

[0086] The rank acquisition unit 43 acquires the rank information acquired by the rank estimation device 3. The rank information may be stored in the carcass cross-section image storage unit 41. The rank acquisition unit 43 may be the rank estimation device 3.

[0087] The parameter acquisition unit 44 acquires one or more cow biological parameters of one cow. The one or more cow biological parameters are, for example, the thickness of the trapezius muscle, the thickness of the subcutaneous fat directly above the trapezius muscle, the thickness of the subcutaneous fat, the thickness of the rib cage, the thickness of the latissimus dorsi muscle, and the presence or absence of a lump.

[0088] The parameter acquisition unit 44 performs image processing on one or more ultrasound images of a single cow, for example, to acquire one or more cow biological parameters. An example of the image processing will be described in detail later.

[0089] The parameter acquisition unit 44 may acquire, for example, one or more manually inputted cow biological parameters. The one or more manually inputted cow biological parameters are, for example, parameters acquired by a person by inspecting an ultrasound image. The one or more manually inputted cow biological parameters are information accepted by the acceptance unit 42.

[0090] The carcass cross-section image acquisition unit 45 uses a parameter set having the rank information acquired by the rank acquisition unit 43 and one or more cattle biological parameters acquired by the parameter acquisition unit 44 to acquire one or more carcass cross-section images in the carcass cross-section image storage unit 41, which are one or more carcass cross-section images and correspond to a parameter set that satisfies an approximation condition. The approximation condition is a condition for determining that two parameter sets are approximate. For example, the approximation condition is that the similarity between the two parameter sets is equal to or greater than a threshold. For example, the approximation condition is that the similarity between the two parameter sets is maximum. For example, the approximation condition is that the ranking of the similarity between the two parameter sets is within a threshold or less than a threshold. For example, the approximation condition is that it is determined that they are approximate by a machine learning prediction process.

[0091] The carcass cross-sectional image output section 46 outputs one or more carcass cross-sectional images acquired by the carcass cross-sectional image acquisition section 45.

[0092] The carcass cross-sectional image output unit 46 or an output unit not shown may output one or more cattle biological parameters acquired by the parameter acquisition unit 44.

[0093] Here, output is a concept that includes display on a display, projection using a projector, printing on a printer, sound output, transmission to an external device, storage on a recording medium, and handing over the processing results to other processing devices or other programs, etc.

[0094] The image storage unit 11, learning storage unit 21, teacher data storage unit 211, model storage unit 212, storage unit 31, model storage unit 212, image storage unit 311, and carcass cross-section image storage unit 41 are preferably non-volatile recording media, but can also be realized using volatile recording media.

[0095] There is no restriction on the process by which information is stored in the image storage unit 11 etc. For example, information may be stored in the image storage unit 11 etc. via a recording medium, information transmitted via a communication line etc. may be stored in the image storage unit 11 etc., or information inputted via an input device may be stored in the image storage unit 11 etc.

[0096] The image acquisition unit 13 can use an ultrasonic generator (for example, "DIAGNOSTIC SCANNER HS-1600V").

[0097] The image storage unit 14, the learning processing unit 23, the learning unit 231, the learning storage unit 232, the processing unit 33, the model acquisition unit 331, the prediction unit 332, the rank acquisition unit 43, the parameter acquisition unit 44, and the carcass cross-sectional image acquisition unit 45 can usually be realized by a processor, a memory, etc. The processing procedure of the image storage unit 14, etc. is usually realized by software, and the software is recorded in a recording medium such as a ROM. However, it may be realized by hardware (dedicated circuit). The processor is, for example, a CPU, an MPU, a GPU, etc., and the type is not important.

[0098] Receiving unit 12, learning receiving unit 22, receiving unit 32, image receiving unit 321, and receiving unit 42 can be realized by a device driver for an input means such as a touch panel or a keyboard, control software for a menu screen, or the like.

[0099] The output unit 34 and the carcass cross-sectional image output unit 46 may or may not include an output device such as a display or a speaker. The output unit 34 and the like may be realized by driver software for an output device, or a combination of driver software for an output device and an output device, etc.

[0100] Next, a description will be given of an example of the operation of the cow information processing system A. First, a description will be given of an example of the operation of the image acquisition device 1 with reference to the flowchart of FIG.

[0101] (Step S501) The reception unit 12 of the image acquisition device 1 judges whether or not an image acquisition instruction has been received. If an image acquisition instruction has been received, the process proceeds to step S502. If an image acquisition instruction has not been received, the process proceeds to step S506. Note that the image acquisition instruction is an instruction to emit ultrasound and acquire an ultrasound image.

[0102] (Step S502) The reception unit 12 judges whether or not a location identifier has been received. If a location identifier has been received, the process proceeds to step S503, and if a location identifier has not been received, the process returns to step S502.

[0103] (Step S503) The image acquiring unit 13 acquires the gain identifier of the image storage unit 11.

[0104] (Step S504) The image acquiring unit 13 emits ultrasonic waves and acquires an ultrasonic image corresponding to the gain indicated by the gain identifier.

[0105] (Step S505) The image accumulation unit 14 accumulates the ultrasound image acquired in step S504 in the image storage unit 11 in association with the location identifier accepted in step S502 and the gain identifier acquired in step S503. Return to step S501.

[0106] (Step S506) The acceptance unit 12 judges whether or not a gain setting has been accepted. If a gain setting has been accepted, the process proceeds to step S507, and if a gain setting has not been accepted, the process returns to step S501.

[0107] (Step S507) The image acquisition unit 13 acquires a gain identifier corresponding to the setting accepted in step S506. The image accumulation unit 14 accumulates the gain identifier in the image storage unit 11. The process returns to step S501.

[0108] In the flowchart of FIG. 5, the process ends when the power is turned off or an interrupt occurs to end the process.

[0109] Next, an example of the operation of the learning device 2 will be described. When the learning reception unit 22 of the learning device 2 receives input of teacher data, the learning processing unit 23 accumulates the teacher data in the teacher data storage unit 211. Then, after two or more pieces of teacher data are stored, the learning reception unit 22 receives a learning instruction. Next, the learning unit 231 reads two or more pieces of teacher data from the teacher data storage unit 211, provides the two or more pieces of teacher data to a module that performs machine learning learning processing, executes the module, and acquires a learning model. Next, the learning accumulation unit 232 accumulates the learning model in the model storage unit 212.

[0110] Next, an example of the operation of the rank estimation device 3 will be described with reference to the flowchart of FIG.

[0111] (Step S601) The image receiving unit 321 determines whether or not one or more ultrasound images of a live cow for which a rank is to be determined have been received. If one or more ultrasound images have been received, the process proceeds to step S602; if one or more ultrasound images have not been received, the process proceeds to step S603. It is preferable that the one or more ultrasound images received by the image receiving unit 321 are ultrasound images acquired by the image acquisition device 1. It is also preferable that each ultrasound image is associated with a gain identifier and a location identifier.

[0112] (Step S602) The processing unit 33 accumulates the one or more ultrasound images received in step S601 in the image storage unit 311. The process returns to step S601.

[0113] It is preferable that the ultrasound images stored in the image storage unit 311 are associated with a gain identifier and a location identifier.

[0114] (Step S603) The reception unit 32 judges whether or not a prediction instruction has been received. If a prediction instruction has been received, the process proceeds to step S604, and if a prediction instruction has not been received, the process returns to step S601.

[0115] (Step S604) The model acquisition unit 331 acquires the learning model from the model storage unit 212.

[0116] (Step S605) The prediction unit 332 acquires one or more ultrasound images from the image storage unit 311.

[0117] (Step S606) The prediction unit 332 provides the learning model acquired in step S604 and one or more ultrasound images acquired in step S605 to a module that performs machine learning prediction processing, executes the module, and acquires rank information.

[0118] (Step S607) The output unit 34 outputs the rank information acquired in step S606. The process returns to step S601.

[0119] In the flowchart of FIG. 6, the process ends when the power is turned off or an interrupt occurs to end the process.

[0120] Next, an example of the operation of the carcass cross-sectional image output device 4 will be described with reference to the flow chart of FIG.

[0121] (Step S701) The reception unit 42 judges whether or not a carcass cross-sectional image output instruction has been received. If a carcass cross-sectional image output instruction has been received, the process proceeds to step S702, and if a carcass cross-sectional image output instruction has not been received, the process returns to step S701.

[0122] (Step S702) The rank acquisition unit 43 acquires rank information of the target live cattle. An example of the rank acquisition process for acquiring rank information will be described with reference to the flowchart of Fig. 8. Here, the rank acquisition unit 43 may perform a process for reading out the rank information stored in the carcass cross-section image storage unit 41.

[0123] (Step S703) The parameter acquisition unit 44 acquires one or more bovine biological parameters of the target bovine living body. An example of such bovine biological parameter acquisition processing will be described with reference to the flowcharts of Figs.

[0124] Here, the parameter acquisition unit 44 may perform a process of reading out one or more cattle biological parameters stored in the carcass cross-section image storage unit 41. The parameter acquisition unit 44 may also acquire some of the multiple cattle biological parameters by a cattle biological parameter acquisition process described in the flowcharts of Figures 9 and 10, and read out other cattle biological parameters from the carcass cross-section image storage unit 41. The cattle biological parameters in the carcass cross-section image storage unit 41 are, for example, cattle biological parameters input by the user.

[0125] (Step S704) The carcass cross-sectional image acquisition unit 45 constructs a vector having as its elements the rank information acquired in step S702 and one or more of the cattle biological parameters acquired in step S703.

[0126] (Step S705) The carcass cross-sectional image acquisition unit 45 assigns 1 to the counter i.

[0127] (Step S706) The carcass cross-sectional image acquisition unit 45 judges whether or not the i-th carcass cross-sectional image exists in the carcass cross-sectional image storage unit 41. If the i-th carcass cross-sectional image exists, the process proceeds to step S707, and if not, the process proceeds to step S710.

[0128] (Step S707) The carcass cross-sectional image acquisition unit 45 acquires from the carcass cross-sectional image storage unit 41 a vector paired with the i-th carcass cross-sectional image.

[0129] (Step S708) The carcass cross-section image acquisition unit 45 calculates the similarity between the vector constructed in step S704 and the vector acquired in step S707. Note that the similarity between two vectors is usually the distance (cosine value) between the two vectors, but the similarity may be acquired using another algorithm. An example of another algorithm is to calculate the proportion of matching elements between the two vectors as the similarity. Also, since the technology for calculating the similarity between two vectors is a publicly known technology, detailed explanation will be omitted.

[0130] (Step S709) The carcass cross-sectional image acquisition unit 45 increments the counter i by 1. The process returns to step S706.

[0131] (Step S710) The carcass cross-sectional image acquisition unit 45 determines one or more vectors whose similarity calculated in step S708 satisfies an approximation condition. The approximation condition is, for example, "the similarity is maximum," "the similarity is in the top N (N is a natural number)," or "the similarity is equal to or greater than a threshold."

[0132] (Step S711) The carcass cross-sectional image acquisition unit 45 acquires from the carcass cross-sectional image storage unit 41 a carcass cross-sectional image paired with each of the one or more vectors determined in step S710.

[0133] (Step S712) The carcass cross-sectional image output unit 46 outputs one or more carcass cross-sectional images acquired in step S711. Return to step S701.

[0134] In the flowchart of FIG. 7, the process ends when the power is turned off or an interrupt is issued to end the process.

[0135] Next, an example of the rank acquisition process in step S702 will be described with reference to the flowchart in Fig. 8. The example of the rank acquisition process in step S702 is a process in which rank acquisition unit 43 performs the processes in steps S604, S605, and S606 in Fig. 6 to acquire rank information.

[0136] 8, it is assumed that the learning model is stored in the carcass cross-section image storage unit 41. It is also assumed that the rank acquisition unit 43 performs the same processing as the prediction unit 332.

[0137] Next, a first example of the cow biological parameter acquisition process in step S703 will be described with reference to the flowchart in FIG.

[0138] (Step S901) The parameter acquiring unit 44 assigns 1 to a counter i.

[0139] (Step S902) The parameter acquisition unit 44 judges whether or not to acquire the i-th parameter. If the i-th parameter is to be acquired, the process proceeds to step S903, and if the i-th parameter is not to be acquired, the process returns to the upper processing. The parameters to be acquired are determined in advance.

[0140] (Step S903) The parameter acquisition unit 44 acquires from the carcass cross-section image storage unit 41 an ultrasonic image paired with the location identifier corresponding to the i-th parameter.

[0141] (Step S904) The parameter acquisition unit 44 determines whether to acquire the i-th parameter by image processing or by machine learning processing. If the parameter is acquired by image processing, the process proceeds to step S905. If the parameter is acquired by machine learning processing, the process proceeds to step S907. The method of acquiring the i-th parameter is determined in advance.

[0142] (Step S905) The parameter acquisition unit 44 acquires a boundary line for acquiring the i-th parameter by performing a contour extraction process on the ultrasound image acquired in step S903. For example, when the i-th parameter is the "thickness of the trapezius muscle", the parameter acquisition unit 44 determines two boundary lines, an upper boundary and a lower boundary of the trapezius muscle.

[0143] (Step S906) The parameter acquisition unit 44 acquires the thickness (i-th parameter) corresponding to the distance between the boundaries for acquiring the i-th parameter, associates the thickness with the identifier of the i-th parameter, and temporarily stores the thickness in a buffer (not shown). Then, the process proceeds to step S909.

[0144] (Step S907) The parameter acquisition unit 44 acquires a learning model for acquiring the i-th parameter from the carcass cross-section image storage unit 41.

[0145] (Step S908) The parameter acquisition unit 44 provides the ultrasound image acquired in step S903 and the learning model acquired in step S908 to a module that performs machine learning prediction processing, executes the module, acquires the i-th parameter, associates it with the identifier of the i-th parameter, and temporarily stores it in a buffer (not shown). Note that the i-th parameter is, for example, "presence or absence of a lump." Furthermore, the learning model in this case is information acquired by performing a learning process using one or more images (positive examples) of "presence of a lump" and one or more images (negative examples) of "no lump" as teacher data.

[0146] (Step S909) The parameter acquiring unit 44 increments the counter i by 1. The process returns to step S902.

[0147] Next, a second example of the cow biological parameter acquisition process in step S703 will be described with reference to the flowchart of Fig. 10. The second example is a more specific version of the first example.

[0148] (Step S1001) The parameter acquisition unit 44 acquires an ultrasound image corresponding to “location identifier=trapezius muscle, etc.” from the carcass cross-sectional image storage unit 41.

[0149] (Step S1002) The parameter acquisition unit 44 determines the boundary between the skin and subcutaneous fat, and the boundary between the trapezius muscle, by performing a contour extraction process on the ultrasound image acquired in step S1001. Note that the contour extraction process for determining the boundary of an image is a known technique, and therefore a detailed description thereof will be omitted.

[0150] (Step S1003) The parameter acquisition unit 44 acquires the "thickness of the subcutaneous fat directly above the trapezius muscle" using the information on the boundary between the skin and the subcutaneous fat determined in step S1002. Note that the parameter acquisition unit 44 acquires, for example, a representative value (for example, maximum value, average value, or corner thickness) of the thicknesses of the subcutaneous fat at multiple locations as the "thickness of the subcutaneous fat directly above the trapezius muscle".

[0151] (Step S1004) The parameter acquiring unit 44 acquires the "trapezius muscle thickness" which is the distance between the boundary lines of the trapezius muscle determined in step S1002. Note that the parameter acquiring unit 44 acquires, for example, a representative value (for example, maximum value, average value, corner thickness) of a plurality of distances between the boundary lines as the "trapezius muscle thickness".

[0152] (Step S1005) The parameter acquiring unit 44 acquires a set of pixels in a predetermined region below the trapezius muscle.

[0153] (Step S1006) Parameter acquisition unit 44 judges whether the set of pixels acquired in step S1005 meets the lump condition. If the lump condition is met, proceed to step S1007, and if the lump condition is not met, proceed to step S1008. Note that the lump condition is a condition for judging whether or not there is a lump, and in this case, a condition for determining "lump = present." An example of a lump condition is that a representative value related to pixel values ​​(e.g., average pixel value, number of pixel values ​​equal to or greater than a threshold, percentage of pixel values ​​equal to or greater than a threshold) is equal to or greater than a threshold.

[0154] (Step S1007) The parameter acquiring unit 44 acquires the parameter "lump = present." Then, the process proceeds to step S1009.

[0155] (Step S1008) The parameter acquiring unit 44 acquires the parameter "lump=none".

[0156] (Step S1009) The parameter acquisition unit 44 acquires an ultrasound image corresponding to “location identifier=subcutaneous fat, etc.” from the carcass cross-sectional image storage unit 41.

[0157] (Step S1010) The parameter acquiring unit 44 determines the boundary line between the subcutaneous fat and the latissimus dorsi muscle by performing a contour extraction process on the ultrasound image acquired in step S1009.

[0158] (Step S1011) The parameter acquiring unit 44 acquires the thickness of the subcutaneous fat by using the information of the boundary line determined in step S1010.

[0159] (Step S1012) The parameter acquisition unit 44 determines the boundary line between the latissimus dorsi muscle and the internal fat by performing a contour extraction process on the ultrasound image acquired in step S1009.

[0160] (Step S1013) The parameter acquisition unit 44 acquires the thickness of the latissimus dorsi muscle by using the information of the boundary line determined in step S1012.

[0161] (Step S1014) The parameter acquisition unit 44 determines the boundary line between the internal fat and the ribs by performing a contour extraction process on the ultrasound image acquired in step S1009.

[0162] (Step S1015) The parameter acquisition unit 44 acquires the variation thickness using the information on the boundary line determined in step S1014, and returns to the upper process.

[0163] A specific example of the operation of the cow information processing system A in this embodiment will be described below. A conceptual diagram of the cow information processing system A is shown in Figure 1.

[0164] Now, let us say that a user uses the image acquisition device 1 to apply generated ultrasound to an area 1102 of the torso immediately behind an area 1101 of the scapula of a live cow within a threshold range, and acquire an ultrasound image (see FIG. 11). It should be noted that the image acquisition device 1 acquires eight ultrasound images for one cow at four locations in the area 1102, each with two levels of gain, and stores the images in the image storage unit 11. It should be noted that the "DIAGNOSTIC SCANNER HS-1600V" (URL: https: / / www.honda-el.co.jp / medical / hs1600v.html) can be used as the image acquisition device 1.

[0165] Moreover, one of the two stages of gain used by the image acquisition device 1 is a fixed value (for example, "74" here). The other of the two stages of gain used by the image acquisition device 1 is the gain at which the image becomes clear when the gain is successively lowered. If there is a lot of marbling, the image will be clear when the gain is low (for example, "66") (see FIG. 12). On the other hand, if there is little marbling, the image will be clear even when the gain is high (for example, "71") (see FIG. 13). In other words, the gain adopted here differs depending on the cow. FIGS. 12 and 13 are examples of ultrasound images acquired to create training data used by the learning device 2.

[0166] It is also assumed that the image acquisition device 1 has stored eight ultrasound images for one cow in the image storage unit 11, each paired with a gain identifier and a location identifier.

[0167] Furthermore, it is assumed that the image acquisition device 1 stores eight ultrasound images for a large number of cows in the image storage unit 11 through the above-mentioned processing.

[0168] Next, it is assumed that the user associates the eight ultrasound images of each cow with a BMS number and stores the images in the teacher data storage unit 211 of the learning device 2. It is also assumed that the teacher data management table shown in Fig. 14 is stored in the teacher data storage unit 211.

[0169] The teacher data management table manages a large number of teacher data. The teacher data management table has eight ultrasound images, numbered "1" to "8," and a BMS number for each cow. Each ultrasound image is associated with a "gain identifier" and a "location identifier." The "location identifier" is one of "1," "2," "3," and "4." The location identifier "1," for example, is the location identifier "trapezius muscle, etc.", and is information that identifies the location where the thickness of the trapezius muscle, the thickness of the subcutaneous fat directly above the trapezius muscle, etc. can be obtained. "2," for example, is the location identifier "subcutaneous fat, etc.", and is information that identifies the location where the thickness of the subcutaneous fat, the thickness of the latissimus dorsi muscle, and the thickness of the rib cage can be obtained.

[0170] Next, it is assumed that the user inputs a learning instruction to the learning device 2. Then, the learning receiving unit 22 of the learning device 2 receives the learning instruction. Next, the learning unit 231 reads two or more pieces of teacher data ( FIG. 14 ) from the teacher data storage unit 211, provides the two or more pieces of teacher data to a module that performs machine learning learning processing (e.g., a deep learning module), executes the module, and acquires a learning model. Next, the learning accumulation unit 232 accumulates the learning model in the model storage unit 212.

[0171] Next, the user provides the rank estimation device 3 with one or more ultrasound images of the cow whose BMS number is to be predicted, along with a gain identifier and a location identifier corresponding to each ultrasound image, and inputs a prediction instruction.

[0172] Here, the rank estimation device 3 may acquire the BMS number without using the gain identifier and the location identifier. In such a case, the learning model is information constructed without using the gain identifier and the location identifier.

[0173] Next, the receiving unit 32 of the rank estimation device 3 receives the prediction instruction. Next, the model acquisition unit 331 acquires the learning model acquired by the learning device 2 from the model storage unit 212.

[0174] Next, the prediction unit 332 provides the acquired learning model and the received set of one or more ultrasound images, gain identifiers, and location identifiers to a module that performs machine learning prediction processing, executes the module, and acquires a BMS number. The output unit 34 outputs the acquired BMS number. The concept of this processing is shown in FIG. 15.

[0175] Next, a specific example of the operation of the carcass cross-section image output device 4 will be described. It is assumed that a large number of carcass cross-section images are stored in the carcass cross-section image storage unit 41 of the carcass cross-section image output device 4, as shown in Fig. 16. It is also assumed that the carcass cross-section images in the carcass cross-section image storage unit 41 correspond to the BMS number of the cow that is the subject of the carcass cross-section image, and one or more cow biological parameters. It is also assumed that the one or more cow biological parameters are the thickness of the trapezius muscle, the thickness of the subcutaneous fat directly above the trapezius muscle, the thickness of the subcutaneous fat, the thickness of the rib, the thickness of the latissimus dorsi muscle, and the presence or absence of a lump.

[0176] It is assumed that the receiving unit 42 of the carcass cross-sectional image output device 4 receives a carcass cross-sectional image output instruction. The carcass cross-sectional image output instruction includes one or more ultrasound images received by the rank estimation device 3 and the BMS number output by the rank estimation device 3.

[0177] Next, the rank acquisition unit 43 acquires the rank information "BMS number: 12" contained in the dressed carcass cross-sectional image output instruction.

[0178] Next, the parameter acquisition unit 44 acquires an ultrasound image corresponding to location identifier "1" from among the ultrasound images contained in the carcass cross-sectional image output instruction. This image is shown in Fig. 17(a). Next, the parameter acquisition unit 44 determines the boundary line of the trapezius muscle by performing the above-mentioned image processing on the image in Fig. 17(a). The parameter acquisition unit 44 then acquires the thickness of the trapezius muscle of "3.8" and the thickness of the subcutaneous fat directly above the trapezius muscle of "2.0".

[0179] Furthermore, the parameter acquisition unit 44 acquires an ultrasound image corresponding to location identifier "2" from among the ultrasound images contained in the carcass cross-sectional image output instruction. This image is shown in Fig. 17(b). Next, the parameter acquisition unit 44 performs the above-mentioned image processing on the image in Fig. 17(b) to acquire the subcutaneous fat thickness of "3.9", the latissimus dorsi thickness of "3.6", and the rib thickness of "8.6".

[0180] The parameter acquisition unit 44 may acquire cow biological parameters input by the user.

[0181] Next, the carcass cross-sectional image acquisition unit 45 constructs a vector (12, 3.8, 2.0, 3.9, 3.6, 8.6) having the acquired rank information "BMS number: 12" and one or more acquired cattle biological parameters as elements.

[0182] Next, the carcass cross-sectional image acquisition unit 45 searches for a vector that is most similar to the acquired vector (12, 3.8, 2.0, 3.9, 3.6, 8.6) from the carcass cross-sectional image storage unit 41. Next, the carcass cross-sectional image acquisition unit 45 acquires from the carcass cross-sectional image storage unit 41 a carcass cross-sectional image that is paired with the most similar vector.

[0183] Next, the carcass cross-sectional image output unit 46 outputs the acquired carcass cross-sectional image. An example of such output is shown in Fig. 18. In Fig. 18, the BMS number, the cow biological parameters, and the carcass cross-sectional image are output.

[0184] As described above, the image acquiring device 1 of the present embodiment can obtain suitable ultrasonic images that enable early assessment of the meat quality of live cattle.

[0185] Furthermore, according to the learning device 2 of the present embodiment, an appropriate learning model can be obtained that enables early determination of the meat quality of live cattle.

[0186] Moreover, according to the rank estimation device 3 of this embodiment, rank information for early meat quality assessment of live cattle can be obtained.

[0187] Moreover, the rank estimation device 3 of this embodiment can provide accurate rank information for early meat quality assessment of live cattle.

[0188] Moreover, according to the carcass cross-section image output device 4 of the present embodiment, a future predicted carcass cross-section image of a cow can be easily obtained.

[0189] The processing in this embodiment may be realized by software. This software may be distributed by software download or the like. This software may also be recorded on a recording medium such as a CD-ROM and distributed. This also applies to other embodiments in this specification. The software that realizes the image acquisition device 1 in this embodiment is a program as follows. In other words, this program is a program that causes a computer to function as an image acquisition unit that acquires one or more ultrasound images with a gain of about 60 to about 80 using ultrasound emitted to a predetermined area of ​​a living cow, and an image storage unit that stores the one or more ultrasound images.

[0190] The software for realizing the rank estimation device 3 in this embodiment is the following program. That is, this program causes a computer to: (1) operate the computer in such a way that a model acquisition unit acquires a learning model from a model storage unit in which a learning model obtained by a machine learning learning process is stored for two or more pieces of teacher data having one or more ultrasound images obtained by using ultrasound emitted to a predetermined region of each of two or more live cattle and rank information indicating the rank of the beef quality of the cattle; The program functions as an image receiving unit that receives one or more ultrasound images obtained using ultrasound emitted to a specified area of ​​a live cow, a prediction unit that performs machine learning prediction processing using the one or more ultrasound images received by the image receiving unit and the learning model acquired by the model acquisition unit, and acquires rank information, and an output unit that outputs the rank information acquired by the prediction unit.

[0191] The software for realizing the carcass cross-sectional image output device 4 in this embodiment is the following program. That is, this program causes a computer to function as a rank acquisition unit that acquires rank information of one cow, which is acquired by the rank estimation device according to any one of claims 3 to 8, a parameter acquisition unit that acquires one or more cow biological parameters of the one cow, a carcass cross-sectional image acquisition unit that acquires, using a parameter set having the rank information acquired by the rank acquisition unit and the one or more cow biological parameters acquired by the parameter acquisition unit, one or more carcass cross-sectional images that are one or more carcass cross-sectional images in a carcass cross-sectional image storage unit in which one or more carcass cross-sectional images associated with a parameter set having rank information and one or more cow biological parameters are stored and that correspond to a parameter set that satisfies an approximation condition, and a carcass cross-sectional image output unit that outputs the one or more carcass cross-sectional images acquired by the carcass cross-sectional image acquisition unit.

[0192] Moreover, Fig. 19 shows the appearance of a computer that executes the programs described in this specification to realize each device that constitutes the cow information processing system A of the various embodiments described above. The above-mentioned embodiments can be realized by computer hardware and a computer program executed thereon. Fig. 19 is an overview of this computer system 300, and Fig. 20 is a block diagram of the system 300.

[0193] In FIG. 19, computer system 900 includes a computer 901 including a CD-ROM drive, a keyboard 902, a mouse 903, and a monitor 904.

[0194] 20, computer 901 includes, in addition to CD-ROM drive 905, MPU 911, bus 915 connected to CD-ROM drive 905 etc., ROM 912 for storing programs such as a boot-up program, RAM 913 connected to MPU 911 for temporarily storing instructions of application programs and providing temporary storage space, and hard disk 914 for storing application programs, system programs, and data. Although not shown here, computer 301 may further include a network card for providing connection to a LAN.

[0195] A program that causes the computer system 900 to execute the functions of each device that constitutes the cow information processing system A of the embodiment described above may be stored on a CD-ROM 921, inserted into the CD-ROM drive 905, and then transferred to the hard disk 914. Alternatively, the program may be transmitted to the computer 901 via a network (not shown) and stored on the hard disk 914. The program is loaded into the RAM 913 when executed. The program may also be loaded directly from the CD-ROM 921 or the network.

[0196] The program does not necessarily have to include an operating system (OS) or a third party program that causes the computer 901 to execute the functions of each device that constitutes the cow information processing system A of the embodiment described above. The program only needs to include an instruction portion that calls appropriate functions (modules) in a controlled manner and produces the desired results. How the computer system 300 operates is well known, and a detailed description will be omitted.

[0197] In addition, in the above program, the steps of transmitting information and receiving information do not include processing performed by hardware, such as processing performed by a modem or interface card in the transmitting step (processing that is performed only by hardware).

[0198] Furthermore, the computer that executes the above program may be a single computer or multiple computers. In other words, centralized processing or distributed processing may be performed. In other words, each device that constitutes the cow information processing system A may be a single standalone device, or may be composed of two or more devices.

[0199] Furthermore, in each of the above embodiments, it goes without saying that two or more communication means present in one device may be physically realized by one medium.

[0200] Furthermore, in each of the above embodiments, each process may be realized by centralized processing in a single device, or may be realized by distributed processing in a plurality of devices.

[0201] The present invention is not limited to the above-described embodiment, and various modifications are possible, and it goes without saying that these modifications are also included within the scope of the present invention. [Industrial Applicability]

[0202] INDUSTRIAL APPLICABILITY As described above, the rank estimation device according to the present invention has the effect of obtaining more accurate rank information for early meat quality judgment of live cattle, and is useful as a rank estimation device, etc. [Explanation of symbols]

[0203] 1. Image acquisition device 2 Learning device 3. Rank Estimation Device 4. Carcass cross-sectional image output device 11, 311 Image storage section 12, 32, 42 Reception 13 Image acquisition section 14 Image storage unit 21 Learning Storage Unit 22 Learning Reception Department 23 Learning processing unit 31 Storage area 33 Processing section 34 Output section 41 Carcass cross-sectional image storage section 43 Rank Acquisition Department 44 Parameter Acquisition Section 45 Carcass cross-sectional image acquisition section 46 Carcass Cross-Sectional Image Output Unit 211 Teacher data storage unit 212 Model Storage Unit 231 Learning Department 232 Learning and storage unit 321 Image Reception Department 331 Model Acquisition Department 332 Prediction Department

Claims

1. a model acquisition unit that acquires a learning model from a model storage unit that stores a learning model obtained by a machine learning learning process for two or more pieces of teacher data that include ultrasound images obtained by using ultrasound emitted to a specific region of each of two or more live cattle, a gain identifier that identifies the gain when acquiring the ultrasound images, a location identifier that identifies the location where the ultrasound was applied, and rank information that indicates the rank of the beef quality of the cattle; and an image receiving unit that receives an ultrasound image obtained by using ultrasound emitted to a predetermined region of a living cow, a gain identifier, and a location identifier; a prediction unit that performs machine learning prediction processing using the ultrasound image, gain identifier, location identifier, and learning model acquired by the model acquisition unit, and acquires rank information; and an output unit that outputs the rank information acquired by the prediction unit.

2. The teacher data is The image set includes two or more ultrasound images obtained by using ultrasound emitted to two or more locations of a living cow, a gain identifier paired with each of the two or more ultrasound images, and a location identifier paired with each of the two or more ultrasound images, and rank information; The image receiving unit is The rank estimation device according to claim 1 , further comprising: a first section for receiving two or more ultrasound images; a second section for receiving a gain identifier paired with each of the two or more ultrasound images; and a third section for receiving each of the two or more ultrasound images.

3. The ultrasound image is 3. The rank estimation device according to claim 1, wherein the image is acquired by using ultrasonic waves emitted to a predetermined region of a living cow with a gain of about 60 to about 80 dB.

4. The ultrasound image is 4. The rank estimation device according to claim 3, wherein the image is acquired by using ultrasonic waves emitted to a predetermined region of a living cow with a gain of 66 to 74 dB.

5. A model acquisition unit that acquires a learning model from a model storage unit in which a learning model obtained by machine learning learning processing is stored for two or more teacher data having one or more ultrasound images obtained by using ultrasound emitted to a specified area of ​​each of two or more live cows and rank information indicating the rank of the cow's meat quality, an image receiving unit that accepts one or more ultrasound images obtained by using ultrasound emitted to a specified area of ​​a live cow, and a rank acquisition unit that acquires rank information of a cow acquired by a rank estimation device having a prediction unit that performs machine learning prediction processing using the one or more ultrasound images accepted by the image receiving unit and the learning model acquired by the model acquisition unit, and acquires rank information of the cow. A parameter acquisition unit for acquiring one or more bovine biological parameters of the one cow; a carcass cross-sectional image acquisition unit that acquires, using a parameter set having the rank information acquired by the rank acquisition unit and the one or more cattle biological parameters acquired by the parameter acquisition unit, one or more carcass cross-sectional images in a carcass cross-sectional image storage unit in which one or more carcass cross-sectional images corresponding to the parameter set having the rank information and the one or more cattle biological parameters are stored, and the one or more carcass cross-sectional images corresponding to the parameter set satisfying an approximation condition; a carcass cross-sectional image output unit that outputs one or more carcass cross-sectional images acquired by the carcass cross-sectional image acquisition unit.

6. The parameter acquisition unit 6. A carcass cross-sectional image output device according to claim 5, further comprising an image recognition process for an ultrasonic image acquired using ultrasonic waves emitted to a specific region of the one cow, thereby acquiring the one or more cow biological parameters.

7. A rank estimation method implemented by a model acquisition unit, an image reception unit, a prediction unit, and an output unit, comprising: A model acquisition step in which the model acquisition unit acquires a learning model from a model storage unit in which a learning model obtained by a machine learning learning process is stored for two or more pieces of teacher data having ultrasonic images obtained by using ultrasonic waves emitted to a predetermined region of each of two or more live cattle, a gain identifier that identifies a gain when acquiring the ultrasonic images, a location identifier that identifies the location where the ultrasonic waves are applied, and rank information that indicates the rank of the beef quality of the cattle; An image receiving step in which the image receiving unit receives an ultrasound image obtained by using ultrasound emitted to a predetermined region of a living cow, a gain identifier, and a location identifier; a prediction step in which the prediction unit performs a machine learning prediction process using one or more ultrasound images accepted in the image acceptance step and the learning model acquired by the model acquisition unit, and acquires rank information; and an output step in which the output unit outputs the rank information acquired in the prediction step.

8. A carcass cross-sectional image output method implemented by a carcass cross-sectional image storage unit storing one or more carcass cross-sectional images associated with a parameter set having rank information and one or more bovine biological parameters, a rank acquisition unit, a parameter acquisition unit, a carcass cross-sectional image acquisition unit, and a carcass cross-sectional image output unit, a model acquisition unit that acquires a learning model from a model storage unit that stores a learning model obtained by a machine learning learning process for two or more teacher data items having one or more ultrasound images obtained by using ultrasound emitted to a specified area of ​​each of two or more live cows and rank information indicating the rank of the cow's meat quality; an image reception unit that receives one or more ultrasound images obtained by using ultrasound emitted to a specified area of ​​a live cow; and a rank acquisition step that acquires rank information of a cow by a rank estimation device having a prediction unit that performs machine learning prediction processing using the one or more ultrasound images received by the image reception unit and the learning model acquired by the model acquisition unit; A parameter acquisition step in which the parameter acquisition unit acquires one or more cow biological parameters of the one cow; a carcass cross-sectional image acquisition step in which the carcass cross-sectional image acquisition unit acquires one or more carcass cross-sectional images in the carcass cross-sectional image storage unit, the carcass cross-sectional images corresponding to the parameter set that satisfies an approximation condition, using a parameter set having the rank information acquired in the rank acquisition step and one or more cattle biological parameters acquired in the parameter acquisition step; A carcass cross-sectional image output method comprising a carcass cross-sectional image output step in which the carcass cross-sectional image output unit outputs one or more carcass cross-sectional images acquired by the carcass cross-sectional image acquisition unit.

9. Computer, a model acquisition unit that acquires a learning model from a model storage unit that stores a learning model obtained by a machine learning learning process for two or more pieces of teacher data that include ultrasound images obtained by using ultrasound emitted to a specific region of each of two or more live cattle, a gain identifier that identifies the gain when acquiring the ultrasound images, a location identifier that identifies the location where the ultrasound was applied, and rank information that indicates the rank of the beef quality of the cattle; and an image receiving unit that receives an ultrasound image obtained by using ultrasound emitted to a predetermined region of a living cow, a gain identifier, and a location identifier; a prediction unit that performs machine learning prediction processing using the ultrasound image, gain identifier, location identifier, and learning model acquired by the model acquisition unit, and acquires rank information; A program for causing the prediction unit to function as an output unit that outputs the rank information acquired by the prediction unit.

10. Computer, a model acquisition unit that acquires a learning model from a model storage unit in which a learning model obtained by a machine learning learning process is stored for two or more pieces of teacher data having one or more ultrasound images obtained by using ultrasound emitted to a specified area of ​​each of two or more live cows and rank information indicating the rank of the cow's meat quality; an image reception unit that accepts one or more ultrasound images obtained by using ultrasound emitted to a specified area of ​​a live cow; and a rank acquisition unit that acquires rank information of a cow acquired by a rank estimation device having a prediction unit that performs machine learning prediction processing using the one or more ultrasound images accepted by the image reception unit and the learning model acquired by the model acquisition unit to acquire rank information; A parameter acquisition unit for acquiring one or more bovine biological parameters of the one cow; a carcass cross-sectional image acquisition unit that acquires, using a parameter set having the rank information acquired by the rank acquisition unit and the one or more cattle biological parameters acquired by the parameter acquisition unit, one or more carcass cross-sectional images in a carcass cross-sectional image storage unit in which one or more carcass cross-sectional images corresponding to the parameter set having the rank information and the one or more cattle biological parameters are stored, and the one or more carcass cross-sectional images corresponding to the parameter set satisfying an approximation condition; A program for causing the carcass cross-sectional image acquisition unit to function as a carcass cross-sectional image output unit that outputs one or more carcass cross-sectional images acquired by the carcass cross-sectional image acquisition unit.

Citation Information

Patent Citations

  • DOBUTSUNODANMENZOHYOJISOCHI

    JP1976039190A

  • Plural tomographic image display system of ultrasonic diagnostic apparatus

    JP1983044040A

  • Ultrasonic diagnostic apparatus

    JP1983165826A

  • Subcutaneous fat measuring instrument

    JP1993176925A

  • Image processor and beef carcass judging system using the same

    JP1994318244A