Data processing system and data processing method

Abstract

To provide a data processing system having a learning processing execution unit for executing machine learning processing related to a learning model, and a data processing method.SOLUTION: A data processing system 10 comprises a central server, a plurality of data processing servers (MEC), and a plurality of data collection devices (edge devices). Each of the plurality of data collection devices has: an object data acquisition unit which acquires object data related to an object in a primary industry; an object data analysis unit which analyzes the object data; and a device transmission unit which transmits analysis results by the object data analysis unit to the data processing server. The data processing server has: a server acquisition unit which acquires the analysis results transmitted by the device transmission unit; and a learning processing execution unit which executes machine learning processing related to a learning model where the object data serves as input and features of the object serve as an output by using the analysis results.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to a data processing system and a data processing method. [Background technology]

[0002] Patent document 1 describes an automatic feeding device for aquatic organisms that uses image analysis of the behavior of aquatic organisms to automatically feed them the appropriate amount of food at all times, thereby achieving a high feeding rate, reducing feed costs, reducing human burden, and preventing pollution of environmental waters. [Prior art document] [Patent documents] [Patent Document 1] Japanese Patent Application Laid-Open No. 9-262040 Summary of the Invention [Means for solving the problem]

[0003] According to one embodiment of the present invention, there is provided a data processing system. The data processing system may include a data processing server and a plurality of data collection devices. Each of the plurality of data collection devices may have an object data acquisition unit that acquires object data related to objects in a primary industry. Each of the plurality of data collection devices may have an object data analysis unit that analyzes the object data. Each of the plurality of data collection devices may have a device transmission unit that transmits analysis results by the object data analysis unit to the data processing server. The data processing server may have a server acquisition unit that acquires the analysis results transmitted by the device transmission unit. The data processing server may have a learning process execution unit that uses the analysis results to execute machine learning processing associated with a learning model that uses the object data as input and features of the object as output.

[0004] In the data processing system, the data processing server may include a model transmission unit that transmits the learning model to at least one of the plurality of data collection devices. Each of the plurality of data collection devices may include a storage unit that stores the learning model received from the data processing server. The object data analysis unit may input the object data to the learning model and obtain the characteristics of the object output from the learning model as the analysis result. The learning process execution unit may execute a machine learning process that updates the learning model using the analysis result.

[0005] The data processing system may include a plurality of data processing servers, each corresponding to a plurality of types of objects in a primary industry, and the device transmission unit may transmit the analysis result of the object data to one of the plurality of data processing servers that corresponds to the type of object.

[0006] Any of the data processing systems may include a central server. The multiple data processing servers may include a first fish data processing server corresponding to a first type of fish and a second fish data processing server corresponding to a second type of fish. The learning process execution unit of the first fish data processing server may use the analysis results to execute a machine learning process to update a first fish learning model corresponding to the first type of fish. The learning process execution unit of the second fish data processing server may use the analysis results to execute a machine learning process to update a second fish learning model corresponding to the second type of fish. The first fish data processing server may have a first fish learning model transmission unit that transmits the first fish learning model to the central server. The second fish data processing server may have a second fish learning model transmission unit that transmits the second fish learning model to the central server. The central server may have a learning execution unit that uses the first fish learning model received from the first fish data processing server and the second fish learning model received from the second fish data processing server to generate or update a general-purpose fish learning model that inputs fish data related to fish and outputs characteristics of the fish.

[0007] In the data processing system, the multiple data processing servers may include a first meat data processing server corresponding to a first type of meat and a second meat data processing server corresponding to a second type of meat. The learning process execution unit of the first meat data processing server may use the analysis results to execute a machine learning process to update a first meat learning model corresponding to the first type of meat. The learning process execution unit of the second meat data processing server may use the analysis results to execute a machine learning process to update a second meat learning model corresponding to the second type of fish. The first meat data processing server may have a first meat learning model transmission unit that transmits the first meat learning model to the central server. The second meat data processing server may have a second meat learning model transmission unit that transmits the second meat learning model to the central server. The learning execution unit may use the first meat learning model received from the first meat data processing server and the second meat learning model received from the second meat data processing server to generate or update a general-purpose meat learning model that takes meat data related to meat as input and outputs characteristics of the meat.

[0008] In any of the data processing systems, the object data acquisition unit may acquire measurement data obtained by measuring the object. The object data analysis unit may analyze the measurement data. Each of the multiple data collection devices may further include an object measurement image acquisition unit that acquires an object measurement image of the object being measured. The device transmission unit may transmit the analysis results, features of the object, and the object measurement image to the data processing server. The server acquisition unit may acquire the analysis results, features of the object, and the object measurement image. The learning process execution unit may execute the machine learning process using the analysis results, features of the object, and the object measurement image.

[0009] In the data processing system, the object data acquisition unit may acquire the measurement data obtained by measuring a portion of the object. The learning process execution unit may analyze the object measurement image to identify a measurement portion of the object that was the measurement target, and perform the machine learning process using the measurement portion, the analysis results, and features of the object. The object data acquisition unit may acquire the measurement data obtained by measuring a portion of the object with a spectrometer scanner, and the learning process execution unit may perform machine learning processing associated with the learning model that uses the measurement data as input and outputs features including the freshness of the object. The learning process execution unit may analyze the object measurement image to identify a measurement time when the object was measured, and perform the machine learning process using the identified measurement time, the analysis results, and features of the object. The learning process execution unit may analyze the object measurement image to identify a measurement method used to measure the object, and perform the machine learning process using the identified measurement method, the analysis results, and features of the object.

[0010] In any of the data processing systems, the object data acquisition unit may acquire imaging data of the object. The object data analysis unit may analyze the imaging data. The device transmission unit may transmit the analysis results and features of the object to the data processing server. The server acquisition unit may acquire the analysis results and features of the object. The learning process execution unit may execute the machine learning process using the analysis results and features of the object.

[0011] According to one embodiment of the present invention, there is provided a data processing system. The data processing method may include an object data acquisition step in which a data collection device acquires object data related to objects in a primary industry. The data processing method may include an object data analysis step in which the data collection device analyzes the object data. The data processing method may include a device transmission step in which the data collection device transmits analysis results from the object data analysis step to a data processing server. The data processing method may include a server acquisition step in which the data processing server acquires the analysis results. The data processing method may include a learning process execution step in which the data processing server uses the analysis results to execute a machine learning process associated with a learning model that takes the object data as input and outputs features of the object.

[0012] The above summary of the invention does not list all of the necessary features of the present invention, and subcombinations of these features may also constitute inventions. [Brief explanation of the drawings]

[0013] [Figure 1] An example of a data processing system 10 is shown schematically. [Figure 2] FIG. 2 is an explanatory diagram for explaining an example of the content of data processing in the data processing system 10. [Figure 3] 2 shows an example of a functional configuration of the central server 100. [Figure 4] 2 shows an example of a functional configuration of the MEC 200. [Figure 5] 2 illustrates an example of a functional configuration of the edge device 300. [Figure 6] 1 illustrates an example of a hardware configuration of a computer 1200 that functions as the central server 100, the MEC 200, or the edge device 300. DETAILED DESCRIPTION OF THE INVENTION

[0014] The present invention will be described below through embodiments of the invention, but the following embodiments do not limit the scope of the invention according to the claims. Furthermore, not all of the combinations of features described in the embodiments are necessarily essential to the solution of the invention.

[0015] To improve the efficiency of aquaculture and fattening, it is important to collect various data on-site. However, because we are dealing with nature, it is not easy to collect distinctive data, and a large amount of repetitive data without any particular characteristics is collected. For example, when attempting to collect video images of objects such as fish or beef cattle captured on-site via a network, the amount of data becomes enormous. This puts a strain on data storage and processing. In the data processing system 10 according to this embodiment, for example, an edge device that collects on-site data performs a simple analysis and sends only the necessary data to a higher level. Then, the higher level performs learning using the collected data.

[0016] 1 schematically illustrates an example of a data processing system 10. The data processing system 10 may include a central server 100. The data processing system 10 may include multiple MECs 200. The data processing system 10 may include multiple edge devices 300.

[0017] The control server 100 may be a server that controls a plurality of MECs 200. The control server 100 may be located in, for example, a cloud. The control server 100 may also be located in a core network of a mobile communication operator that manages a plurality of MECs 200.

[0018] The MEC 200 may be a server disposed between a radio base station and a core network. One MEC 200 may be realized by one device. Multiple MECs 200 may be realized by one device. The MEC 200 may be an example of a data processing server.

[0019] The edge device 300 is a device that can use the MEC 200. The edge device 300 may be a smartphone, a tablet terminal, a PC (Personal Computer), a wearable device, an IoT (Internet of Things) device, an IoE (Internet of Everything) device, or the like.

[0020] In this embodiment, the edge device 300 has a function of collecting object data related to the object 50 in the primary industry and analyzing the collected object data. The edge device 300 may be an example of a data collection device.

[0021] The edge device 300, for example, acquires measurement data obtained by measuring the object 50 as object data of the object 50 and analyzes the acquired measurement data. The edge device 300, for example, measures the object 50 itself to acquire the measurement data. The edge device 300 may also acquire the measurement data from a measuring instrument that measured the object 50.

[0022] The measurement of the object 50 may be performed using any measuring device. The edge device 300 and the measuring device may be connected by wire. The edge device 300 and the measuring device may be connected wirelessly. The measuring device may be built into the edge device 300.

[0023] For example, the measurement of the object 50 may be a scan by a spectrometer scanner, and the measurement data may be a scan result. The edge device 300 may analyze the measurement data to obtain characteristics of the object 50, including the freshness, as an analysis result.

[0024] A spectrometer scanner separates light into wavelengths and measures the intensity of each wavelength by receiving it with a detector. A characteristic of spectrometer scanners is that they are known to absorb specific wavelengths depending on the chemicals contained in the object 50. The scan results of a spectrometer scanner vary depending on the type and amount of chemicals contained in the object 50. Since a spectrometer scanner can measure the wavelengths absorbed by the chemicals contained in the object 50, it is possible to estimate the chemicals contained in the object 50 from the scan results. Each object 50 contains different types of chemicals in different amounts, resulting in complex variations in wavelength absorption. However, for example, by performing machine learning using the scan results and the type and amount of chemicals contained in the object 50 as training data, a learning model can be generated that inputs the scan results and outputs the type and amount of chemicals. Using this learning model, the type and amount of chemicals contained in the object 50 can be estimated from the scan results, making it possible to indicate indicators of the quality of the object 50.

[0025] For example, if the object 50 is a fish, the edge device 300 may estimate the K value of the fish as the freshness of the fish. The characteristics of the fish may include the umami of the fish. The characteristics of the fish may include free amino acids contained in the fish. Examples of free amino acids estimated by the edge device 300 include inosinic acid, glutamic acid, and histamine. The characteristics of the fish may include nucleic acids of the fish. The characteristics of the fish may include the water content of the fish. The characteristics of the fish may include lipids of the fish. The characteristics of the fish may include color differences of the fish. The characteristics of the fish may include microorganisms contained in the fish. The characteristics of the fish may include the texture of the fish. The characteristics of the fish may include fatty acids of the fish. The characteristics of the fish may include the blood content of the fish.

[0026] The scan results of a spectrometer scanner vary depending on the freshness of the fish. Therefore, by measuring fish of known freshness with a spectrometer scanner and generating GT data including the scan results and the fish's freshness for various freshness levels, a learning model can be generated that can estimate the freshness of the fish from the scan results.

[0027] The scan results of the spectrometer scanner vary depending on the umami of the fish. Therefore, by measuring fish with known umami flavors with the spectrometer scanner and generating GT data containing the scan results and the umami of the fish for various umami flavors, a learning model can be generated that can estimate the umami of the fish from the scan results.

[0028] The scan results of the spectrometer scanner vary depending on the moisture content of the fish. Therefore, by measuring fish with known moisture contents with the spectrometer scanner and generating GT data including the scan results and the moisture content of the fish for various moisture contents, a learning model can be generated that can estimate the moisture content of the fish from the scan results.

[0029] The scan results of the spectrometer scanner vary depending on the lipid content of the fish. Therefore, by measuring fish with known lipid content using the spectrometer scanner and generating GT data containing the scan results and the lipid content of the fish for various lipids, a learning model can be generated that can estimate the lipid content of the fish from the scan results.

[0030] The scan results of the spectrometer scanner vary depending on the color difference of the fish. Therefore, by measuring fish with known color differences with the spectrometer scanner and generating GT data including the scan results and the color difference of the fish for various color differences, it is possible to generate a learning model that can estimate the color difference of the fish from the scan results.

[0031] The scan results of the spectrometer scanner vary depending on the microorganisms contained in the fish. Therefore, by measuring fish with known microorganisms using the spectrometer scanner and generating GT data containing the scan results and the microorganisms contained in the fish for various contained microorganisms, a learning model can be generated that can estimate the microorganisms contained in the fish from the scan results.

[0032] The scan results of the spectrometer scanner vary depending on the fatty acids of the fish. Therefore, by measuring fish with known fatty acids with the spectrometer scanner and generating GT data containing the scan results and the fish's fatty acids for various fatty acids, a learning model can be generated that can estimate the fatty acids of the fish from the scan results.

[0033] The scan results of the spectrometer scanner vary depending on the blood content of the fish. Therefore, by measuring fish with known blood content using the spectrometer scanner and generating GT data including the scan results and the blood content of the fish for various blood contents, a learning model can be generated that can estimate the blood content of the fish from the scan results.

[0034] The scan results of the spectrometer scanner vary depending on the texture of the fish. Therefore, by measuring fish with known textures with the spectrometer scanner and generating GT data including the scan results and the texture of the fish for various textures, a learning model can be generated that can estimate the texture of the fish from the scan results.

[0035] If the object 50 is from other primary industries, such as meat, fruits, vegetables, and milk, various characteristics of the object 50 can also be estimated by scanning the object 50 with a spectrometer scanner.

[0036] For example, the measurement of the object 50 may be a measurement using a saccharometer. The measurement data may be the sugar content of the object 50. For example, the measurement of the object 50 may be a measurement using a moisture meter. The measurement data may be the moisture content of the object 50. For example, the measurement of the object 50 may be a measurement using a fat meter. The measurement data may be the fat percentage of the object 50. The measurement of the object 50 may be a measurement other than these.

[0037] In the data processing system 10 according to this embodiment, for example, the central server 100 first uses a plurality of GT (Ground Truth) data sets including measurement data and features of the objects 50, without distinguishing between primary industry objects 50, to generate a general-purpose learning model that takes the measurement data of the objects 50 as input and outputs the features of the objects 50. The amount of GT data that can be prepared here is usually insufficient, and the general-purpose learning model at this point often does not have very high estimation accuracy. The central server 100 provides the general-purpose learning model to a plurality of MECs 200.

[0038] Each of the multiple MECs 200 performs machine learning processing to generate a learning model corresponding to each type (sometimes referred to as a type-generalized learning model) based on a general-purpose learning model for each of the general types of primary industry objects 50. For example, each of the multiple MECs 200 performs machine learning processing to generate a type-generalized learning model for each of multiple types corresponding to industry types in the primary industry. The multiple types corresponding to industry types in the primary industry may include, for example, fish, meat, fruit, vegetables, and milk. As an example, each of the multiple MECs 200 may generate a type-generalized learning model by performing transfer learning based on the general-purpose learning model.

[0039] For example, a certain MEC 200 may use GT data of multiple fish to generate a general-purpose learning model for fish (sometimes referred to as a general-purpose fish learning model). The amount of GT data available here is usually insufficient, and the general-purpose fish learning model at this point often does not have very high estimation accuracy. Similarly, a certain MEC 200 may use GT data of multiple meats to generate a general-purpose learning model for meat (sometimes referred to as a general-purpose meat learning model). Similarly, a certain MEC 200 may use GT data of multiple fruits to generate a general-purpose learning model for fruit (sometimes referred to as a general-purpose fruit learning model). Similarly, a certain MEC 200 may use GT data of multiple vegetables to generate a general-purpose learning model for vegetables (sometimes referred to as a general-purpose vegetable learning model). Similarly, a certain MEC 200 may use GT data of multiple milks to generate a general-purpose learning model for milk (sometimes referred to as a general-purpose milk learning model). Similarly, a certain MEC 200 may generate a general-purpose learning model for a type of object 50 other than these.

[0040] Each of the multiple MECs 200 provides the type-general learning model that it has generated to the edge device 300 that collects information on that type of object 50. For example, the MEC 200 that generated the fish-general learning model provides the fish-general learning model to the edge device 300 that collects fish information.

[0041] The edge device 300 transmits to the MEC 200, for example, measurement data obtained by measuring the object 50, estimation results of the features of the object 50 output from the type-generalized learning model after inputting the measurement data into the type-generalized learning model, a measured image of the object 50 captured during measurement, and the features of the object 50. The measurement data may include attribute information of the object 50 indicating the type of the object 50. The features of the object 50 may be features determined by an expert or specialist in the object 50, or may be features measured by a dedicated measuring device. The edge device 300 may acquire the features of the object 50 by receiving input of the features of the object 50 or output from a dedicated measuring device.

[0042] The MEC 200 updates the type-general learning model using the data received from the edge device 300. At this time, the MEC 200 may analyze the object measurement image to identify which part of the object 50 was measured, how long the measurement took, the actual measurement method, and the like, and use the identification results. For example, the MEC 200 may determine whether the measurement method was correct based on the measured part, measurement time, and measurement method, and update the type-general learning model using only data for which the measurement method was correct.

[0043] Each of the multiple MECs 200 may transmit an updated type-generalized learning model to the central server 100. The central server 100 may update the general-purpose learning model using the multiple type-generalized learning models received from the multiple MECs 200. For example, the central server 100 updates the general-purpose learning model using the updated fish-generalized learning model, meat-generalized learning model, fruit-generalized learning model, and vegetable-generalized learning model received from the multiple MECs 200. This improves the estimation accuracy of the general-purpose learning model. If the estimation accuracy of the general-purpose learning model improves, it becomes possible to generate a type-generalized learning model with high estimation accuracy when a new primary industry object 50 is added, thereby making learning more efficient.

[0044] The edge devices 300 may be arranged at various locations in the distribution process of the object 50. For example, if the object 50 is a fish, one edge device 300 may be used by a producer to acquire object data of fish at a fish farm or fish preserve. Another edge device 300 may be arranged at a fish processing plant to acquire object data of fish before and after processing. Another edge device 300 may be arranged at a food processing company to acquire object data of fish before and after processing into food. Another edge device 300 may be arranged at a wholesaler or supermarket to acquire object data of raw fish, frozen fish, thawed fish, etc. Another edge device 300 may be arranged at a retail site to acquire object data of thawed fish, fish ready for sale, etc. The data processing system 10 according to this embodiment can collect data on fish in such various situations and update the fish species learning model, general fish learning model, and general learning model, thereby expanding the application range of the learning models and improving the accuracy of estimating fish characteristics.

[0045] Instead of the central server 100 generating a general-purpose learning model and each of the multiple MECs 200 generating a type-specific general-purpose learning model, the central server 100 may generate a general-purpose learning model and multiple type-specific general-purpose learning models, and each of the multiple MECs 200 may use the type-specific general-purpose learning model to perform machine learning processing to generate a learning model corresponding to each of the more specific types of primary industry objects 50 (sometimes referred to as a type-specific learning model). For example, each of the multiple MECs 200 may generate a type-specific learning model for each of the multiple types of objects 50. As specific examples, the type of fish may include known fish types such as red sea bream and yellowtail. The type of meat may include known meat types such as beef, pork, and chicken. The type of fruit may include known fruit types such as strawberries and apples. The type of vegetables may include known vegetable types such as tomatoes and cabbage. As an example, each of the multiple MECs 200 may generate a type-specific learning model by performing transfer learning based on a type-general learning model.

[0046] For example, each of a plurality of MECs 200 may generate a type-specific learning model (sometimes referred to as a fish-type-specific learning model) for each of a plurality of types of fish. Also, each of a plurality of MECs 200 may generate a type-specific learning model (sometimes referred to as a meat-type-specific learning model) for each of a plurality of types of meat. Also, each of a plurality of MECs 200 may generate a type-specific learning model (sometimes referred to as a fruit-type-specific learning model) for each of a plurality of types of vegetables. Also, each of a plurality of MECs 200 may generate a type-specific learning model (sometimes referred to as a vegetable-type-specific learning model) for each of a plurality of types of other objects 50.

[0047] Each of the multiple MECs 200 provides the type-specific learning model that it has generated to the edge device 300 that collects information on that type of object 50. For example, an MEC 200 that generated a first type of fish type-specific learning model provides the first type of fish type-specific learning model to the edge device 300 that collects information on the first type of fish, and an MEC 200 that generated a fish type-specific learning model corresponding to a second type of fish provides the fish type-specific learning model corresponding to the second type of fish to the edge device 300 that collects information on the second type of fish.

[0048] The edge device 300 transmits to the MEC 200, for example, measurement data obtained by measuring the object 50, estimation results of the features of the object 50 output from the type-specific learning model after inputting the measurement data into the type-specific learning model, a measured image of the object 50 captured during measurement, and the features of the object 50. The features of the object 50 may be features determined by an expert or specialist in the object 50, or may be features measured by a dedicated measuring device. The edge device 300 may acquire the features of the object 50 by receiving input of the features of the object 50 or receiving output from a dedicated measuring device.

[0049] The MEC 200 updates the type-specific learning model using the data received from the edge device 300. At this time, the MEC 200 may analyze the object measurement image to identify which part of the object 50 was measured, how long the measurement took, the actual measurement method, and the like, and use the identification results. For example, the MEC 200 determines whether the measurement method was correct based on the measured part, measurement time, and measurement method, and updates the type-specific learning model using only data for which the measurement method was correct.

[0050] For example, each of the multiple MECs 200 may transmit an updated type-specific learning model to the central server 100. The central server 100 may update a type-general learning model using the multiple type-specific learning models received from the multiple MECs 200. The central server 100 may also update a general-purpose learning model using the updated multiple type-general learning models. As a specific example, the central server 100 may update a fish-general learning model, a meat-general learning model, a fruit-general learning model, a vegetable-general learning model, and a milk-general learning model using the multiple type-specific learning models received from the multiple MECs 200, and then update the general-purpose learning model using the updated fish-general learning model, meat-general learning model, fruit-general learning model, vegetable-general learning model, and milk-general learning model.

[0051] As another example, the edge device 300 acquires imaging data of the object 50 as object data of the object 50 and analyzes the acquired imaging data. For example, the edge device 300 acquires the imaging data by capturing an image of the object 50 itself. The edge device 300 may also acquire the imaging data from an imaging device that captured the object 50.

[0052] By performing machine learning using image data of the object 50 and features such as the freshness of the object 50, it is possible to generate a learning model that inputs the image data of the object 50 and outputs the features of the object 50. For example, it is known that the freshness of fish meat can be determined to some extent from the appearance of the fish meat. Therefore, it is possible to estimate the freshness of fish meat from an image of the fish meat.

[0053] In the data processing system 10 according to this embodiment, for example, the central server 100 receives image data of the object 50 as input and generates a general-purpose learning model that outputs characteristics of the object 50, such as the freshness of the object 50. The central server 100 provides the general-purpose learning model to a plurality of MECs 200. Each of the plurality of MECs 200 may execute a machine learning process to generate a general-purpose learning model.

[0054] For example, one MEC 200 may generate a general-purpose fish learning model for fish. Another MEC 200 may generate a general-purpose meat learning model for meat. Another MEC 200 may generate a general-purpose fruit learning model for fruit. Another MEC 200 may generate a general-purpose vegetable learning model for vegetables. Another MEC 200 may generate a general-purpose milk learning model for milk. Another MEC 200 may generate a type-general learning model for an object 50 other than these.

[0055] Each of the multiple MECs 200 provides the type-general learning model that it has generated to the edge device 300 that collects information on that type of object 50. For example, the MEC 200 that generated a fish learning model provides the fish learning model to the edge device 300 that collects information on fish.

[0056] The edge device 300 transmits, for example, imaging data of the object 50, an estimation result of the characteristics of the object 50 output from the type-general learning model after inputting the imaging data into the type-general learning model, and the characteristics of the object 50 to the MEC 200. The MEC 200 updates the type-general learning model using the data received from the edge device 300.

[0057] Each of the multiple MECs 200 may transmit the updated type-general learning model to the central server 100. The central server 100 may update the general-purpose learning model using the multiple type-general learning models received from the multiple MECs 200. For example, the central server 100 updates the general-purpose learning model using the updated fish-general learning model, meat-general learning model, fruit-general learning model, and vegetable-general learning model received from the multiple MECs 200.

[0058] In addition, instead of the central server 100 generating a general-purpose learning model and each of the multiple MECs 200 generating a type-specific general-purpose learning model, the central server 100 may generate a general-purpose learning model and multiple type-specific general-purpose learning models, and each of the multiple MECs 200 may use the type-specific general-purpose learning model to perform machine learning processing to generate a type-specific learning model for each of the more specific types of primary industry objects 50.

[0059] Each of the multiple MECs 200 provides the type-specific learning model that it has generated to the edge device 300 that collects information on that type of object 50. For example, an MEC 200 that generated a first type of fish type-specific learning model provides the first type of fish type-specific learning model to the edge device 300 that collects information on the first type of fish, and an MEC 200 that generated a fish type-specific learning model corresponding to a second type of fish provides the fish type-specific learning model corresponding to the second type of fish to the edge device 300 that collects information on the second type of fish.

[0060] The edge device 300 transmits, for example, imaging data of the object 50, estimation results of the features of the object 50 output from the type-specific learning model after inputting the imaging data into the type-specific learning model, and the features of the object 50 to the MEC 200. The MEC 200 updates the type-specific learning model using the data received from the edge device 300.

[0061] For example, each of the multiple MECs 200 may transmit an updated type-specific learning model to the central server 100. The central server 100 may update a type-general learning model using the multiple type-specific learning models received from the multiple MECs 200. The central server 100 may also update a general-purpose learning model using the updated multiple type-general learning models. As a specific example, the central server 100 may update a fish-general learning model, a meat-general learning model, a fruit-general learning model, a vegetable-general learning model, and a milk-general learning model using the multiple type-specific learning models received from the multiple MECs 200, and then update the general-purpose learning model using the updated fish-general learning model, meat-general learning model, fruit-general learning model, vegetable-general learning model, and milk-general learning model.

[0062] In the present embodiment, the MEC 200 is a data processing server and the edge device 300 is a data collection device, but the present invention is not limited to this. For example, the data processing server may be a server located on the Internet or in a cloud, and the data collection device may be any device that can use a data processing server located on the Internet or in a cloud.

[0063] 2 is an explanatory diagram for explaining an example of the data processing content in the data processing system 10. Here, an example will be described in which the data processing system 10 performs learning to estimate the characteristics of an object 50 from measurement data of the object 50. Also, here, the example will be described in which fish and meat are used as examples of the object 50 in the primary industry.

[0064] In the example shown in FIG. 2, the central server 100 performs machine learning processing using GT data including measurement data of the object 50 and features of the object 50, without distinguishing between primary industry objects 50, to generate a general-purpose learning model 400 that takes the measurement data of the object 50 as input and outputs the features of the object 50. The central server 100 also generates a general-purpose fish learning model 410 that takes fish measurement data as input and outputs the features of the fish, and a general-purpose meat learning model 420 that takes meat measurement data as input and outputs the features of the meat. The central server 100 may use the general-purpose learning model 400 to generate the general-purpose fish learning model 410 and the general-purpose meat learning model 420. The central server 100 may also generate the general-purpose fish learning model 410 and the general-purpose meat learning model 420 without using the general-purpose learning model 400.

[0065] The central server 100 transmits the generic fish learning model 410 to the MEC 200 that performs learning about fish. The MEC 200 that receives the generic fish learning model 410 may generate, from the generic fish learning model 410, a type-specific learning model that corresponds to the type of fish for which it is performing learning. For example, an MEC 200 that performs learning about red sea bream may generate a type-specific learning model (red sea bream learning model 412) that corresponds to red sea bream. The MEC 200 may generate the red sea bream learning model 412, for example, based on the generic fish learning model 410 and using GT data that includes red sea bream data. For example, an MEC 200 that performs learning about yellowtail may generate a type-specific learning model (yellowtail learning model 414) that corresponds to yellowtail. The MEC 200 may generate the yellowtail learning model 414, for example, based on the generic fish learning model 410 and using GT data that includes yellowtail data.

[0066] The central server 100 transmits the general-purpose meat learning model 420 to the MEC 200 that performs learning related to meat. The MEC 200 that receives the general-purpose meat learning model 420 may generate, from the general-purpose meat learning model 420, a type-specific learning model corresponding to the type of meat for which it is performing learning. For example, the MEC 200 that performs learning related to beef cattle may generate a type-specific learning model corresponding to beef cattle (beef cattle learning model 422). The MEC 200 may generate the beef cattle learning model 422 based on the general-purpose meat learning model 420 using GT data including beef cattle data. For example, the MEC 200 that performs learning related to pigs may generate a type-specific learning model corresponding to pigs (pig learning model 424). The MEC 200 may generate the pig learning model 424 based on the general-purpose meat learning model 420 using GT data including pig data.

[0067] The MEC 200 having the red sea bream learning model 412 may provide the red sea bream learning model 412 to an edge device 300 that collects information on red sea bream. The MEC 200 having the yellowtail learning model 414 may provide the yellowtail learning model 414 to an edge device 300 that collects information on yellowtail. An edge device 300 that collects information on both red sea bream and yellowtail will obtain both the red sea bream learning model 412 and the yellowtail learning model 414.

[0068] An MEC 200 having a beef cattle learning model 422 may provide the beef cattle learning model 422 to an edge device 300 that collects beef cattle information. An MEC 200 having a pig learning model 424 may provide the pig learning model 424 to an edge device 300 that collects pig information. An edge device 300 that collects both beef cattle and pig information will obtain both the beef cattle learning model 422 and the pig learning model 424.

[0069] The edge device 300, which collects information about red sea bream, transmits measurement data obtained by measuring the red sea bream, estimation results of the red sea bream's characteristics output from the red sea bream learning model 412 after inputting the measurement data into the red sea bream learning model 412, an object measurement image of the red sea bream being measured, and the red sea bream's characteristics to the MEC 200, which has the red sea bream learning model 412. The measurement data may include attribute information indicating the red sea bream. The red sea bream's characteristics may be characteristics determined by a red sea bream connoisseur or expert, or characteristics measured using a dedicated measuring device. The MEC 200 updates the red sea bream learning model 412 using the data received from the edge device 300.

[0070] In this case, the MEC 200 may identify the measurement portion of the target red sea bream by analyzing the object measurement image. The measurement results may differ depending on whether the belly or back portion of the red sea bream is measured. Furthermore, the measurement results may differ depending on whether the white meat or the dark blood of the red sea bream is measured. The MEC 200 according to this embodiment may identify the measurement portion by analyzing the object measurement image and determine whether or not to use the data for updating the red sea bream learning model 412. As a result, for example, by using only data measuring the white meat of the red sea bream to update the red sea bream learning model 412, it is possible to realize a red sea bream learning model 412 that has high accuracy in estimating red sea bream characteristics based on the white meat. The MEC 200 may also use the measurement portion to update the red sea bream learning model 412. That is, the MEC 200 may use the measurement portion as one of the parameters of the red sea bream learning model 412. This will contribute to the realization of a red sea bream learning model 412 that can make estimations based on the measured part.

[0071] The MEC 200 may identify the measurement time by analyzing the object measurement image. Measurement results may vary depending on the measurement time. For example, if a measurement is required for one minute but is only performed for 30 seconds, the reliability of the measurement result will be reduced. The MEC 200 according to this embodiment may determine whether or not to use the data to update the red sea bream learning model 412 by identifying the measurement time by analyzing the object measurement image. This can prevent a decrease in the accuracy of the red sea bream learning model 412 due to using measurement results that were measured for an insufficient time to update the red sea bream learning model 412. The MEC 200 may also use the measurement time to update the red sea bream learning model 412. That is, the MEC 200 may use the measurement time as one of the parameters of the red sea bream learning model 412. This contributes to realizing the red sea bream learning model 412 that can perform estimations based on the measurement time.

[0072] The MEC 200 may identify the measurement method by analyzing the object measurement image. Measurement results may vary depending on the measurement method. For example, if a measurement is performed using a measurement method other than a predetermined measurement method, the reliability of the measurement results may be reduced. The MEC 200 according to this embodiment may identify the measurement method by analyzing the object measurement image and determine whether or not to use the data to update the red sea bream learning model 412. This can prevent a decrease in the accuracy of the red sea bream learning model 412 due to using measurement results obtained using an inappropriate measurement method to update the red sea bream learning model 412. The MEC 200 may also use the measurement method to update the red sea bream learning model 412. That is, the MEC 200 may use the measurement method as one of the parameters of the red sea bream learning model 412. This can contribute to realizing the red sea bream learning model 412 that can perform estimations according to the measurement method.

[0073] The MEC 200 transmits the updated red sea bream learning model 412 to the central server 100. The MEC 200 may update the red sea bream learning model 412 for a predetermined period of time, and after the predetermined period has elapsed, transmit the red sea bream learning model 412 to the central server 100. The MEC 200 may update the red sea bream learning model 412 by a predetermined amount of data, and after the amount of data used for learning exceeds the predetermined amount of data, transmit the red sea bream learning model 412 to the central server 100.

[0074] The edge device 300, which collects information about yellowtail, transmits to the MEC 200, which has the yellowtail learning model 414, the measurement data obtained by measuring the yellowtail, the estimated results of the yellowtail's characteristics output from the yellowtail learning model 414 after inputting the measurement data into the yellowtail learning model 414, a measurement image of the yellowtail being measured, and the yellowtail's characteristics. The measurement data may include attribute information identifying the yellowtail. The yellowtail's characteristics may be characteristics determined by a yellowtail connoisseur or expert, or characteristics measured using a dedicated measuring device. The MEC 200 updates the yellowtail learning model 414 using the data received from the edge device 300. At this time, the MEC 200 may analyze the measurement image of the object to identify the measurement part of the yellowtail, the measurement time, the measurement method, etc., and use them for learning.

[0075] The MEC 200 transmits the updated yellowtail learning model 414 to the central server 100. The MEC 200 may update the yellowtail learning model 414 for a predetermined period of time, and after the predetermined period has elapsed, transmit the yellowtail learning model 414 to the central server 100. The MEC 200 may update the yellowtail learning model 414 by a predetermined amount of data, and after the amount of data used for learning exceeds the predetermined amount of data, transmit the yellowtail learning model 414 to the central server 100.

[0076] An edge device 300 that collects information about beef cattle transmits measurement data obtained by measuring the beef cattle, estimation results of the beef cattle's characteristics output from the beef cattle learning model 422 after inputting the measurement data into the beef cattle learning model 422, an object measurement image of the beef cattle being measured, and the beef cattle's characteristics to an MEC 200 that has the beef cattle learning model 422. The measurement data may include attribute information that identifies the beef cattle. The beef cattle's characteristics may be characteristics determined by a beef cattle connoisseur or expert, or may be characteristics measured using a dedicated measuring device. The MEC 200 updates the beef cattle learning model 422 using the data received from the edge device 300.

[0077] In this case, the MEC 200 may identify the measurement part of the target beef cattle by analyzing the object measurement image. For example, when measuring beef cattle meat, the measurement results may differ depending on whether the measurement is of the lean part or the so-called marbling part. The MEC 200 according to this embodiment may identify the measurement part by analyzing the object measurement image and determine whether or not to use the data for updating the beef cattle learning model 422. As a result, for example, by using only data obtained by measuring the lean part of the beef cattle to update the beef cattle learning model 422, it is possible to realize a beef cattle learning model 422 that has high accuracy in estimating beef cattle characteristics based on the lean part. The MEC 200 may also use the measurement part to update the beef cattle learning model 422. That is, the MEC 200 may use the measurement part as one of the parameters of the beef cattle learning model 422. This contributes to realizing a beef cattle learning model 422 that can make estimations based on the measured part.

[0078] The MEC200 may identify the measurement time by analyzing the object measurement image. Measurement results may vary depending on the measurement time. For example, if a measurement is performed for only 30 seconds when it is believed that measurement should be performed for one minute, the reliability of the measurement result will be reduced. The MEC200 according to this embodiment may determine whether or not to use the data for updating the beef cattle learning model 422 by identifying the measurement time by analyzing the object measurement image. This can prevent a decrease in the accuracy of the beef cattle learning model 422 due to using measurement results that were measured for an insufficient time for updating the beef cattle learning model 422. The MEC200 may also use the measurement time to update the beef cattle learning model 422. That is, the MEC200 may use the measurement time as one of the parameters of the beef cattle learning model 422. This can contribute to realizing a beef cattle learning model 422 that can make estimations based on the measurement time.

[0079] The MEC200 may identify the measurement method by analyzing the object measurement image. Measurement results may vary depending on the measurement method. For example, if a measurement is performed using a measurement method different from a predetermined measurement method, the reliability of the measurement results may be reduced. The MEC200 according to this embodiment may identify the measurement method by analyzing the object measurement image, and then determine whether or not to use the data to update the beef cattle learning model 422. This can prevent a decrease in the accuracy of the beef cattle learning model 422 due to using measurement results obtained using an inappropriate measurement method to update the beef cattle learning model 422. The MEC200 may also use the measurement method to update the beef cattle learning model 422. That is, the MEC200 may use the measurement method as one of the parameters of the beef cattle learning model 422. This can contribute to realizing a beef cattle learning model 422 that can perform estimations according to the measurement method.

[0080] MEC200 transmits the updated beef cattle learning model 422 to central server 100. MEC200 may update the beef cattle learning model 422 for a predetermined period of time, and after the predetermined period has elapsed, transmit the beef cattle learning model 422 to central server 100. MEC200 may update the beef cattle learning model 422 by a predetermined amount of data, and after the amount of data used for learning exceeds the predetermined amount of data, transmit the beef cattle learning model 422 to central server 100.

[0081] The edge device 300, which collects pig information, transmits to the MEC 200, which has the pig learning model 424, measurement data obtained by measuring the pig, the estimated results of the yellowtail characteristics output from the pig learning model 424 after inputting the measurement data into the pig learning model 424, a target measurement image of the pig being measured, and the pig's characteristics. The measurement data may include attribute information identifying the pig. The pig's characteristics may be characteristics determined by a pig connoisseur or expert, or characteristics measured using a dedicated measuring device. The MEC 200 updates the pig learning model 424 using the data received from the edge device 300. At this time, the MEC 200 may analyze the target measurement image to identify the measurement part of the pig, the measurement time, the measurement method, etc., and use them for learning.

[0082] The MEC 200 transmits the updated pig learning model 424 to the central server 100. The MEC 200 may update the pig learning model 424 for a predetermined period of time, and transmit the pig learning model 424 to the central server 100 after the predetermined period has elapsed. The MEC 200 may update the pig learning model 424 by a predetermined amount of data, and transmit the pig learning model 424 to the central server 100 after the amount of data used for learning exceeds the predetermined amount of data.

[0083] The central server 100 may update the general-purpose fish learning model 410 using the received red sea bream learning model 412 and yellowtail learning model 414. The central server 100 may update the general-purpose fish learning model 410 using the received beef cattle learning model 422 and pig learning model 424. The central server 100 may update the general-purpose learning model 400 using the updated general-purpose fish learning model 410 and general-purpose meat learning model 420.

[0084] 3 shows an example of the functional configuration of the central server 100. The central server 100 includes a data acquisition unit 112, a learning execution unit 114, a model transmission unit 116, and a model acquisition unit 118.

[0085] The data acquisition unit 112 acquires various data. The data acquisition unit 112 stores the acquired data in the data acquisition unit 112. For example, the data acquisition unit 112 acquires general-purpose learning data for generating a general-purpose learning model. The general-purpose learning data may include object data of the object 50 in the primary industry and GT data including features of the object 50. The object data may be measurement data obtained by measuring the object 50.

[0086] The data acquiring unit 112 may acquire general-purpose learning data prepared in advance by an administrator or the like of the data processing system 10. The data acquiring unit 112 may acquire GT data generated by the edge device 300, the GT data including object data of the object 50 from which the edge device 300 collects information and features of the object 50.

[0087] The learning execution unit 114 executes machine learning processing using the data acquired by the data acquisition unit 112. For example, the learning execution unit 114 executes machine learning processing using the data acquired by the data acquisition unit 112 to generate a learning model. The learning execution unit 114 stores the generated learning model in the storage unit 110. For example, the learning execution unit 114 executes machine learning processing using the data acquired by the data acquisition unit 112 to generate a general-purpose learning model. The learning execution unit 114 may execute machine learning processing using the data acquired by the data acquisition unit 112 to generate a type-general-purpose learning model. The learning execution unit 114 may generate the type-general-purpose learning model based on the general-purpose learning model.

[0088] The model sending unit 116 sends the learning model generated by the learning execution unit 114 to multiple MECs 200. For example, the model sending unit 116 sends a general-purpose learning model to multiple MECs 200. For example, the model sending unit 116 sends a type-general-purpose learning model to one or more MECs 200 that perform learning on the object 50 corresponding to the type.

[0089] The model acquisition unit 118 acquires the type-general learning model updated by the MEC 200 from the MEC 200. The model acquisition unit 118 stores the acquired type-general learning model in the memory unit 110. The learning execution unit 114 may replace the type-general learning model stored in the memory unit 110 with the updated type-general learning model acquired by the model acquisition unit 118. The learning execution unit 114 may update the general-purpose learning model using the type-general learning model acquired by the model acquisition unit 118.

[0090] 4 shows an example of the functional configuration of the MEC 200. The MEC 200 includes a storage unit 210, a model acquisition unit 212, a model generation unit 214, a model transmission unit 216, a server acquisition unit 218, and a learning process execution unit 220.

[0091] The model acquisition unit 212 acquires a learning model from the central server 100. The model acquisition unit 212 stores the acquired learning model in the storage unit 210. For example, the model acquisition unit 212 acquires a general-purpose learning model from the central server 100. For example, the model acquisition unit 212 acquires a type-general-purpose learning model from the central server 100.

[0092] The model generation unit 214 generates a learning model. The model generation unit 214 stores the generated learning model in the storage unit 210. For example, the model generation unit 214 generates a type-generalized learning model using the general-purpose learning model acquired by the model acquisition unit 212. The model generation unit 214 generates the type-generalized learning model using, for example, GT data of a certain type of object 50 stored in advance in the storage unit 210, based on the general-purpose learning model.

[0093] For example, the model generation unit 214 generates a type-specific learning model using a type-generalized learning model acquired from the central server 100 or a type-generalized learning model generated by itself. The model generation unit 214 generates a type-specific learning model using GT data of a certain type of object 50 stored in advance in the storage unit 210, for example, based on the type-generalized learning model.

[0094] The model sending unit 216 sends the learning model to at least one of the multiple MECs 200. For example, the model sending unit 216 sends the type-general learning model generated by the model generation unit 214 to the edge device 300. For example, the model sending unit 216 sends the general-purpose learning model to the edge device 300 that collects information on objects 50 of the type targeted by the type-general learning model generated by the model generation unit 214. For example, the model sending unit 216 sends the type-specific learning model to the general-purpose learning model 400 that collects information on objects 50 of the type targeted by the type-specific learning model generated by the model generation unit 214.

[0095] 5 illustrates an example of the functional configuration of the edge device 300. The edge device 300 includes a storage unit 310, a model acquisition unit 312, an object data acquisition unit 314, an object data analysis unit 316, a feature information acquisition unit 318, an object measurement image acquisition unit 320, and a device transmission unit 322. It is not essential that the edge device 300 include all of these units.

[0096] The model acquisition unit 312 acquires a learning model from the MEC 200. The model acquisition unit 312 stores the acquired learning model in the storage unit 310. For example, the model acquisition unit 312 acquires a type-general learning model from the MEC 200. For example, the model acquisition unit 312 acquires a type-specific learning model from the MEC 200.

[0097] The object data acquiring unit 314 acquires object data related to the object 50 about which the edge device 300 is to collect information. The object data acquiring unit 314 acquires, for example, measurement data obtained by measuring the object 50. The object data acquiring unit 314 may acquire the measurement data of the object 50 measured by a measuring device included in the edge device 300 from the measuring device. The object data acquiring unit 314 may receive the measurement data of the object 50 measured by a measuring device separate from the edge device 300 from the measuring device.

[0098] The object data analysis unit 316 analyzes the object data acquired by the object data acquisition unit 314 .

[0099] For example, the object data analysis unit 316 analyzes the object data using a type-generalized learning model stored in the storage unit 310. The object data analysis unit 316 may input the object data into the type-generalized learning model and obtain the characteristics of the object 50 output from the type-generalized learning model as the analysis result. The analysis result may include the analyzed object data.

[0100] For example, the object data analysis unit 316 analyzes the object data using a type-specific learning model stored in the storage unit 310. The object data analysis unit 316 may input the object data to the type-specific learning model and obtain the features of the object 50 output from the type-specific learning model as the analysis result.

[0101] The feature information acquisition unit 318 acquires feature information indicating the features of the object 50 corresponding to the object data analyzed by the object data analysis unit 316. The feature information acquisition unit 318 may acquire feature information indicating the features of the object 50 judged by an expert or a connoisseur of the object 50. The feature information acquisition unit 318 may acquire feature information indicating the results of measurement by a dedicated measuring device that measures the features of the object 50.

[0102] When the object data is measurement data, the object measurement image acquisition unit 320 acquires an object measurement image obtained by capturing an image of the object 50 during measurement. The object measurement image acquisition unit 320 stores the acquired object measurement image in the memory unit 310. The image of the object 50 during measurement may be captured by a camera provided in the edge device 300. The image of the object 50 during measurement may be captured by a measuring instrument. For example, when the measurement is a scan using a spectrometer scanner, the object measurement image includes at least the portion of the object 50 being measured. It is desirable that the object measurement image include the entire object 50 and the portion of the object 50 being measured. The object measurement image may be a moving image captured from before the start of measurement to after the start of measurement.

[0103] The device transmitter 322 transmits various data to the MEC 200. For example, the device transmitter 322 transmits the analysis results by the object data analyzer 316 to the MEC 200. For example, the device transmitter 322 transmits the feature information acquired by the feature information acquirer 318 to the MEC 200. For example, the device transmitter 322 transmits the object measurement image acquired by the object measurement image acquirer 320 to the MEC 200.

[0104] The device transmitter 322 may transmit the analysis results of the object data to one of the multiple MECs 200 that corresponds to the type of object 50 corresponding to the object data. The device transmitter 322 may transmit the analysis results of the object data and feature information to one of the multiple MECs 200 that corresponds to the type of object 50 corresponding to the object data. The device transmitter 322 may transmit the analysis results of the object data, feature information, and object measurement image to one of the multiple MECs 200 that corresponds to the type of object 50 corresponding to the object data.

[0105] The server acquisition unit 218 of the MEC 200 acquires various data transmitted by the device transmission unit 322. For example, the server acquisition unit 218 acquires the analysis results transmitted by the device transmission unit 322. For example, the server acquisition unit 218 acquires the feature information transmitted by the device transmission unit 322. For example, the server acquisition unit 218 acquires the object measurement image transmitted by the device transmission unit 322.

[0106] The learning process execution unit 220 uses the analysis results acquired by the server acquisition unit 218 to execute machine learning processing related to a learning model that takes object data as input and outputs the features of the object 50. For example, the learning process execution unit 220 executes machine learning processing to update the learning model using the analysis results. As a specific example, the learning process execution unit 220 uses the analysis results to update a type-specific learning model. As a specific example, the learning process execution unit 220 uses the analysis results to update a type-general learning model.

[0107] The learning process execution unit 220 may execute machine learning processing using the analysis results and feature information acquired by the server acquisition unit 218. For example, when the degree of match between the estimated features indicated by the analysis results and the features indicated by the feature information is lower than a predetermined threshold, the learning process execution unit 220 updates the type-specific learning model so that the features output when object data included in the analysis results is input into the type-specific learning model will approach the features indicated by the feature information. For example, when the degree of match between the estimated features indicated by the analysis results and the features indicated by the feature information is lower than a predetermined threshold, the learning process execution unit 220 updates the type-specific learning model so that the features output when object data included in the analysis results is input into the type-specific learning model will approach the features indicated by the feature information.

[0108] The learning process execution unit 220 may execute machine learning processing using the analysis results and feature information acquired by the server acquisition unit 218 and the object measurement image.

[0109] The learning process execution unit 220 may identify a measurement portion of the object 50 that is the measurement target by analyzing the object measurement image. For example, if the identified measurement portion is a predetermined portion, the learning process execution unit 220 may execute a machine learning process using the analysis result and the feature information, and if the identified measurement portion is not a predetermined portion, may not execute a machine learning process using the analysis result and the feature information. For example, the learning process execution unit 220 may use the measurement portion as one of the parameters of a learning model.

[0110] The learning process executing unit 220 may identify the measurement time of the object 50 by analyzing the object measurement image. For example, if the difference between the identified measurement time and a predetermined time is within a predetermined threshold range, the learning process executing unit 220 may execute a machine learning process using the analysis result and the feature information, and if the difference is outside the threshold range, may not execute the machine learning process using the analysis result and the feature information. For example, the learning process executing unit 220 may use the measurement time as one of the parameters of a learning model.

[0111] The learning process executing unit 220 may identify the measurement method of the object 50 by analyzing the object measurement image. For example, if the learning process executing unit 220 determines that the identified measurement method is a legitimate measurement method, it may execute machine learning processing using the analysis results and the feature information, and if it determines that the measurement method is not a legitimate measurement method, it may not execute machine learning processing using the analysis results and the feature information. The learning process executing unit 220 may use the measurement method as one of the parameters of a learning model, for example.

[0112] As described above, the object data may be imaging data of the object 50. In this case, the object data acquisition unit 314 acquires imaging data of the object 50. The object data acquisition unit 314 may acquire imaging data of the object 50 captured by a camera included in the edge device 300 from the camera. The object data acquisition unit 314 may receive imaging data of the object 50 captured by an imaging device separate from the edge device 300 from the imaging device.

[0113] The object data analysis unit 316 may analyze the imaging data. The device transmission unit 322 may transmit the analysis results of the imaging data by the object data analysis unit 316 and the feature information acquired by the feature information acquisition unit 318 to the MEC 200. The server acquisition unit 218 may acquire the analysis results and the feature information transmitted by the device transmission unit 322. The learning process execution unit 220 may execute machine learning processing using the analysis results and feature information acquired by the server acquisition unit 218.

[0114] 6 schematically illustrates an example of the hardware configuration of a computer 1200 functioning as the central server 100, the MEC 200, or the edge device 300. A program installed on the computer 1200 can cause the computer 1200 to function as one or more "parts" of an apparatus according to the present embodiment, or to execute operations associated with the apparatus according to the present embodiment or one or more "parts," and / or to execute a process according to the present embodiment or steps of the process. Such a program can be executed by the CPU 1212 to cause the computer 1200 to perform specific operations associated with some or all of the blocks in the flowcharts and block diagrams described herein.

[0115] The computer 1200 according to this embodiment includes a CPU 1212, a RAM 1214, and a graphics controller 1216, which are interconnected by a host controller 1210. The computer 1200 also includes input / output units such as a communications interface 1222, a storage device 1224, a DVD drive, and an IC card drive, which are connected to the host controller 1210 via an input / output controller 1220. The DVD drive may be a DVD-ROM drive, a DVD-RAM drive, or the like. The storage device 1224 may be a hard disk drive, a solid-state drive, or the like. The computer 1200 also includes a ROM 1230 and legacy input / output units such as a keyboard, which are connected to the input / output controller 1220 via an input / output chip 1240.

[0116] The CPU 1212 operates according to programs stored in the ROM 1230 and the RAM 1214, thereby controlling each unit. The graphics controller 1216 acquires image data generated by the CPU 1212 into a frame buffer or the like provided in the RAM 1214 or into the graphics controller itself, and causes the image data to be displayed on the display device 1218.

[0117] The communication interface 1222 communicates with other electronic devices via a network. The storage device 1224 stores programs and data used by the CPU 1212 in the computer 1200. The DVD drive reads programs or data from a DVD-ROM or the like and provides them to the storage device 1224. The IC card drive reads programs and data from an IC card and / or writes programs and data to an IC card.

[0118] The ROM 1230 stores therein a boot program or the like that is executed by the computer 1200 upon activation, and / or programs that depend on the hardware of the computer 1200. The input / output chip 1240 may also connect various input / output units to the input / output controller 1220 via a USB port, a parallel port, a serial port, a keyboard port, a mouse port, etc.

[0119] The programs are provided by a computer-readable storage medium such as a DVD-ROM or an IC card. The programs are read from the computer-readable storage medium, installed in the storage device 1224, RAM 1214, or ROM 1230, which are also examples of computer-readable storage media, and executed by the CPU 1212. Information processing described in these programs is read by the computer 1200, and causes cooperation between the programs and the various types of hardware resources described above. An apparatus or method may be configured by implementing operations or processing of information in accordance with the use of the computer 1200.

[0120] For example, when communication is performed between the computer 1200 and an external device, the CPU 1212 may execute a communication program loaded into the RAM 1214 and instruct the communication interface 1222 to perform communication processing based on the processing described in the communication program. Under the control of the CPU 1212, the communication interface 1222 reads transmission data stored in a transmission buffer area provided in the RAM 1214, the storage device 1224, a DVD-ROM, or a recording medium such as an IC card, and transmits the read transmission data to the network, or writes reception data received from the network to a reception buffer area or the like provided on the recording medium.

[0121] Furthermore, the CPU 1212 may cause all or a necessary portion of a file or database stored in an external recording medium such as the storage device 1224, a DVD drive (DVD-ROM), an IC card, etc. to be read into the RAM 1214, and may perform various types of processing on the data on the RAM 1214. The CPU 1212 may then write back the processed data to the external recording medium.

[0122] Various types of information, such as various types of programs, data, tables, and databases, may be stored on the recording medium and may undergo information processing. The CPU 1212 may perform various types of processing on data read from the RAM 1214, including various types of operations, information processing, conditional judgment, conditional branching, unconditional branching, information search / replacement, etc., as described throughout this disclosure and specified by the instruction sequences of the programs, and write the results back to the RAM 1214. The CPU 1212 may also search for information in a file, database, etc. on the recording medium. For example, if multiple entries, each having an attribute value of a first attribute associated with an attribute value of a second attribute, are stored on the recording medium, the CPU 1212 may search for an entry whose attribute value of the first attribute matches a specified condition from among the multiple entries, read the attribute value of the second attribute stored in the entry, and thereby obtain the attribute value of the second attribute associated with the first attribute that satisfies a predetermined condition.

[0123] The above-described programs or software modules may be stored in a computer-readable storage medium on or near the computer 1200. A recording medium such as a hard disk or RAM provided in a server system connected to a dedicated communication network or the Internet can also be used as a computer-readable storage medium, thereby providing the programs to the computer 1200 via the network.

[0124] The blocks in the flowcharts and block diagrams in the present embodiments may represent stages of a process in which an operation is performed or "parts" of an apparatus responsible for performing the operation. Particular stages and "parts" may be implemented by dedicated circuitry, programmable circuitry provided with computer-readable instructions stored on a computer-readable storage medium, and / or a processor provided with computer-readable instructions stored on a computer-readable storage medium. The dedicated circuitry may include digital and / or analog hardware circuits, including integrated circuits (ICs) and / or discrete circuits. The programmable circuitry may include reconfigurable hardware circuits, such as field programmable gate arrays (FPGAs) and programmable logic arrays (PLAs), including AND, OR, XOR, NAND, NOR, and other logical operations, flip-flops, registers, and memory elements.

[0125] A computer-readable storage medium may include any tangible device capable of storing instructions that are executed by an appropriate device, such that a computer-readable storage medium having instructions stored thereon comprises an article of manufacture, including instructions that can be executed to create means for performing the operations specified in the flowcharts or block diagrams. Examples of computer-readable storage media may include electronic storage media, magnetic storage media, optical storage media, electromagnetic storage media, semiconductor storage media, etc. More specific examples of computer-readable storage media may include floppy disks, diskettes, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), electrically erasable programmable read-only memory (EEPROM), static random access memory (SRAM), compact disc read-only memory (CD-ROM), digital versatile disc (DVD), Blu-ray disc, memory stick, integrated circuit card, etc.

[0126] The computer readable instructions may include either assembler instructions, instruction set architecture (ISA) instructions, machine instructions, machine-dependent instructions, microcode, firmware instructions, state-setting data, or source or object code written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Smalltalk®, JAVA®, C++, etc., and conventional procedural programming languages ​​such as the “C” programming language or similar programming languages.

[0127] Computer-readable instructions may be provided to a general-purpose computer, a special-purpose computer, or another programmable data processing device, or a programmable circuit, either locally or via a local area network (LAN) or a wide area network (WAN) such as the Internet, so that the processor of the programmable data processing device, such as a computer, or the programmable circuit executes the computer-readable instructions to generate means for performing the operations specified in the flowcharts or block diagrams. Here, the computer may be a personal computer (PC), a tablet computer, a smartphone, a workstation, a server computer, a general-purpose computer, a special-purpose computer, or the like, or may be a computer system in which multiple computers are connected. Such a computer system in which multiple computers are connected is also called a distributed computing system, and is a broad definition of computers. In a distributed computing system, multiple computers collectively execute a program by each executing a portion of the program and passing data between the computers as needed during program execution.

[0128] Examples of processors include computer processors, central processing units, processing units, microprocessors, digital signal processors, controllers, microcontrollers, etc. A computer may have one or more processors. In a multiprocessor system with multiple processors, each processor executes a portion of a program and passes data between processors as needed during program execution, allowing the multiple processors to collectively execute the program. For example, in multitasking, each of the multiple processors may execute a portion of each task in small chunks by switching tasks at time slice intervals. In this case, which portion of a program each processor executes changes dynamically. Which portion of a program each of the multiple processors executes may also be statically determined by multiprocessor-aware programming.

[0129] Although the present invention has been described above using the embodiments, the technical scope of the present invention is not limited to the scope described in the above embodiments. It will be apparent to those skilled in the art that various modifications and improvements can be made to the above embodiments. It is clear from the claims that such modifications and improvements can also be included within the technical scope of the present invention.

[0130] It should be noted that the execution order of each process, such as operations, procedures, steps, and stages, in the devices, systems, programs, and methods shown in the claims, specifications, and drawings is not specifically stated as "before," "prior to," etc., and that the processes can be performed in any order unless the output of a previous process is used in a later process. Even if the operational flow in the claims, specifications, and drawings is described using "first," "next," etc. for convenience, this does not mean that the processes must be performed in this order. [Explanation of symbols]

[0131] 50 object, 100 central server, 110 memory unit, 112 data acquisition unit, 114 learning execution unit, 116 model transmission unit, 118 model acquisition unit, 200 MEC, 210 memory unit, 212 model acquisition unit, 214 model generation unit, 216 model transmission unit, 218 server acquisition unit, 220 learning processing execution unit, 300 edge device, 310 memory unit, 312 model acquisition unit, 314 object data acquisition unit, 316 object data analysis unit, 318 feature information acquisition unit, 320 object measurement image acquisition unit, 322 device transmission unit, 410 general-purpose fish learning model, 412 red sea bream learning model, 414 yellowtail learning model, 420 general-purpose meat learning model, 422 beef cattle learning model, 424 pig learning model, 1200 computer, 1210 host controller, 1212 CPU, 1214 RAM, 1216 graphics controller, 1218 display device, 1220 input / output controller, 1222 communication interface, 1224 storage device, 1230 ROM, 1240 input / output chip

Claims

1. a data processing server; Multiple data collection devices and Equipped with Each of the plurality of data collection devices an object data acquisition unit that acquires object data related to objects in the primary industry; an object data analysis unit that analyzes the object data; a device transmission unit that transmits an analysis result by the object data analysis unit to the data processing server; and The data processing server a server acquisition unit that acquires the analysis result transmitted by the device transmission unit; a learning process execution unit that uses the analysis results to execute machine learning processing related to a learning model that takes the object data as input and outputs the features of the object; A data processing system comprising:

2. the data processing server has a model transmission unit that transmits the learning model to at least one of the plurality of data collection devices; each of the plurality of data collection devices has a storage unit that stores the learning model received from the data processing server; the object data analysis unit inputs the object data into the learning model and acquires the feature of the object output from the learning model as the analysis result; The data processing system according to claim 1 , wherein the learning process execution unit executes a machine learning process that updates the learning model using the analysis result.

3. the data processing system comprises a plurality of the data processing servers, each corresponding to a respective one of a plurality of types of primary industry objects; the device transmission unit transmits the analysis result of the object data to a data processing server corresponding to the type of the object, among the plurality of data processing servers; 3. The data processing system of claim 2.

4. the data processing system includes a central server; the plurality of data processing servers includes a first fish data processing server corresponding to a first type of fish and a second fish data processing server corresponding to a second type of fish; The learning process execution unit of the first fish data processing server executes a machine learning process to update a first fish learning model corresponding to the first type of fish using the analysis result; The learning process execution unit of the second fish data processing server executes a machine learning process to update a second fish learning model corresponding to the second type of fish using the analysis result; The first fish data processing server has a first fish learning model transmission unit that transmits the first fish learning model to the central server, The second fish data processing server has a second fish learning model transmission unit that transmits the second fish learning model to the central server, The data processing system described in claim 3, wherein the central server has a learning execution unit that uses the first fish learning model received from the first fish data processing server and the second fish learning model received from the second fish data processing server to generate or update a general-purpose fish learning model that inputs fish data related to a fish and outputs characteristics of the fish.

5. the plurality of data processing servers include a first meat data processing server corresponding to a first type of meat and a second meat data processing server corresponding to a second type of meat; the learning process execution unit of the first meat data processing server executes a machine learning process to update a first meat learning model corresponding to the first type of meat using the analysis result; the learning process execution unit of the second meat data processing server executes a machine learning process to update a second meat learning model corresponding to the second type of fish using the analysis result; The first meat data processing server has a first meat learning model transmission unit that transmits the first meat learning model to the central server, The second meat data processing server has a second meat learning model transmission unit that transmits the second meat learning model to the central server, The data processing system of claim 4, wherein the learning execution unit uses the first meat learning model received from the first meat data processing server and the second meat learning model received from the second meat data processing server to generate or update a general-purpose meat learning model that inputs meat data related to meat and outputs characteristics of the meat.

6. the object data acquisition unit acquires measurement data obtained by measuring the object, the object data analysis unit analyzes the measurement data, Each of the plurality of data collection devices further includes an object measurement image acquisition unit that acquires an object measurement image obtained by capturing an image of the object being measured, the device transmission unit transmits the analysis result, the characteristics of the object, and the object measurement image to the data processing server; the server acquisition unit acquires the analysis result, the characteristics of the object, and the object measurement image; The data processing system according to claim 1 , wherein the learning process execution unit executes the machine learning process using the analysis result, the features of the object, and the object measurement image.

7. the object data acquisition unit acquires the measurement data measured on a part of the object, The data processing system of claim 6, wherein the learning process execution unit identifies a measurement portion of the object that is the subject of measurement by analyzing the object measurement image, and executes the machine learning process using the identified measurement portion, the analysis results, and the characteristics of the object.

8. the object data acquisition unit acquires the measurement data measured by a spectrometer scanner on a part of the object; The data processing system according to claim 7 , wherein the learning process execution unit executes machine learning processing related to the learning model, using the measurement data as input and outputting features including the freshness of the object.

9. The data processing system of claim 6, wherein the learning processing execution unit identifies a measurement time at which the object was measured by analyzing the object measurement image, and executes the machine learning processing using the identified measurement time, the analysis result, and the characteristics of the object.

10. The data processing system of claim 6, wherein the learning processing execution unit identifies a measurement method used to measure the object by analyzing the object measurement image, and executes the machine learning processing using the identified measurement method, the analysis results, and the characteristics of the object.

11. the object data acquisition unit acquires imaging data of the object, the object data analysis unit analyzes the imaging data, the device transmission unit transmits the analysis result and the characteristics of the object to the data processing server; the server acquisition unit acquires the analysis result and the characteristics of the object; The data processing system according to claim 1 , wherein the learning process execution unit executes the machine learning process using the analysis result and the features of the object.

12. an object data acquisition stage in which a data collection device acquires object data relating to an object in the primary industry; an object data analysis step in which the data collection device analyzes the object data; a device transmission step in which the data collection device transmits the analysis result of the object data analysis step to a data processing server; a server acquisition step in which the data processing server acquires the analysis results; a learning process execution step in which the data processing server uses the analysis result to execute machine learning processing related to a learning model that takes the object data as input and outputs the features of the object; A data processing method comprising:

Images