Information processing device, information processing method, and information processing program

Abstract

To accurately estimate the age of geological strata. [Solution] The classification unit 102 includes a first generator 121, a second generator 122, and a third generator 123. The first generator 121 generates an intermediate representation 113 based on the fossil image group 112. The second generator 122 generates type information 114 indicating the likelihood of each type of microfossil in each fossil image based on the intermediate representation 113. The third generator 123 generates age information 115 indicating the likelihood of each age of the strata corresponding to the fossil image group 112 based on the intermediate representation 113. The classification unit 102 is trained using training data 130 to output type information 114 indicating the likelihood of each type of microfossil in each fossil image and age information 115 indicating the likelihood of each age of the strata corresponding to the fossil image group 112, based on the fossil image group 112.

Description

Technical Field

[0001] The present invention relates to an information processing apparatus, an information processing method, and an information processing program.

Background Art

[0002] Conventionally, there are cases where it is desirable to estimate the age of a stratum. For example, in studies such as correlation of contemporaneous strata or rock samples for each region or well, geological structure interpretation or estimation of resource amount and storage capacity in exploration for oil and natural gas or survey of suitable sites for CO₂ underground storage, and geological evaluation such as prediction of the depth and lithofacies of the target horizon during drilling, the age of strata is used as important knowledge. Specifically, in drilling operations, if the age of the stratum being drilled can be estimated, it is expected that this will contribute to recognizing the discrepancy between the expected stratigraphy and the actual stratigraphy, correcting the prediction of the depth at which the target horizon is reached, or predicting the physical properties of the horizon during drilling, and thus contributing to the optimization of drilling operations. In contrast, there exists a technique for estimating the age of a stratum based on the combination of fossil species of microfossils including calcareous nannofossils appearing in samples collected from the stratum.

[0003] As a prior art, for example, there is a technique that analyzes image data of microfossils in a sample collected from a stratum using a machine learning result, so as to analyze the taxon or type of microfossils in the image data.

Prior Art Documents

Patent Documents

[0004]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0005] However, conventional techniques make it difficult to accurately estimate the age of geological strata. For example, consider a case where a worker visually inspects a sample taken from a geological stratum and examines the combination of fossil species of microfossils appearing in the sample. In this case, it is difficult to secure workers because there are few experts who can visually inspect the sample and accurately identify the fossil species of microfossils. Furthermore, it is difficult to estimate the age of geological strata in real time. In addition, conventional techniques such as those described in Patent Document 1 above have challenges in improving the accuracy of estimating the age of geological strata.

[0006] In one aspect, the present invention aims to accurately estimate the age of geological strata. [Means for solving the problem]

[0007] According to one embodiment, an information processing device, information processing method, and information processing program are proposed, which detect a group of fossil images containing microfossils within a geological sample image, and use a classification unit trained with training data that associates a group of specimen images containing microfossils with information indicating the correct type of microfossil in each specimen image of the group of specimen images and information indicating the correct age of the geological layer corresponding to the group of specimen images. The classification unit then generates information indicating the likelihood for each type of microfossil in each fossil image of the group of fossils and information indicating the likelihood for each age of the geological layer corresponding to the group of fossils, according to the detected group of fossil images. The classification unit includes a first generator that generates an intermediate representation based on the group of fossil images, a second generator that generates information indicating the likelihood for each type of microfossil in each fossil image of the group of fossil images based on the intermediate representation, and a third generator that generates information indicating the likelihood for each age of the geological layer corresponding to the group of fossil images based on the intermediate representation. [Effects of the Invention]

[0008] According to one embodiment, it becomes possible to accurately estimate the age of geological strata. [Brief explanation of the drawing]

[0009] [Figure 1]Figure 1 is an explanatory diagram showing one embodiment of the information processing method according to the embodiment. [Figure 2] Figure 2 is an explanatory diagram showing an example of the information processing system 200. [Figure 3] Figure 3 is a block diagram showing an example of the hardware configuration of the information processing device 100. [Figure 4] Figure 4 is a block diagram showing an example of the functional configuration of the information processing device 100. [Figure 5] Figure 5 is an explanatory diagram (part 1) showing an example of classification model 510. [Figure 6] Figure 6 is an explanatory diagram (part 2) showing an example of classification model 510. [Figure 7] Figure 7 is an explanatory diagram (part 1) showing an example of how to generate training data. [Figure 8] Figure 8 is an explanatory diagram (part 2) showing an example of how to generate training data. [Figure 9] Figure 9 is an explanatory diagram (part 3) showing an example of how to generate training data. [Figure 10] Figure 10 is an explanatory diagram (part 1) showing a specific example of the 900 pseudo-age data points used as training data. [Figure 11] Figure 11 is an explanatory diagram (part 2) showing a specific example of the 900 pseudo-age data points used as training data. [Figure 12] Figure 12 is an explanatory diagram (part 1) showing an example of how to estimate the geological age of each stratum. [Figure 13] Figure 13 is an explanatory diagram (part 2) showing an example of how to estimate the geological age of each stratum. [Figure 14] Figure 14 is an explanatory diagram (part 1) showing an example of estimating the geological age of each stratum. [Figure 15] Figure 15 is an explanatory diagram (part 2) showing an example of estimating the geological age of each stratum. [Figure 16] Figure 16 is an explanatory diagram (part 3) showing an example of estimating the geological age of each stratum. [Figure 17]FIG. 17 is an explanatory diagram (part 4) showing an example of estimating the geological age of each stratum. [Figure 18] FIG. 18 is a flowchart showing an example of a generation processing procedure. [Figure 19] FIG. 19 is a flowchart showing an example of an estimation processing procedure. [Mode for Carrying Out the Invention]

[0010] Hereinafter, embodiments of the information processing apparatus, information processing method, and information processing program according to the present invention will be described in detail with reference to the drawings.

[0011] (Example of the information processing method according to the embodiment) FIG. 1 is an explanatory diagram showing an example of the information processing method according to the embodiment. The information processing apparatus 100 is a computer for estimating the age of a stratum. The information processing apparatus 100 is, for example, a server or a PC (Personal Computer).

[0012] The age of a stratum is, for example, a division of the era in which the stratum is considered to have been formed in the past. Specifically, the age of a stratum is a unit such as the Cretaceous period and the Jurassic period, or a finer-grained unit than the Cretaceous period and the Jurassic period. The age refers to the geological age.

[0013] Conventionally, there are cases where it is desired to estimate the age of a stratum. For example, in studies such as correlation of contemporaneous strata or rock samples for each region or well, geological structure interpretation or estimation of resource amount and storage capacity in exploration for oil and natural gas or survey of suitable sites for CO2 underground storage, and geological evaluation such as prediction of the depth and lithofacies of the target stratigraphic horizon during drilling, the age of strata is used as important knowledge.

[0014] Furthermore, for example, at an oil and gas drilling site, it is desirable to determine whether or not oil and gas is present in the geological formation at the drilling tip and to carry out drilling operations appropriately. For example, oil and gas is expected to be present in geological formations of a specific age. If the age of the geological formation being drilled can be estimated during drilling operations, it is expected that this will contribute to recognizing the discrepancy between the expected stratigraphy and the actual stratigraphy, correcting the predicted depth to be reached at the target layer, or predicting the physical characteristics of the layer being drilled, thereby contributing to the optimization of drilling operations.

[0015] In contrast, there are techniques to estimate the age of a geological layer based on the combination of microfossil types found in samples taken from the layer, such as surface or seabed sediments, outcrops, cores recovered through drilling or boring, or excavated material transported to the surface by drilling fluid. Excavated material is also called cuttings. The samples are rocks or sediments.

[0016] Microfossils are a general term for tiny fossils. Because microfossils appear and become extinct over relatively short periods, they can serve as a guide for estimating the age of geological strata. Specifically, microfossils include foraminifera, radiolaria, calcareous nannofossils, diatoms, pollen and spores, conodonts, and ostracods.

[0017] Specifically, when geological strata are deposited on the Earth's surface or on the seabed, the microfossils contained in the rocks (clastic materials) within the strata provide important clues about the age of the sample's deposition or the environment in which the rock sample was deposited. Therefore, the above technique creates slides containing samples taken from geological strata. Specifically, slides containing the sample are created by fixing clastic materials obtained by performing physical or chemical treatments on the sample (rock sample), such as crushing or removing impurities by acid dissolution, onto a slide with a transparent resin.

[0018] Subsequently, in the above technique, the operator visually inspects the slide using various microscopes, investigates the combination of types of microfossils appearing in the sample, and estimates the age of the stratum based on the investigated combinations. Examples of microscopes include conventional microscopes, optical microscopes, or polarizing microscopes. Preferably, the operator is a specialist with knowledge of microfossils. Specifically, the operator can estimate the depositional age or depositional environment of the sample. As mentioned above, the estimated depositional age or depositional environment can be used as important knowledge in studies such as correlation of contemporary stratum or rock samples for each region or well, interpretation of geological structures or estimation of resource amounts and storage capacity in oil and natural gas exploration or CO2 underground storage site surveys, and geological evaluation such as prediction of the depth and lithology of the target stratum during drilling.

[0019] On the other hand, conventional techniques have the problem of being unable to accurately estimate the age of geological strata. For example, there is the problem of increased costs when investigating combinations of types of microfossils that appear in samples.

[0020] Costs can be human, time, or financial resources. Specifically, costs include the expense of hiring experts or the time spent by experts investigating combinations of microfossil types found in samples. Furthermore, securing appropriate experts with knowledge of microfossils and the ability to accurately identify different types of microfossils can be challenging.

[0021] In response to this, a method can be conceivable that uses machine learning to estimate the age of a geological layer by utilizing a model that estimates the types of microfossils visible in the input image. The model could be, for example, a distance learner. In this method, for example, a group of fossil images containing microfossils is extracted from various microscope images of a sample taken from a geological layer. In this method, for example, the model is used to estimate the types of microfossils visible in each fossil image in the extracted group of fossil images, and the combination of microfossil types appearing in the sample is estimated.

[0022] In this method, for example, a score indicating the likelihood that a particular type of microfossil appears in a sample may be calculated for each type of microfossil, based on the likelihood estimated by the model for each type of microfossil captured in the fossil image. Specifically, for each type of microfossil, one or more fossil images containing that type of microfossil are sorted in descending order of likelihood, and the statistical value of the likelihood estimated by the model for the types of microfossils captured in the top N fossil images is calculated as the score. The statistical value may be, for example, a sum or an average. If the top N fossil images are missing, a penalty may be added to the score.

[0023] In this method, for example, a list is used to identify combinations of microfossil types that appear in a given geological layer for each age, and the age of the layer is estimated based on the estimated combinations of microfossil types according to a predetermined algorithm. Specifically, a score is calculated for each age based on the estimated combinations of microfossil types according to a predetermined algorithm, and the age with the highest calculated score is set as the age of the geological layer from which the sample was taken.

[0024] The predetermined algorithm calculates the score for a given age by subtracting the statistical value of the score of each type of microfossil that does not appear in the strata of a given age from the statistical value of the score of each type of microfossil that appears in the strata of that age. The statistical value is, for example, a sum or an average value. The predetermined algorithm may also correct the score for an age according to the estimated number of types of microfossils.

[0025] However, this method has a problem in that it is difficult to improve the predetermined algorithm for estimating the age of the strata. Specifically, it is difficult to incorporate existing knowledge into the predetermined algorithm. Existing knowledge consists of example data that shows which combinations of microfossil types correspond to which ages of strata.

[0026] Furthermore, this method has the problem that the plausibility of the estimated age of the geological layers is not clear. For example, the calculated score does not necessarily represent the plausibility of the estimated age of the geological layers. Therefore, it is difficult to determine how much confidence to have in the estimated age of the geological layers at an oil and gas drilling site and how to formulate future drilling plans accordingly.

[0027] Furthermore, this method has the problem that it is difficult to improve the model. For example, when training the model, training data is prepared that associates specimen images showing microfossils with the correct types of microfossils shown in those specimen images. In this case, it is desirable that experts with knowledge of microfossils identify the correct types of microfossils shown in the specimen images in order to prepare the training data. This leads to an increase in the cost of preparing the training data.

[0028] Costs can be human, time, or financial resources. Specifically, costs include the expense of hiring experts or the time spent by experts investigating combinations of microfossil types found in samples. Furthermore, securing suitable experts with knowledge of microfossils can be challenging.

[0029] Therefore, this embodiment describes an information processing method that can accurately estimate the age of geological strata.

[0030] In Figure 1, the information processing device 100 has a detection unit 101 and a classification unit 102. The detection unit 101 has the function of detecting a group of fossil images 112, each depicting a microfossil, contained in the geological sample image 111. The geological sample image 111 is, for example, a microscope image of a slide containing a sample taken from a geological stratum. Microfossils specifically include foraminifera, radiolarians, calcareous nannofossils, diatoms, pollen and spores, conodonts, or ostracods. Samples specifically include rocks and sediments. The various microscopes include, for example, a conventional microscope, an optical microscope, or a polarizing microscope.

[0031] Geological sample images 111 are, specifically, various microscopic images of slides containing samples, which were created by fixing clastic material obtained by physically or chemically pre-treating samples taken from geological strata onto a slide with transparent resin. Pre-treatment includes the removal of impurities by crushing or acid dissolution. Geological sample images 111 may also specifically be various microscopic images of slides containing samples, which were created by directly fixing samples taken from geological strata onto a slide with transparent resin. Each fossil image in the fossil image group 112 corresponds to a part of the geological sample image 111.

[0032] The classification unit 102 includes a first generator 121, a second generator 122, and a third generator 123. The first generator 121 has the function of generating an intermediate representation 113 based on the fossil image group 112. The second generator 122 has the function of generating type information 114 that shows the likelihood of each type of microfossil in each fossil image of the fossil image group 112, based on the intermediate representation 113. The third generator 123 has the function of generating age information 115 that shows the likelihood of each age of the strata corresponding to the fossil image group 112, based on the intermediate representation 113.

[0033] The classification unit 102 is trained using training data 130. The training data 130 is data that associates a group of specimen images 131, each depicting a microfossil, with information 132 indicating the correct type of microfossil in each specimen image and information 133 indicating the correct age of the geological layer corresponding to the group of specimen images 131.

[0034] The classification unit 102 receives, for example, a group of fossil images 112 detected by the detection unit 101. The classification unit 102 is trained to output, for example, type information 114 indicating the likelihood of each type of microfossil in each fossil image, and age information 115 indicating the likelihood of each age of the strata corresponding to the group of fossil images 112, based on the received group of fossil images 112.

[0035] (1-1) Specifically, the information processing device 100 acquires geological sample images 111. Specifically, the information processing device 100 acquires geological sample images 111 by receiving input of geological sample images 111 based on user input. Specifically, the information processing device 100 may acquire geological sample images 111 by receiving geological sample images 111 from another computer.

[0036] (1-2) The information processing device 100 uses the detection unit 101 to detect a group of fossil images 112, each containing a microfossil, that are included in the acquired geological sample image 111. This allows the information processing device 100 to reduce the cost incurred in detecting the group of fossil images 112. The information processing device 100 enables the detection of the group of fossil images 112 even if the user does not have knowledge of microfossils.

[0037] (1-3) The information processing device 100, using the classification unit 102, obtains type information 114 and age information 115 based on the detected fossil image group 112. The type information 114 indicates the likelihood of each type of microfossil in each fossil image of the detected fossil image group 112. The age information 115 indicates the likelihood of each age of the strata corresponding to the detected fossil image group 112.

[0038] The information processing device 100, for example, inputs the detected fossil image group 112 to the classification unit 102, thereby acquiring the type information 114 and age information 115 output by the classification unit 102. This allows the information processing device 100 to accurately estimate the age of the geological strata. The information processing device 100 can also reduce the cost associated with estimating the age of the geological strata.

[0039] The information processing device 100 can train the classification unit 102 by integrating the first generator 121, the second generator 122, and the third generator 123. Therefore, the information processing device 100 can reflect existing knowledge in the second generator 122, which generates type information 114 that shows the likelihood for each type of microfossil in each fossil image. As a result, the information processing device 100 can improve the accuracy of estimating the type of microfossil using the second generator 122.

[0040] Similarly, the information processing device 100 can incorporate existing knowledge into a third generator 123 that generates age information 115 indicating the likelihood for each age of the strata corresponding to the fossil image group 112. As a result, the information processing device 100 can improve the accuracy of estimating the age of the strata using the third generator 123.

[0041] The information processing device 100 can generate age information 115, which indicates the likelihood of each age of the strata corresponding to the fossil image group 112, using the third generator 123. Therefore, the information processing device 100 makes it easier for the user to grasp the likelihood of the estimated age of the strata. Consequently, the information processing device 100 makes it easier for the user to determine how much confidence to have in the estimated age of the strata at the drilling site and how to formulate future drilling plans.

[0042] Furthermore, how the training data 130 is prepared will be described later using Figures 5 to 13. When the information processing device 100 prepares the training data 130 as described later in Figures 5 to 13, it is possible to reduce the cost incurred in preparing the training data 130.

[0043] Here, we have described the case where the functions of the information processing device 100 are realized by a single computer, but this is not the only case. For example, the functions of the information processing device 100 may be realized through the collaboration of multiple computers. For example, the functions of the information processing device 100 may be realized on the cloud. Specifically, the case where the functions of the information processing device 100 are realized through the collaboration of multiple computers will be described later using Figure 2.

[0044] (An example of information processing system 200) Next, using Figure 2, we will describe an example of an information processing system 200 that realizes the functions of the information processing device 100 shown in Figure 1 through the cooperation of multiple computers.

[0045] Figure 2 is an explanatory diagram showing an example of an information processing system 200. In Figure 2, the information processing system 200 includes an image analysis device 201, an estimation processing device 202, and a client device 203.

[0046] In the information processing system 200, the image analysis device 201 and the estimation processing device 202 are connected via a wired or wireless network 210. The network 210 is, for example, a LAN (Local Area Network), a WAN (Wide Area Network), or the Internet. Also in the information processing system 200, the estimation processing device 202 and the client device 203 are connected via a wired or wireless network 210.

[0047] The estimation processing unit 202 is a computer for estimating the age of geological strata. The estimation processing unit 202 has a model corresponding to the classification unit 102. The estimation processing unit 202 has a model, for example, by training the model based on training data. The training data is data that associates a group of specimen images depicting microfossils, information indicating the correct type of microfossil in each specimen image in the group of specimen images, and information indicating the correct age of the geological strata corresponding to the group of specimen images. The estimation processing unit 202 may have a model, for example, by receiving the model from another computer. The other computer is, for example, a client device 203.

[0048] The estimation processing device 202 receives a processing request that requests the estimation of the age of a geological stratum. The processing request includes, for example, one or more geological sample images. The geological sample images are images corresponding to geological strata. The geological sample images are, for example, various microscopic images of slides containing samples taken from the geological strata.

[0049] Geological sample images are, specifically, various microscope images taken of slides containing samples, which are created by fixing clastic material obtained by physically or chemically pre-treating samples taken from geological strata onto a slide with transparent resin. Pre-treatment includes the removal of impurities by crushing or acid dissolution. Geological sample images may also be, specifically, various microscope images taken of slides containing samples, which are created by directly fixing samples taken from geological strata onto a slide with transparent resin.

[0050] The estimation processing unit 202 acquires a processing request, for example, by receiving a processing request from another computer. The other computer is, for example, a client device 203.

[0051] The estimation processing unit 202 acquires one or more geological sample images from the acquired processing request. The estimation processing unit 202 sends a detection request to the image analysis device 201, requesting the detection of fossil image groups contained in the geological sample images. Fossil images are regions of the geological sample image that show microfossils. The detection request includes, for example, one or more geological sample images. As a result of sending the detection request, the estimation processing unit 202 receives from the image analysis device 201 the results of detecting fossil image groups contained in each of the one or more geological sample images.

[0052] The estimation processing unit 202, for each geological sample image, uses a model to acquire type information indicating the likelihood of each type of microfossil in each fossil image, and age information indicating the likelihood of each age of the geological layer corresponding to the fossil image group, based on the detected fossil image group. The estimation processing unit 202 transmits the acquired type information and age information, associated with each geological sample image, to the client device 203. The estimation processing unit 202 is, for example, a server or a PC.

[0053] The image analysis device 201 is a computer for analyzing geological sample images. The image analysis device 201 corresponds to the detection unit 101 shown in Figure 1. The image analysis device 201 receives a detection request from the estimation processing device 202. The image analysis device 201 acquires one or more fossil images from the received detection request. For each of the one or more geological sample images, the image analysis device 201 detects the group of fossil images contained in that geological sample image. The image analysis device 201 transmits the results of detecting the group of fossil images contained in each of the one or more geological sample images to the estimation processing device 202. The image analysis device 201 is, for example, a server or a PC.

[0054] The client device 203 is a computer used by a user who wishes to estimate the age of a geological stratum. Based on the user's input, the client device 203 acquires one or more geological sample images. Based on the user's input, the client device 203 generates a processing request that includes one or more geological sample images and requests the estimation of the age of the geological stratum, and transmits it to the estimation processing device 202.

[0055] The client device 203 receives type information, indicating the likelihood of each type of microfossil in each fossil image contained in the geological sample image, and age information, indicating the likelihood of each age of the geological layer, from the estimation processing device 202, corresponding to each geological sample image. The client device 203 outputs the received type information and age information, corresponding to each geological sample image, so that the user can refer to them. The client device 203 may be, for example, a PC, a tablet terminal, or a smartphone.

[0056] This section describes a case where the estimation processing device 202 is a different device from the image analysis device 201, but it is not limited to this case. For example, the estimation processing device 202 may have the functionality of an image analysis device 201 and may operate as an image analysis device 201. In this case, the information processing system 200 does not need to include an image analysis device 201.

[0057] This section describes a case where the estimation processing unit 202 is a different device from the client device 203, but it is not limited to this case. For example, the estimation processing unit 202 may have the functionality of a client device 203 and may operate as a client device 203. In this case, the information processing system 200 does not need to include a client device 203.

[0058] This explanation describes a case where the image analysis device 201 is a different device from the client device 203, but it is not limited to this case. For example, the image analysis device 201 may have the functionality of a client device 203 and may operate as a client device 203. In this case, the information processing system 200 does not need to include a client device 203.

[0059] (Example of hardware configuration of information processing device 100) Next, an example of the hardware configuration of the information processing device 100 will be described using Figure 3.

[0060] Figure 3 is a block diagram showing an example of the hardware configuration of the information processing device 100. In Figure 3, the information processing device 100 includes a CPU (Central Processing Unit) 301, a memory 302, and a network interface 303. The information processing device 100 also includes a recording medium interface 304, a recording medium 305, a display 306, and an input device 307. Each component is connected by a bus 300.

[0061] Here, the CPU 301 is responsible for the overall control of the information processing device 100. The memory 302 includes, for example, ROM (Read Only Memory), RAM (Random Access Memory), and flash ROM. Specifically, for example, flash ROM and ROM store various programs, and RAM is used as the work area for the CPU 301. Programs stored in memory 302 are loaded into the CPU 301, causing the CPU 301 to execute the coded processes.

[0062] The network interface 303 is connected to network 210 via a communication line, and then connects to other computers via network 210. The network interface 303 manages the internal interface with network 210 and controls the input and output of data from other computers. The network interface 303 is, for example, a modem or a LAN adapter.

[0063] The recording medium interface (I / F) 304 controls the reading and writing of data to the recording medium 305 according to the control of the CPU 301. The recording medium interface (I / F) 304 is, for example, a disk drive, an SSD (Solid State Drive), or a USB (Universal Serial Bus) port. The recording medium 305 is a non-volatile memory that stores the data written under the control of the recording medium interface (I / F) 304. The recording medium 305 is, for example, a disk, semiconductor memory, or USB memory. The recording medium 305 may be detachable from the information processing device 100.

[0064] Display 306 displays data such as cursors, icons, toolboxes, documents, images, or functional information. Display 306 is, for example, a CRT (Cathode Ray Tube), a liquid crystal display, or an organic EL (Electroluminescence) display. Input device 307 has keys for inputting characters, numbers, or various instructions, and performs data input. Input device 307 is, for example, a keyboard or a mouse. Input device 307 may also be, for example, a touch panel input pad or a numeric keypad.

[0065] The information processing device 100 may have, in addition to the components described above, a camera, for example. Furthermore, the information processing device 100 may have, in addition to the components described above, a printer, scanner, microphone, or speaker, for example. Also, the information processing device 100 may have multiple recording medium interfaces 304 and recording mediums 305, for example. Furthermore, the information processing device 100 does not necessarily have, for example, a display 306 or an input device 307. Also, the information processing device 100 does not necessarily have, for example, recording medium interfaces 304 and recording mediums 305.

[0066] (Example hardware configuration of image analysis device 201) The hardware configuration example of the image analysis device 201 is the same as the hardware configuration example of the information processing device 100 shown in Figure 3, so the explanation is omitted.

[0067] (Example hardware configuration of the estimation processing unit 202) The hardware configuration example of the estimation processing unit 202 is the same as the hardware configuration example of the information processing unit 100 shown in Figure 3, so the explanation is omitted.

[0068] (Example hardware configuration for client device 203) The hardware configuration example for client device 203 is the same as the hardware configuration example for information processing device 100 shown in Figure 3, so the explanation is omitted.

[0069] The following explanation will primarily describe the case where the information processing device 100 operates independently, similar to Figure 1. As shown in Figure 2, an example of the information processing device 100 operating independently can be modified as appropriate to enable the functions of the information processing device 100 to be realized through the cooperation of multiple computers.

[0070] (Example of the functional configuration of the information processing device 100) Next, an example of the functional configuration of the information processing device 100 will be described using Figure 4.

[0071] Figure 4 is a block diagram showing an example of the functional configuration of the information processing device 100. The information processing device 100 includes a storage unit 400, an acquisition unit 401, a detection unit 402, a classification unit 403, a training unit 404, an estimation unit 405, and an output unit 406.

[0072] The storage unit 400 is implemented by a storage area such as the memory 302 or recording medium 305 shown in Figure 3. The following description will focus on the case where the storage unit 400 is included in the information processing device 100, but is not limited to this case. For example, the storage unit 400 may be included in a device different from the information processing device 100, and the contents of the storage unit 400 may be accessible from the information processing device 100.

[0073] The acquisition unit 401 to the output unit 406 function as an example of a control unit. Specifically, the acquisition unit 401 to the output unit 406 realize their functions, for example, by having the CPU 301 execute a program stored in a storage area such as the memory 302 or recording medium 305 shown in Figure 3, or by using the network I / F 303. The processing results of each functional unit are stored in a storage area such as the memory 302 or recording medium 305 shown in Figure 3.

[0074] The memory unit 400 stores various types of information that are referenced or updated during the processing of each functional unit. For example, the memory unit 400 stores geological sample images. Geological sample images are images corresponding to geological strata. Geological sample images are, for example, various microscope images taken of slides containing samples taken from geological strata.

[0075] Geological sample images are, specifically, various microscope images taken of slides containing samples, which are created by fixing clastic material obtained by physically or chemically pre-treating samples taken from geological strata onto a slide with transparent resin. Pre-treatment includes the removal of impurities by crushing or acid dissolution. Geological sample images may also be, specifically, various microscope images taken of slides containing samples, which are created by directly fixing samples taken from geological strata onto a slide with transparent resin.

[0076] Specifically, the memory unit 400 stores one or more training images of geological samples. Specifically, the memory unit 400 stores one or more training images of geological samples corresponding to strata of different ages. Specifically, the memory unit 400 stores one or more geological sample images for estimation. Specifically, the memory unit 400 may store multiple geological sample images for estimation that have an order in the depth direction. Geological sample images are acquired, for example, by the acquisition unit 401. Geological sample images are referenced, for example, by the detection unit 402.

[0077] The memory unit 400 stores, for example, a training population of specimen images depicting microfossils. Microfossils specifically include foraminifera, radiolarians, calcareous nannofossils, diatoms, pollen and spores, conodonts, or ostracods. Specimen images are, for example, images extracted from training geological sample images that depict microfossils. The population is, for example, a collection of specimen images depicting microfossils in strata of different ages. The population is acquired, for example, by the acquisition unit 401. The population may be generated, for example, by the detection unit 402. The population is referenced, for example, by the training unit 404.

[0078] The memory unit 400 stores, for example, one or more training data. Each training data set is data that associates a group of specimen images depicting microfossils with information indicating the correct type of microfossil in each specimen image in the group and information indicating the correct age of the geological layer corresponding to the group of specimen images. The group of specimen images is generated, for example, from a training population. The training data is acquired, for example, by the acquisition unit 401. The training data may also be generated, for example, by the training unit 404. The training data is referenced, for example, by the training unit 404.

[0079] The memory unit 400 stores, for example, a collection of fossil images for estimation, which are images of microfossils. The fossil images are, for example, images extracted from geological sample images for estimation that show areas containing microfossils. The fossil image collection is acquired, for example, by the acquisition unit 401. The fossil image collection may also be generated, for example, by the detection unit 402. The fossil image collection is referenced, for example, by the classification unit 403.

[0080] The memory unit 400 stores, for example, correspondence information indicating predetermined types of microfossils that appear in strata of at least a predetermined age. Specifically, the correspondence information shows, for each age, the combination of types of microfossils that appear in strata of that age. The correspondence information is acquired, for example, by the acquisition unit 401. The correspondence information is referenced, for example, by the training unit 404.

[0081] The acquisition unit 401 acquires various types of information used in the processing of each functional unit. The acquisition unit 401 stores the acquired information in the storage unit 400 or outputs it to each functional unit. The acquisition unit 401 may also output the information stored in the storage unit 400 to each functional unit. The acquisition unit 401 acquires various types of information, for example, based on user input. The acquisition unit 401 may also receive various types of information from a device other than the information processing device 100, for example.

[0082] The acquisition unit 401 acquires, for example, a training request that requests training of the classification unit 403. The training request may include, for example, one or more geological sample images for training. The training request may include, for example, a training population of specimen images of microfossils. The training request may include, for example, one or more training data. Specifically, the acquisition unit 401 acquires a training request by accepting input of a training request based on user operation input. Specifically, the acquisition unit 401 may acquire a training request by receiving a training request from another computer.

[0083] The acquisition unit 401 acquires, for example, one or more geological sample images for training. Specifically, the acquisition unit 401 acquires one or more geological sample images for training by extracting one or more geological sample images for training from a training request. Specifically, the acquisition unit 401 may acquire one or more geological sample images for training by accepting input of one or more geological sample images for training based on user operation input. Specifically, the acquisition unit 401 may acquire one or more geological sample images for training by receiving one or more geological sample images for training from another computer.

[0084] The acquisition unit 401 acquires, for example, a training population. Specifically, the acquisition unit 401 acquires a training population by extracting a training population from a training request. Specifically, the acquisition unit 401 may acquire a training population by accepting input of a training population based on user input. Specifically, the acquisition unit 401 may acquire a training population by receiving a training population from another computer.

[0085] The acquisition unit 401 acquires, for example, one or more training data. Specifically, the acquisition unit 401 acquires one or more training data by extracting one or more training data from a training request. Specifically, the acquisition unit 401 may acquire one or more training data by accepting input of one or more training data based on user operation input. Specifically, the acquisition unit 401 may acquire one or more training data by receiving one or more training data from another computer.

[0086] The acquisition unit 401 acquires, for example, an estimation request that the classification unit 403 requests to estimate the age of a geological stratum. The estimation request may include, for example, one or more geological sample images for estimation. The estimation request may also include, for example, a group of fossil images for estimation. Specifically, the acquisition unit 401 acquires estimation requests by accepting input of estimation requests based on user operation input. Specifically, the acquisition unit 401 may acquire estimation requests by receiving estimation requests from other computers.

[0087] The acquisition unit 401 acquires, for example, one or more geological sample images for estimation. Specifically, the acquisition unit 401 acquires one or more geological sample images for estimation by extracting one or more geological sample images for estimation from an estimation request. Specifically, the acquisition unit 401 may acquire one or more geological sample images for estimation by accepting input of one or more geological sample images for estimation based on user operation input. Specifically, the acquisition unit 401 may acquire one or more geological sample images for estimation by receiving one or more geological sample images for estimation from another computer.

[0088] The acquisition unit 401 acquires, for example, a group of fossil images for estimation. Specifically, the acquisition unit 401 acquires a group of fossil images for estimation by extracting a group of fossil images for estimation from an estimation request. Specifically, the acquisition unit 401 may acquire a group of fossil images for estimation by accepting input of a group of fossil images for estimation based on user input. Specifically, the acquisition unit 401 may acquire a group of fossil images for estimation by receiving a group of fossil images for estimation from another computer.

[0089] The acquisition unit 401 acquires, for example, correspondence information indicating a predetermined type of microfossil that appears in a geological layer of at least a predetermined age. Specifically, the acquisition unit 401 acquires correspondence information by receiving input of correspondence information based on user operation input. Specifically, the acquisition unit 401 may acquire correspondence information by receiving correspondence information from another computer.

[0090] The acquisition unit 401 may receive a start trigger to initiate processing in any of the functional units. A start trigger may be, for example, a predetermined operation input by a user. A start trigger may also be, for example, the reception of predetermined information from another computer. A start trigger may also be, for example, the output of predetermined information by any of the functional units. Specifically, the acquisition unit 401 may receive the acquisition of a training request as a start trigger to initiate processing in the detection unit 402 and the training unit 404. Specifically, the acquisition unit 401 may receive the acquisition of an estimation request as a start trigger to initiate processing in the detection unit 402 and the classification unit 403.

[0091] The detection unit 402 detects one or more sample images containing microfossils within each of the training geological sample images acquired by the acquisition unit 401. The detection unit 402 identifies regions within the geological sample image that have image features corresponding to microfossils, based on the results of image analysis of the geological sample image. The detection unit 402 detects one or more sample images that form the training population by extracting each of the identified regions as sample images. This allows the detection unit 402 to reduce the costs incurred when performing image analysis of geological sample images. Costs include human, time, and financial resources. The detection unit 402 can also make it easier to prepare the training population.

[0092] The detection unit 402 detects a group of fossil images for estimation, each containing microfossils, for each of the one or more geological sample images for estimation acquired by the acquisition unit 401. For example, the detection unit 402 identifies a region in the geological sample image that has image features corresponding to microfossils based on the results of image analysis of the geological sample image, and extracts the identified region as a fossil image for estimation, thereby detecting a group of fossil images for estimation.

[0093] The detection unit 402 detects, for example, a group of fossil images for estimation if the acquisition unit 401 does not acquire such a group. This allows the detection unit 402 to reduce the costs associated with image analysis of geological sample images. Costs include human, time, and financial resources. The detection unit 402 can also make it easier to prepare a group of fossil images for estimation.

[0094] The classification unit 403 includes a first generator, a second generator, and a third generator. The first generator has the function of generating an intermediate representation based on a group of fossil images. The intermediate representation is, for example, a vector. The second generator has the function of generating information indicating the likelihood of each type of microfossil in each fossil image in the group of fossil images based on the intermediate representation. The information indicating the likelihood of each type of microfossil may be information that shows a value of 1 or less for one type of likelihood, and 0 or undetermined for other types of likelihood. The third generator has the function of generating information indicating the likelihood of each age of the strata corresponding to the group of fossil images based on the intermediate representation.

[0095] The classification unit 403 specifically corresponds to the classification model 510 described later in Figures 5 and 6. The first generator specifically corresponds to the encoder 610 described later in Figure 6. The second generator specifically corresponds to the fossil species classification network 620 described later in Figure 6. The third generator specifically corresponds to the combination of the fossil aggregation network 630 and the age classification network 640 described later in Figure 6.

[0096] The third generator, for example, enables the generation of an aggregate vector 602 that includes information other than that used to identify fossil species, based on the feature vector 601, using the fossil aggregation network 630 described later in Figure 6. The third generator, for example, enables the determination of the age of a geological layer more accurately than when determining the age of a layer based solely on the combination of fossil species, using the age classification network 640 described later in Figure 6.

[0097] The training unit 404 generates a training population by collecting sample images of microfossils from strata of different ages. The training unit 404 generates a training population by collecting one or more sample images detected by the detection unit 402 for each of the training geological sample images, each corresponding to a strata of different ages.

[0098] The training unit 404 generates a training population, for example, if the acquisition unit 401 does not acquire a training population. The training unit 404 does not need to generate a training population if the acquisition unit 401 has already acquired a training population. The training unit 404 may, for example, integrate the training population acquired by the acquisition unit 401 with the generated training population to generate a new training population. This allows the training unit 404 to set a training population from which to generate training data.

[0099] The training unit 404 generates training data based on the training population acquired by the acquisition unit 401 or generated by the training unit 404, by referring to the correspondence information. The training unit 404 selects a predetermined age, for example. The training unit 404 may randomly select multiple predetermined ages, for example. The training unit 404 identifies a predetermined type of microfossil that appears in the strata of the selected predetermined age, by referring to the correspondence information. The training unit 404 sets a predetermined number of specimen images from the training population, for example, including specimen images of the identified predetermined type of microfossil. The training unit 404 generates training data by associating the set specimen image group with information indicating the correct type of microfossil in each specimen image of the specimen image group and information indicating a predetermined age which is the correct age of the strata corresponding to the specimen image group.

[0100] This allows the training unit 404 to prepare training data and enable the classification unit 403 to be trained. The training unit 404 can reduce the costs incurred in preparing training data. Costs include human, time, and financial resources. The training unit 404 can refer to correspondence information and prepare training data corresponding to a predetermined age, regardless of the age of the strata shown in the geological sample image that forms the basis of the sample image. Therefore, the training unit 404 can easily increase the number of training data prepared.

[0101] The training unit 404 may, for example, set a predetermined number of sample images randomly selected from a training population so that sample images of a specified type of microfossil are selected with a higher probability than sample images of microfossils of other types. The training unit 404 generates training data that associates the set sample image set with information indicating the correct type of microfossil in each sample image of the sample image set and information indicating a predetermined age which is the correct age of the geological layer corresponding to the sample image set.

[0102] This allows the training unit 404 to prepare training data that includes specimen images of microfossils other than the specified type as noise. Therefore, the training unit 404 can train the classification unit 403 with high accuracy. The training unit 404 can reduce the costs incurred in preparing training data. Costs include human, time, and financial resources.

[0103] The training unit 404 trains the classification unit 403 using training data acquired by the acquisition unit 401 or generated by the training unit 404. The training unit 404 trains the classification unit 403 so that, in response to a group of fossil images being input, it outputs information indicating the likelihood of each type of microfossil in each fossil image, and information indicating the likelihood of each age of the geological layer corresponding to the group of fossil images. Training is performed, for example, by backpropagation.

[0104] As a result, the training unit 404 can train the classification unit 403 and improve the accuracy of the information indicating the likelihood of each type of microfossil in each fossil image output by the classification unit 403, and the information indicating the likelihood of each age of the strata corresponding to the group of fossil images. The training unit 404 can, for example, train the classification unit 403 by integrating the first generator, the second generator, and the third generator. Therefore, the training unit 404 can reflect existing knowledge in the third generator. As a result, the information processing device 100 can improve the accuracy of estimating the age of the strata using the third generator.

[0105] The classification unit 403 accepts input of a group of fossil images for estimation acquired by the acquisition unit 401 or detected by the detection unit 402 for any of the geological layers. The classification unit 403 may, for example, accept input of a group of fossil images for estimation acquired by the acquisition unit 401 or detected by the detection unit 402 for each of a group of geological layers that have a sequence in the depth direction.

[0106] The classification unit 403 outputs information indicating the likelihood for each type of microfossil in each fossil image in the group of fossil images, and information indicating the likelihood for each age of the strata corresponding to the group of fossil images, in response to the input of a group of fossil images for estimation. For example, the classification unit 403 outputs information indicating the likelihood for each type of microfossil in each fossil image and information indicating the likelihood for each age of a strata in a group of strata that have an order in the depth direction.

[0107] Specifically, the classification unit 403 generates an intermediate representation for any of the geological strata using the first generator based on the set of fossil images for estimation that it has received as input. Specifically, the classification unit 403 generates information showing the likelihood of each type of microfossil in each fossil image using the second generator based on the intermediate representation for any of the geological strata. Specifically, the classification unit 403 generates information showing the likelihood of each age of any of the geological strata using the third generator based on the intermediate representation for any of the geological strata.

[0108] Specifically, the classification unit 403 outputs information indicating the likelihood of each type of microfossil in each generated fossil image and information indicating the likelihood of each age of the stratum. This allows the classification unit 403 to estimate the type of microfossil in each fossil image and the age of the stratum. The classification unit 403 can reduce the cost incurred in estimating the age of the stratum. The classification unit 403 can make it easier for the user to understand the likelihood of the estimated age of the stratum.

[0109] The estimation unit 405 estimates the age of the strata corresponding to the fossil image group based on the information output by the classification unit 403, which shows the likelihood of each age of the strata corresponding to the fossil image group. For example, the estimation unit 405 can estimate the age with the highest likelihood as the age of the strata corresponding to the fossil image group. This allows the estimation unit 405 to enable the user to understand the age of the strata corresponding to the fossil image group without having to refer to the information showing the likelihood of each age of the strata corresponding to the fossil image group.

[0110] The estimation unit 405 may estimate the age of each of a group of strata that have an order in the depth direction, based on the information indicating the likelihood of each age of the strata output by the classification unit 403. For example, the estimation unit 405 estimates the age of each strata such that the age of the relatively shallower strata is earlier than the age of the relatively deeper strata. This allows the estimation unit 405 to improve the accuracy of estimating the age of each strata by taking into account the order in the depth direction.

[0111] The output unit 406 outputs the processing result of at least one of the functional units. The output format can be, for example, display on a screen, print to a printer, transmit to an external device via the network interface 303, or store in a storage area such as the memory 302 or recording medium 305. This allows the output unit 406 to notify the user of the processing result of at least one of the functional units, thereby improving the usability of the information processing device 100.

[0112] The output unit 406 outputs, for example, information indicating the likelihood of each type of microfossil in each fossil image, and information indicating the likelihood of each age of the said geological layer, which were generated by the classification unit 403, for any given geological layer. Specifically, the output unit 406 outputs information indicating the likelihood of each type of microfossil in each fossil image and information indicating the likelihood of each age of the said geological layer, in association with any given geological layer, so that the user can refer to them.

[0113] Specifically, the output unit 406 may transmit to another computer information indicating the likelihood of each type of microfossil in each fossil image, and information indicating the likelihood of each age in that fossil layer, in association with any of the geological layers. This allows the output unit 406 to access externally the information indicating the likelihood of each type of microfossil in each fossil image and the information indicating the likelihood of each age in that fossil layer.

[0114] The output unit 406 outputs, for example, the age of any of the geological layers that corresponds to the group of fossil images, as estimated by the estimation unit 405. Specifically, the output unit 406 outputs the age of any of the geological layers that corresponds to the group of fossil images, as estimated by the estimation unit 405, in a way that can be referenced by the user. Specifically, the output unit 406 may transmit the age of any of the geological layers that corresponds to the group of fossil images, as estimated by the estimation unit 405, to another computer. This makes it possible for the estimation unit 405 to make the age of any of the geological layers that corresponds to the group of fossil images, as estimated by the estimation unit 405, accessible externally.

[0115] The output unit 406 outputs, for example, the age of each of the multiple strata that have a sequence in the depth direction, which was estimated by the estimation unit 405. Specifically, the output unit 406 outputs the age of each of the multiple strata that have a sequence in the depth direction, which was estimated by the estimation unit 405, so that it can be referenced by the user. Specifically, the output unit 406 may transmit the age of each of the multiple strata that have a sequence in the depth direction, which was estimated by the estimation unit 405, to another computer. This makes it possible for the estimation unit 405 to make the age of each of the strata that it estimated to have been estimated by the estimation unit 405 accessible externally.

[0116] Here, we have described a case where the information processing device 100 includes an acquisition unit 401, a detection unit 402, a classification unit 403, and a training unit 404, but it is not limited to this. For example, the information processing device 100 may not include any of the functional units. Specifically, the information processing device 100 may not include a detection unit 402. In this case, the information processing device 100 cooperates with another computer that includes a detection unit 402.

[0117] (Example of operation of the information processing device 100) Next, we will explain an example of the operation of the information processing device 100 using Figures 5 to 17. For convenience, in this example, we will assume that the microfossils are "calcareous nannofossils." First, we will explain an example of the classification model 510 used by the information processing device 100 using Figures 5 and 6.

[0118] Figures 5 and 6 are explanatory diagrams showing an example of the classification model 510. In Figure 5, the information processing device 100 has an object detector 500 and a classification model 510. The object detector 500 receives a geological sample image as an input image 521. The geological sample image is an image corresponding to a geological layer. The geological sample image is, for example, various microscope images taken of slides containing samples collected from geological layers.

[0119] The object detector 500 has the function of detecting a group of candidate fossil images 530 that are determined to contain calcareous nannofossils, by identifying regions in the input image 521 that have image features corresponding to calcareous nannofossils. Candidate fossil images 531 are, for example, regions in the input image 521 that show objects that are thought to be calcareous nannofossils. Candidate fossil images 531 may contain objects that are not calcareous nannofossils.

[0120] In the example shown in Figure 5, the object detector 500 specifically receives an input image set 520, which includes, as input image 521, geological sample images corresponding to each of a plurality of strata having a sequence in the depth direction. Specifically, the object detector 500 detects a group of candidate fossil images 530 contained in each input image 521.

[0121] The classification model 510 is a single-depth AI (Artificial Intelligence) model. The classification model 510 is trained through multi-task learning. Multi-task learning involves shaping and training the classification model 510 to perform multiple related tasks. Specific examples of training will be described later using Figures 7 to 11.

[0122] The classification model 510 receives input of a group of candidate fossil images 530 detected by the object detector 500 for any of the geological layers. Based on the received group of candidate fossil images 530 for any of the geological layers, the classification model 510 outputs fossil species information 541 indicating the likelihood for each fossil species corresponding to each candidate fossil image 531, and age information 542 indicating the likelihood for each geological age of the said geological layer.

[0123] The fossil species information 541 specifically has fields for nannofossil species and fossil species score. The nannofossil species field is set to the calcareous nannofossil fossil species. The fossil species score field is set to the fossil species score indicating the likelihood of the fossil species corresponding to the fossil candidate image 531. The age information 542 specifically has fields for age, age score, and number of detections. The age field is set to the geological age. The age score field is set to the age score indicating the likelihood of the geological age. The number of detections field is set to the number of detections indicating the likelihood of the overall age score.

[0124] In the example shown in Figure 5, the classification model 510 specifically receives input of a group of fossil candidate images 530 detected by the object detector 500 for each of several strata that have an order in the depth direction. Specifically, for each strata, the classification model 510 outputs fossil species information 541 indicating the likelihood for each fossil species corresponding to each fossil candidate image 531, and age information 542 indicating the likelihood for each geological age of the strata. Now, we will move on to the explanation of Figure 6 and describe the contents of the classification model 510.

[0125] In Figure 6, the classification model 510 includes an encoder 610, a fossil species classification network 620, a fossil aggregation network 630, and a chronological classification network 640. The encoder 610 has the function of generating an intermediate representation, a feature vector 601, based on a group of candidate fossil images 530.

[0126] The fossil species classification network 620 has the function of generating fossil species information 541 by generating a fossil species corresponding to each fossil candidate image 531 and the likelihood of that fossil species based on the feature vector 601. The fossil aggregation network 630 has the function of generating an aggregation vector 602 based on the feature vector 601. The fossil aggregation network 630 aggregates, for example, the features contained in any number of feature vectors 601 output by the first generator from any number of fossil candidate image groups 530 into content suitable for input to the age classification network 640. Suitable content refers to features of the appearance of the fossil, as well as features of the combination of fossil candidate image groups themselves. In other words, the fossil aggregation network 630 can generate an aggregation vector 602 that includes information other than that used to identify the fossil species. Since experts estimate ages based on fossil combinations, considering combinations of candidate fossil images is crucial for accuracy. It also helps absorb errors in the candidate fossil images output by the object detector 500, making the system more robust.

[0127] The age classification network 640 has the function of generating age information 542 by generating the geological age of any stratum corresponding to the fossil candidate image group 530 and the likelihood of that geological age, based on the aggregation vector 602. The age classification network 640 may further have the function of generating the number of detections based on the aggregation vector 602. The age classification network 640 makes it possible to determine the age of a stratum more accurately than when determining the age of a stratum based only on the combination of fossil species.

[0128] This allows the information processing device 100 to directly estimate the geological age of a stratum based on a group of candidate fossil images 530, using the classification model 510. In this way, the information processing device 100 can reduce the costs associated with estimating the geological age of a stratum by utilizing machine learning. These costs include human, time, and financial resources.

[0129] The information processing device 100 can simultaneously learn the correct geological age of strata and the correct fossil species by forming and training a classification model 510 through multitasking learning. Therefore, the information processing device 100 can realize fossil species estimation that takes into account the importance of each element to geological age estimation, thereby improving the accuracy of fossil species estimation.

[0130] Furthermore, the information processing device 100 can improve the fossil aggregation network 630, which corresponds to the algorithm for estimating geological ages, and the age classification network 640 through multitasking learning. The information processing device 100 can, for example, reflect existing knowledge into the fossil aggregation network 630, which corresponds to the algorithm for estimating geological ages, and the age classification network 640.

[0131] Furthermore, the information processing device 100 can clarify the plausibility of the estimated geological age. Therefore, the information processing device 100 can enable the user to appropriately determine how much confidence they can have in the estimated geological age. The information processing device 100 can easily eliminate arbitrariness in manually set algorithms by utilizing the fossil aggregation network 630, which corresponds to the algorithm for estimating geological ages, and the age classification network 640.

[0132] The information processing device 100 can separately estimate the fossil species corresponding to each candidate fossil image 531 and estimate the geological age of any of the strata corresponding to the group of candidate fossil images 530. Therefore, if the likelihood of the estimated geological age falls below a certain threshold, the information processing device 100 can output an alert prompting the user to estimate the geological age by another means, without displaying the estimated geological age. The information processing device 100 makes it easier for the user to appropriately estimate the geological age of strata.

[0133] Traditionally, there has been a problem in that the cost of preparing training data for machine learning to estimate geological ages tends to increase. For example, one possible method of preparing training data is for experts to visually examine hundreds of geological sample images and determine the correct fossil species and geological age corresponding to each fossil image in the geological sample image.

[0134] Therefore, it is conceivable that the information processing device 100 can reduce the cost of preparing training data by generating training data for training the classification model 510. Here, an example of how the information processing device 100 generates training data for training the classification model 510 will be explained using Figures 7 to 11.

[0135] Figures 7 to 9 are explanatory diagrams illustrating an example of generating training data. In the example shown in Figures 7 to 9, the information processing device 100 aims to prepare a relatively large amount of training data by combining a relatively small amount of real-world data. Furthermore, the information processing device 100 aims to prepare plausible training data that matches the real-world data.

[0136] Here, a given calcareous nannofossil of a specific fossil species may have the property of appearing in common in strata of different geological ages. For example, a given calcareous nannofossil of a specific fossil species may appear in common in geological age x and geological age y. In this case, an example of a fossil image corresponding to geological age x can also be used as an example of a fossil image corresponding to geological age y. Furthermore, different calcareous nannofossils of the same fossil species tend to have common characteristics. The information processing device 100 aims to reduce the cost incurred when preparing training data by utilizing each of the above properties.

[0137] In Figure 7, the information processing device 100 acquires fossil images of each fossil species as actual data, as shown in Table 700. The fossil images are associated with fossil species. The fossil images correspond to areas detected using the object detector 500 from training geological sample images corresponding to strata of any geological age, for example. The fossil images may also correspond to areas detected by experts from training geological sample images corresponding to strata of any geological age, for example.

[0138] The information processing device 100, for example, acquires fossil images as real data corresponding to areas containing calcareous nannofossils detected by experts from training geological sample images, based on user input. The information processing device 100 may also acquire fossil images of any fossil species as real data by, for example, using an object detector 500 to detect areas containing calcareous nannofossils from training geological sample images. Next, we will move on to the explanation of Figure 8.

[0139] In Figure 8, the information processing device 100 acquires trend information 800 representing the occurrence trends of fossil species by geological age. The trend information 800 has, for example, fields for fossil species name, age I, age II, age III, age IV, and age V. The field for fossil species name is set with the fossil species name, which is identification information that identifies the fossil species.

[0140] The field for Age I sets the frequency of occurrence of the above-mentioned calcareous nannofossil species in strata of geological age I. The frequency can be, for example, none, rare, common, or abundant. The field for Age II sets the frequency of occurrence of the above-mentioned calcareous nannofossil species in strata of geological age II. The field for Age III sets the frequency of occurrence of the above-mentioned calcareous nannofossil species in strata of geological age III.

[0141] The field for Age IV sets the frequency of occurrence of the above-mentioned calcareous nannofossil species in strata of geological age IV. The field for Age V sets the frequency of occurrence of the above-mentioned calcareous nannofossil species in strata of geological age V. Next, we will move on to the explanation of Figure 9.

[0142] In Figure 9, the information processing device 100 generates pseudo-chronological data 900 to be used as training data based on the fossil images of each fossil species, which are the actual data shown in Table 700, and the trend information 800. The information processing device 100 generates the pseudo-chronological data 900 to be used as training data by performing a series of processes, for example, as shown in (9-1), (9-2), (9-3), (9-4), (9-5), and (9-6) below.

[0143] (9-1) The information processing device 100 randomly determines the target geological age and the total number of fossil image samples in the input data 910. The total number of samples may be fixed.

[0144] (9-2) The information processing device 100 sets the occurrence rate in the input data 910 for each fossil species based on the trend information 800. For example, the information processing device 100 sets the occurrence rate in the input data 910 for each fossil species such that the occurrence rate in the input data 910 for fossil species that appear in the target geological period is higher than the occurrence rate in the input data 910 for fossil species that do not appear in the target geological period.

[0145] Specifically, the information processing device 100 sets the occurrence rate in the input data 910 for each fossil species so that the sum of the occurrence rates in the input data 910 for each fossil species equals 100%. In this case, the information processing device 100 specifically sets the occurrence rate in the input data 910 for fossil species with no frequency to 0%. Furthermore, the information processing device 100 specifically sets the occurrence rate in the input data 910 for fossil species with a rare frequency to a range of over 0% and up to 10%. Furthermore, the information processing device 100 specifically sets the occurrence rate in the input data 910 for fossil species with a normal frequency to a range of over 10% and up to 30%. Furthermore, the information processing device 100 specifically sets the occurrence rate in the input data 910 for fossil species with a high frequency to a range of over 30% and up to 100%.

[0146] The information processing device 100 sets the number of fossil images for each fossil species according to the occurrence rate in the input data 910 for each set fossil species, so that the sum of the number of fossil images for each fossil species equals the total number of fossil images determined. The information processing device 100 selects a group of fossil images equal to the total number of samples for each fossil species by selecting the set number of fossil images for each fossil species from the acquired fossil images for each fossil species. The information processing device 100 sets the selected group of fossil images as the sample image group to be included in the input data 910 for training data.

[0147] (9-3) The information processing device 100 sets the ground truth data 920 for the fossil species classification network 620 based on the fossil species group, which is a collection of fossil species corresponding to each fossil image in the selected fossil image group. The ground truth data 920 for the fossil species classification network 620 represents the correct answer which is the fossil species information 541 corresponding to the fossil image group. The ground truth data 920 for the fossil species classification network 620 represents, for example, the correct answer for each fossil species corresponding to each fossil image. The ground truth data 920 for the fossil species classification network 620 may also represent, for example, the likelihood for each fossil species corresponding to each fossil image that has the highest likelihood for the correct answer for each fossil image.

[0148] (9-4) The information processing device 100 calculates age scores for each geological period according to predetermined calculation rules. The age scores are continuous values. The calculation rules include calculating the age score of a geological period to be as close to 1 as the proportion of fossil species appearing in any particular geological period in the selected group of fossil images increases. The calculation rules also include calculating the age score of a geological period to be as close to 0 as the proportion of fossil species not appearing in any particular geological period in the selected group of fossil images increases.

[0149] The calculation rules include calculating the age score for each geological period to be closer to 0 the greater the proportion of fossil images other than calcareous nannofossils in the selected set of fossil images. The calculation rules also include calculating the age score for any given geological period to be closer to 0 the greater the number of fossil species that appear in any given geological period but do not appear in the selected set of fossil images. The information processing device 100 sets the ground truth data 930 for the age classification network 640 based on a vector that includes the calculated age scores. The ground truth data 930 for the age classification network 640 represents, for example, the age score for each geological period.

[0150] (9-5) The information processing device 100 generates pseudo-chronological data 900 to be used as training data by associating the set input data 910 with the set ground truth data 920 of the set fossil species classification network 620 and the set ground truth data 930 of the set chronological classification network 640.

[0151] (9-6) The information processing device 100 returns to the process in (9-1) and starts generating training data again until it has generated a predetermined number of training data. The predetermined number is set in advance by the user, for example. This allows the information processing device 100 to reduce the cost incurred in preparing training data.

[0152] Since the information processing device 100 utilizes fossil images of each fossil species, which are real data, it can prepare useful training data that is true to reality, from the perspective of training the classification model 510. Regardless of which geological age the fossil image of each fossil species is based on, the information processing device 100 can adopt the fossil image as an example of a fossil image corresponding to two or more appropriate geological ages, based on the trend information 800. For this reason, the information processing device 100 can easily prepare training data even with a small amount of real data.

[0153] When setting the input data 910, the information processing device 100 can include fossil images of fossil species that do not appear in the target geological period as noise. Therefore, the information processing device 100 can prepare useful training data that is true to reality, from the perspective of training the classification model 510. For example, the information processing device 100 can prepare training data to improve the robustness of the classification model 510.

[0154] The information processing device 100 can calculate age scores as continuous values ​​for each geological age according to predetermined calculation rules. Therefore, the information processing device 100 can calculate age scores for each geological age by reflecting not only the combination of fossil species appearing in the input data 910, but also the abundance of fossil images of each fossil species appearing in the input data 910. As a result, the information processing device 100 can prepare useful training data from the perspective of training the classification model 510. An example of the generation process procedure for generating training data is described later in Figure 18.

[0155] Next, we will explain specific examples of the pseudo-age data 900 used as training data using Figures 10 and 11.

[0156] Figures 10 and 11 are explanatory diagrams showing specific examples of the pseudo-chronological data 900 used as training data. In Figure 10, Table 1000 shows the number of fossil image samples for each fossil species for each pseudo-chronological data 900, as set by the information processing device 100. The leftmost column of Table 1000 is the data_id that identifies the pseudo-chronological data 900. The columns from the second column onward in Table 1000 are the number of fossil image samples for each fossil species. Each row in Table 1000 corresponds to one pseudo-chronological data 900. Next, we will move on to the explanation of Figure 11.

[0157] In Figure 11, Table 1100 shows the age scores for each geological period calculated by the information processing device 100. The leftmost column of Table 1100 is the data_id that identifies the pseudo-age data 900. The columns from the second column onward in Table 1100 are the age scores for each geological period. Each row in Table 1000 corresponds to one pseudo-age data 900. Specifically, as shown in Table 1000, the pseudo-age data 900 includes fossil images for the number of samples of each fossil species as input data 910, and as shown in Table 1100, it includes age scores for each geological period as ground truth data 930 for the age classification network 640.

[0158] The information processing device 100 trains the classification model 510 using the prepared training data. This improves the accuracy with which the classification model 510 generates fossil species information 541, which shows the likelihood for each fossil species corresponding to each candidate fossil image 531, and age information 542, which shows the likelihood for each geological age of the stratum.

[0159] Furthermore, the information processing device 100 can train the classification model 510 to reflect the age scores for each geological age represented by the ground truth data 930 of the age classification network 640. Therefore, the information processing device 100 can clarify the plausibility of the estimated geological age. Consequently, the information processing device 100 can enable the user to appropriately determine how much confidence they can place in the estimated geological age.

[0160] The information processing device 100 can improve the robustness of the classification model 510. Specifically, the information processing device 100 can improve the accuracy with which the classification model 510 generates fossil species information 541 and age information 542, even when the group of fossil candidate images 530 contains a fossil candidate image 531 that has been misdetected by the object detector 500.

[0161] Subsequently, the information processing device 100 uses the trained classification model 510 to estimate the geological age of a stratum based on geological sample images corresponding to each of a plurality of strata that have a sequence in the depth direction. For example, the information processing device 100 uses the object detector 500 to obtain a group of candidate fossil images 530 contained in the input image 521, using the geological sample image corresponding to each stratum as the input image 521.

[0162] For example, the information processing device 100 inputs the acquired group of candidate fossil images 530 for each geological layer into the trained classification model 510. As a result, the information processing device 100 obtains, for each geological layer, fossil species information 541 indicating the likelihood for each fossil species corresponding to each candidate fossil image 531 in the acquired group of candidate fossil images 530, and age information 542 indicating the likelihood for each geological age of the layer.

[0163] The information processing device 100 estimates the geological age of each geological layer based on the acquired age information 542, which shows the likelihood for each geological age of that layer. Specifically, the information processing device 100 estimates the geological age of each layer as the one with the highest likelihood age score. This allows the information processing device 100 to accurately estimate the geological age of each layer.

[0164] Here, we have described a case in which the information processing device 100 estimates the geological age of a stratum without referring to age information 542 that shows the likelihood for each geological age other than that stratum, but it is not limited to this. For example, the information processing device 100 may also estimate the geological age of a stratum by referring to the order in the depth direction of multiple stratums and using age information 542 that shows the likelihood for each geological age other than that stratum. This allows the information processing device 100 to improve the accuracy of estimating the geological age of a stratum.

[0165] Next, using Figures 12 to 17, we will explain how the information processing device 100 improves the accuracy of estimating the geological age of each layer by referring to the order in the depth direction of multiple layers and estimating the geological age of each layer.

[0166] Figures 12 and 13 are explanatory diagrams illustrating an example of estimating the geological age of each stratum. In Figure 12, assume that there are geological ages NN1, NN2, NN3, NN4, NN5, NN6, and NN7. NNx (x=1~7) is such that a larger x indicates a more recent geological age.

[0167] Furthermore, in Figure 12, assume that there are multiple layers with a sequence in the depth direction. In the example in Figure 12, there is a first layer at depths of 10000-10100. There is also a second layer at depths of 11000-11100. There is also a third layer at depths of 12000-12100. There is also a fourth layer at depths of 13000-13100. And there is also a fifth layer at depths of 14000-14100.

[0168] Here, since multiple strata are stacked on top of each other, it is considered that the geological age of relatively shallow strata is later than the geological age of relatively deep strata. "Later" includes the same geological age. Therefore, the information processing device 100 aims to improve the accuracy of estimating the geological age of each strata based on the above property.

[0169] In the example shown in Figure 12, the information processing device 100 uses the classification model 510 to obtain age scores indicating the likelihood of each geological age for each layer at each depth, as shown in Table 1200. Each hatch in Table 1200 classifies the magnitude of the age score. The upward-sloping hatch in Table 1200 is assigned to the largest age score. The geological age of the largest age score is the geological age estimated at a single depth. Next, we will move on to the explanation of Figure 13.

[0170] In Figure 13, the information processing device 100 sets up all geological age groups by combining geological ages at each depth such that the geological age increases with depth. A geological age group is, for example, a combination of geological age NN7 of the first depth layer, geological age NN6 of the second depth layer, geological age NN5 of the third depth layer, geological age NN4 of the fourth depth layer, and geological age NN3 of the fifth depth layer.

[0171] A geological age group may be, for example, a combination of geological age NN7 of the first depth layer, geological age NN7 of the second depth layer, geological age NN3 of the third depth layer, geological age NN2 of the fourth depth layer, and geological age NN1 of the fifth depth layer. A geological age group may also be, for example, a combination of geological age NN7 of the layers at each depth.

[0172] For example, a combination such as the geological age NN5 of the first depth layer, the geological age NN7 of the second depth layer, the geological age NN3 of the third depth layer, the geological age NN2 of the fourth depth layer, and the geological age NN1 of the fifth depth layer is unlikely to form a geological age group, except in special cases such as when older layers are located on top of younger layers due to a reverse fault, or when sediments originating from older layers are redeposited on top of younger layers. Specifically, since the geological age NN7 of the second depth layer, which is deeper than the first depth layer, is newer than the geological age NN5 of the first depth layer, this combination contradicts the above property and is therefore unlikely to form a geological age group.

[0173] The information processing device 100 calculates an integrated score for all set geological age groups by accumulating the age scores of the geological ages of the strata at each depth within each geological age group. In the following explanation, the calculation of the integrated score may be referred to as "inter-depth correction". Based on the calculated integrated score, the information processing device 100 estimates the geological age of the strata at each depth. For example, the information processing device 100 estimates the geological age of the strata at each depth as the geological age of the strata in the geological age group with the largest integrated score.

[0174] Furthermore, with respect to each geological layer at a given depth, if there is another geological layer with the same age score as the estimated geological layer, the information processing device 100 may further estimate that other geological layer as the geological layer's age.

[0175] In the example shown in Figure 13, it is assumed that the cumulative score of the geological age group consisting of geological age NN6 of the first depth layer, geological age NN6 of the second depth layer, geological age NN4 of the third depth layer, geological age NN2 of the fourth depth layer, and geological age NN1 of the fifth depth layer is the highest. Therefore, the information processing device 100 estimates geological ages NN6 of the first depth layer, geological age NN6 of the second depth layer, geological age NN4 of the third depth layer, geological age NN2 of the fourth depth layer, and geological age NN1 of the fifth depth layer as the geological ages of each layer.

[0176] Furthermore, the information processing device 100 further estimates geological ages NN4 and NN5, which have the same age score as geological age NN6 of the second depth layer, as the geological ages of the second depth layer. In the example in Figure 13, as shown in Table 1300, the geological ages of each layer estimated by the information processing device 100 through inter-depth correction are the geological ages enclosed by thick rectangles.

[0177] As a result, the information processing device 100 can improve the accuracy of estimating the geological age of a stratum at a given depth by referring to the order in the depth direction of multiple strata and considering the entire age information 542 that shows the likelihood for each geological age of a stratum at each depth. In an actual well drilling site, it is conceivable that as drilling progresses, samples from multiple strata at different depths are obtained, and it is desirable to estimate the geological age of each stratum. The information processing device 100 can be applied to this situation, accurately estimate the age of each stratum, and support drilling operations. Next, we will move on to the explanation of Figures 14 to 17.

[0178] Figures 14 to 17 are explanatory diagrams illustrating an example of estimating the geological age of each stratum. In the example shown in Figures 14 to 17, it is assumed that during the drilling of wells 1 and 2, it was desired to estimate the geological age of the stratum at each depth of each well using the information processing device 100.

[0179] In Figure 14, the information processing device 100 uses the classification model 510 to obtain age scores indicating the likelihood of each geological age for each layer at each depth in well 1, as shown in Table 1400. In Table 1400, the geological ages enclosed in thick rectangles are the correct geological ages for each layer at each depth. Each hatch indicates a geological age where the age score is above a threshold. The dotted hatch indicates the geological age where the age score is maximum. Next, we will move on to the explanation of Figure 15.

[0180] In Figure 15, Table 1500 shows the number of fossil images of each fossil species in the strata at each depth in well 1. Dotted hatches indicate non-index fossil species. Diagonal hatches indicate fossil species that cannot be found in the correct geological age. Cross-hatched hatches indicate fossil species that can be found in the correct geological age but were not found in expert surveys. Next, we will move on to the explanation of Figure 16.

[0181] In Figure 16, the information processing device 100 uses the classification model 510 to obtain age scores indicating the likelihood of each geological age for each layer at each depth in well 2, as shown in Table 1600. In Table 1600, the geological ages enclosed in thick rectangles are the correct geological ages for each layer at each depth. Each hatch indicates a geological age where the age score is above a threshold. The dotted hatch indicates the geological age where the age score is maximum. Next, we will move on to the explanation of Figure 17.

[0182] In Figure 17, Table 1700 shows the number of fossil images of each fossil species in the strata at each depth in well 2. Dotted hatches indicate non-index fossil species. Diagonal hatches indicate fossil species that cannot be found in the correct geological age. Cross-hatched hatches indicate fossil species that can be found in the correct geological age but were not found in expert surveys. Next, we will move on to the explanation of Figure 16.

[0183] As shown in Tables 1500 and 1700, fossil images of fossil species that do not appear in the correct geological age may appear incorrectly. Therefore, as shown in Tables 1400 and 1600, the geological age with the highest age score may not coincide with the correct geological age. In response to this, the information processing device 100 can improve the accuracy of estimating the geological age of a stratum at each depth by referring to the order in the depth direction of multiple strata and considering the overall likelihood of each geological age for the stratum at each depth. An example of the estimation processing procedure for estimating the geological age of a stratum at each depth using depth correction is described later in Figure 19.

[0184] (Generation process procedure) Next, an example of a generation process procedure executed by the information processing device 100 will be described using Figure 18. The generation process is realized, for example, by the CPU 301 shown in Figure 3, storage areas such as memory 302 and recording medium 305, and network I / F 303.

[0185] Figure 18 is a flowchart illustrating an example of the generation process. In Figure 18, the information processing device 100 acquires multiple pairs of fossil images and fossil species (step S1801). The information processing device 100 also acquires trend information representing the occurrence trends of fossil species by geological age (step S1802).

[0186] Next, the information processing device 100 randomly determines the target geological age and the total number of samples for generating pseudo-age data (step S1803). Then, the information processing device 100 refers to the acquired trend information to identify the occurrence trend of fossil species in the determined target geological age (step S1804).

[0187] Next, the information processing device 100 sets the occurrence rate of each fossil species based on the occurrence trends of the identified fossil species, such that the occurrence rate of fossil species that appear in the target geological period is higher than the occurrence rate of fossil species that do not appear in the target geological period (step S1805). Then, according to the set occurrence rates, the information processing device 100 sets the number of samples for each fossil species so that the sum of the number of samples for each fossil species equals the determined total number of samples (step S1806).

[0188] Next, the information processing device 100 calculates age scores for each geological age according to predetermined calculation rules, and sets the vector containing the calculated age scores as the ground truth data for the age classification network (step S1807). Then, the information processing device 100 selects the total number of pairs for the samples by selecting the number of pairs corresponding to the number of samples for each set fossil species from among the multiple pairs acquired (step S1808).

[0189] Next, the information processing device 100 sets the fossil image group, which is a collection of fossil images of each selected pair, as input data, and sets the fossil species group, which is a collection of fossil species of each selected pair, as the ground truth data for the fossil species classification network (step S1809). Then, the information processing device 100 generates training data by associating the set input data with the ground truth data for the set fossil species classification network and the ground truth data for the set age classification network (step S1810).

[0190] Next, the information processing device 100 determines whether or not a predetermined number of training data has been generated (step S1811). If the predetermined number of training data has not been generated (step S1811: No), the information processing device 100 returns to the process in step S1813. On the other hand, if the predetermined number of training data has been generated (step S1811: Yes), the information processing device 100 terminates the generation process. This allows the information processing device 100 to reduce the cost incurred in preparing training data.

[0191] (Estimated processing steps) Next, an example of an estimation process performed by the information processing device 100 will be described using Figure 19. The estimation process is realized, for example, by the CPU 301 shown in Figure 3, storage areas such as memory 302 and recording medium 305, and network I / F 303.

[0192] Figure 19 is a flowchart illustrating an example of the estimation process. In Figure 19, the information processing device 100 uses a classification model to calculate the age score for each geological age for each depth k by performing single-depth estimation for each depth k which has an order in the depth direction (step S1901). A larger k indicates a deeper depth. For example, when k=0, k represents the shallowest depth. For example, when k=K, k represents the deepest depth.

[0193] Next, the information processing device 100 calculates an integrated score by summing the age scores of each geological age for all geological age groups, combining the geological ages of each depth k such that the geological age x becomes older as the depth i increases at depth [0, K] (step S1902). i = 0 to K.

[0194] Then, the information processing device 100 sets the geological age of each depth k in the geological age group with the highest cumulative score as the correct geological age (step S1903). Here, for example, if there is another geological age with the same age score as the set correct geological age for any depth k, the information processing device 100 may additionally set that other geological age as the correct geological age. After that, the information processing device 100 terminates the estimation process. This allows the information processing device 100 to improve the accuracy of estimating geological ages.

[0195] Here, the information processing device 100 may execute some steps in the flowcharts of Figures 18 and 19 in a different order. Alternatively, the information processing device 100 may omit some steps in the flowcharts of Figures 18 and 19.

[0196] As described above, the information processing device 100 can realize a detection unit 402 that detects a group of fossil images containing microfossils, each of which is included in a geological sample image. The information processing device 100 can realize a classification unit 403 that includes a first generator, a second generator, and a third generator, which are trained using training data. The information processing device 100 can generate an intermediate representation based on the group of fossil images using the first generator. The information processing device 100 can generate information showing the likelihood of each type of microfossil in each fossil image of the group of fossil images using the second generator, based on the intermediate representation. The information processing device 100 can generate information showing the likelihood of each age of the strata corresponding to the group of fossil images using the third generator, based on the intermediate representation. The information processing device 100 can output information showing the likelihood of each type of microfossil in each fossil image and information showing the likelihood of each age of the strata corresponding to the group of fossil images using the classification unit 403, according to the detected group of fossil images. This allows the information processing device 100 to improve the accuracy of estimating the age of the geological layers.

[0197] According to the information processing device 100, it is possible to refer to information indicating a predetermined type of microfossil that appears in a geological layer of a specific age. According to the information processing device 100, it is possible to refer to a population of specimen images that depict microfossils in geological layers of different ages. According to the information processing device 100, it is possible to obtain a predetermined number of specimen images from the population that include specimen images depicting a predetermined type of microfossil. According to the information processing device 100, it is possible to generate training data that associates the specimen image group with information indicating the correct type of microfossil in each specimen image and information indicating a specific age which is the correct age of the geological layer corresponding to the specimen image group. As a result, the information processing device 100 can prepare training data and reduce the cost incurred in preparing training data.

[0198] According to the information processing device 100, a predetermined number of sample images can be randomly selected from the population so that sample images of a predetermined type of microfossil are selected with a higher probability than sample images of other types of microfossils. According to the information processing device 100, training data can be generated that associates the sample image group with information indicating the correct type of microfossil in each sample image and information indicating a specific age which is the correct age of the geological layer corresponding to the sample image group. As a result, the information processing device 100 can improve the robustness of the classification unit 403 when training the classification unit 403.

[0199] According to the information processing device 100, the age of a geological layer corresponding to a group of fossil images can be estimated based on information indicating the likelihood of each age of the geological layer corresponding to the group of fossil images. This allows the information processing device 100 to accurately estimate the age of the geological layer.

[0200] According to the information processing device 100, the classification unit 403 can output information indicating the likelihood of each age of a strata for each of a plurality of strata that have an order in the depth direction. According to the information processing device 100, based on the information indicating the likelihood of each age of a strata output for each strata, the age of each strata can be estimated such that the age of the relatively shallow strata is earlier than the age of the relatively deep strata. In this way, the information processing device 100 can improve the accuracy of estimating the age of strata.

[0201] According to the information processing device 100, a population can be generated using the detection unit 402. This allows the information processing device 100 to reduce the cost of generating the population and the cost of preparing training data.

[0202] According to the information processing device 100, multiple specific time periods can be randomly selected and training data can be generated. This allows the information processing device 100 to generate training data for various time periods. The information processing device 100 can easily generate diverse training data.

[0203] According to the information processing device 100, various microscope images of samples taken from geological strata can be used as geological sample images. The information processing device 100 allows for the use of useful geological sample images, thereby improving the accuracy of estimating the age of geological strata.

[0204] Using the classification model 510 described in this embodiment, verification experiments were conducted on samples from three wells to estimate the age of the geological layers. The results showed that the accuracy of estimating the age of the geological layers improved in all of the wells.

[0205] The information processing method described in this embodiment can be implemented by executing a pre-prepared program on a computer such as a PC or workstation. The information processing program described in this embodiment is recorded on a computer-readable recording medium and executed by being read from the recording medium by the computer. The recording medium can be a hard disk, flexible disk, CD (Compact Disc)-ROM, MO (Magneto Optical Disc), DVD (Digital Versatile Disc), etc. Furthermore, the information processing program described in this embodiment may be distributed via a network such as the Internet.

[0206] With regard to the embodiments described above, the following additional information is disclosed.

[0207] (Note 1) A detection unit that detects groups of fossil images containing microfossils within geological sample images, A classification unit trained using training data that associates a group of specimen images depicting microfossils with information indicating the correct type of microfossil in each specimen image of the group of specimen images and information indicating the correct age of the geological layer corresponding to the group of specimen images, is provided to output information indicating the likelihood of each type of microfossil in each fossil image of the group of fossils and information indicating the likelihood of each age of the geological layer corresponding to the group of fossils, respectively. It has, The classification unit includes a first generator that generates an intermediate representation based on the fossil image group; a second generator that generates information indicating the likelihood of each type of microfossil in each fossil image of the fossil image group based on the intermediate representation; and a third generator that generates information indicating the likelihood of each age of the strata corresponding to the fossil image group based on the intermediate representation.

[0208] (Note 2) A training unit that generates the training data by referring to information indicating a predetermined type of microfossil that appears in strata of a specific age, and associating a predetermined number of specimen images containing specimen images of the predetermined type of microfossil from a population of specimen images of microfossils in strata of different ages, information indicating the correct type of microfossil in each specimen image of the specimen image group, and information indicating the specific age which is the correct age of the strata corresponding to the specimen image group. The information processing apparatus according to Appendix 1, further characterized by having the following features.

[0209] (Note 3) The information processing device according to Note 2, characterized in that the training unit generates training data by associating a predetermined number of sample images randomly selected from the population so as to select sample images of a predetermined type of microfossil with a higher probability than sample images of microfossils of a different type from the population, information indicating the correct type of microfossil in each sample image of the sample image group, and information indicating a specific age which is the correct age of the geological layer corresponding to the sample image group.

[0210] (Note 4) An estimation unit that estimates the age of the strata corresponding to the fossil image group based on the information output by the classification unit that shows the likelihood of each age of the strata corresponding to the fossil image group, An information processing device according to any one of the appendices 1 to 3, further characterized by having

[0211] (Note 5) The classification unit outputs information indicating the likelihood of the age of each of the multiple strata having a sequence in the depth direction, in response to the input of a group of fossil images containing microfossils included in the geological sample image corresponding to the strata. The information processing device according to Appendix 4, characterized in that the estimation unit estimates the age of each of the strata based on the information indicating the likelihood of each age of the strata output by the classification unit, such that the age of the relatively shallow strata among the plurality of strata is earlier than the age of the relatively deep strata.

[0212] (Note 6) The information processing apparatus according to Note 3, characterized in that the training unit generates the population using the detection unit.

[0213] (Note 7) The information processing apparatus according to Note 3, characterized in that the training unit randomly selects multiple specific age groups and generates the training data.

[0214] (Note 8) The information processing device according to any one of Notes 1 to 7, characterized in that the geological sample image is a microscopic image of a sample taken from a geological layer using a conventional microscope, an optical microscope, or a polarizing microscope.

[0215] (Note 9) The information processing apparatus according to any one of Notes 1 to 8, characterized in that the microfossils include at least one of foraminifera, radiolarians, calcareous nannofossils, diatoms, pollen, spores, conodonts, or ostracods.

[0216] (Note 10) Fossil image groups containing microfossils were detected within the geological sample images. A classification unit trained using training data that associates a group of specimen images depicting microfossils, information indicating the correct type of microfossil in each specimen image of the group, and information indicating the correct age of the geological layer corresponding to the group of specimen images, generates information indicating the likelihood for each type of microfossil in each fossil image of the group, and information indicating the likelihood for each age of the geological layer corresponding to the group of fossil images, according to the detected group of fossil images. The computer performs the process, The classification unit includes a first generator that generates an intermediate representation based on the fossil image group; a second generator that generates information indicating the likelihood of each type of microfossil in each fossil image of the fossil image group based on the intermediate representation; and a third generator that generates information indicating the likelihood of each age of the strata corresponding to the fossil image group based on the intermediate representation.

[0217] (Note 11) Fossil image groups containing microfossils were detected within the geological sample images. A classification unit trained using training data that associates a group of specimen images depicting microfossils, information indicating the correct type of microfossil in each specimen image of the group, and information indicating the correct age of the geological layer corresponding to the group of specimen images, generates information indicating the likelihood for each type of microfossil in each fossil image of the group, and information indicating the likelihood for each age of the geological layer corresponding to the group of fossil images, according to the detected group of fossil images. Let the computer perform the process, The classification unit is an information processing program characterized by including a first generator that generates an intermediate representation based on the fossil image group, a second generator that generates information indicating the likelihood of each type of microfossil in each fossil image of the fossil image group based on the intermediate representation, and a third generator that generates information indicating the likelihood of each age of the strata corresponding to the fossil image group based on the intermediate representation. [Explanation of symbols]

[0218] 100 Information Processing Devices 101,402 Detection unit 102,403 Classification section 111 Geological sample images 112 Fossil Images 113 Intermediate Expressions 114 types of information 115,542 Chronological information 121 1st generator 122 2nd generator 123 Third generator 130 training data 131 Sample Image Group Information 132,133 200 Information Processing Systems 201 Image Analysis Device 202 Estimation Processing Unit 203 Client Devices 210 Network 300 bus 301 CPU 302 memory 303 Network I / F 304 Recording medium interface 305 Recording media 400 Storage section 401 Acquisition Department 404 Training Department 405 Estimation section 406 Output section 500 object detectors 510 Classification Models 520 input images 521 Input Images 530 images of potential fossils 531 Candidate fossil images 541 Fossil species information 601 Feature Vectors 602 Aggregated Vectors 610 encoder 620 Fossil Species Classification Network 630 Fossil Aggregation Network Classification network of the 640s 700,1000,1100,1200,1300,1400,1500,1600,1700 Table 800 Trend Information 900 Pseudo-chronological data 910 Input data 920,930 correct data

Claims

1. A detection unit that detects groups of fossil images, each depicting a microfossil, contained within a geological sample image, A classification unit trained using training data that associates a group of specimen images depicting microfossils with information indicating the correct type of microfossil in each specimen image of the group of specimen images and information indicating the correct age of the geological layer corresponding to the group of specimen images, is provided to output information indicating the likelihood of each type of microfossil in each fossil image of the group of fossils and information indicating the likelihood of each age of the geological layer corresponding to the group of fossils, respectively. It has, The classification unit is characterized by including a first generator that generates an intermediate representation based on the fossil image group, a second generator that generates information indicating the likelihood of each type of microfossil in each fossil image of the fossil image group based on the intermediate representation, and a third generator that generates information indicating the likelihood of each age of the strata corresponding to the fossil image group based on the intermediate representation.

2. A training unit generates training data by referencing information indicating a predetermined type of microfossil that appears in strata of a specific age, and associating a predetermined number of specimen images containing specimen images of the predetermined type of microfossil from a population of specimen images of microfossils in strata of different ages, information indicating the correct type of microfossil in each specimen image in the specimen image group, and information indicating the specific age which is the correct age of the strata corresponding to the specimen image group. The information processing apparatus according to claim 1, further comprising the features.

3. The information processing device according to claim 2, characterized in that the training unit generates training data which associates a predetermined number of sample images randomly selected from the population so as to select sample images of a predetermined type of microfossil with a higher probability than sample images of microfossils of other types, information indicating the correct type of microfossil in each sample image of the sample image group, and information indicating a specific age which is the correct age of the geological layer corresponding to the sample image group.

4. Based on the information output by the classification unit indicating the likelihood of each age of the strata corresponding to the fossil image group, an estimation unit estimates the age of the strata corresponding to the fossil image group. The information processing apparatus according to any one of claims 1 to 3, further comprising

5. The classification unit outputs information indicating the likelihood of the age of each of the multiple strata having a sequence in the depth direction, in response to the input of a group of fossil images containing microfossils included in the geological sample image corresponding to the strata. The information processing device according to claim 4, wherein the estimation unit estimates the age of each of the strata based on the information indicating the likelihood of each age of the strata output by the classification unit, such that the age of the relatively shallow strata among the plurality of strata is earlier than the age of the relatively deep strata.

6. We detect groups of fossil images, each depicting a microfossil, contained within geological sample images. A classification unit trained using training data that associates a group of specimen images depicting microfossils, information indicating the correct type of microfossil in each specimen image of the group, and information indicating the correct age of the geological layer corresponding to the group of specimen images, generates information indicating the likelihood for each type of microfossil in each fossil image of the group, and information indicating the likelihood for each age of the geological layer corresponding to the group of fossil images, according to the detected group of fossil images. The computer performs the process, The classification unit includes a first generator that generates an intermediate representation based on the fossil image group; a second generator that generates information indicating the likelihood of each type of microfossil in each fossil image of the fossil image group based on the intermediate representation; and a third generator that generates information indicating the likelihood of each age of the strata corresponding to the fossil image group based on the intermediate representation.

7. We detect groups of fossil images, each depicting a microfossil, contained within geological sample images. A classification unit trained using training data that associates a group of specimen images depicting microfossils, information indicating the correct type of microfossil in each specimen image of the group, and information indicating the correct age of the geological layer corresponding to the group of specimen images, generates information indicating the likelihood for each type of microfossil in each fossil image of the group, and information indicating the likelihood for each age of the geological layer corresponding to the group of fossil images, according to the detected group of fossil images. Let the computer perform the process, The classification unit is an information processing program characterized by including a first generator that generates an intermediate representation based on the fossil image group, a second generator that generates information indicating the likelihood of each type of microfossil in each fossil image of the fossil image group based on the intermediate representation, and a third generator that generates information indicating the likelihood of each age of the strata corresponding to the fossil image group based on the intermediate representation.

Images