Predictor creation device, prediction device, and program

The predictor creation device enhances deterioration prediction accuracy for infrastructure equipment by integrating attribute information into the prediction process, addressing the limitations of existing methods and optimizing maintenance planning to reduce Life Cycle Costs.

WO2024241501A1PCT designated stage expired Publication Date: 2024-11-28NT T INC
View PDF 2 Cites 0 Cited by

Patent Information

Application Number
PCT/JP2023/019174
Authority / Receiving Office
WO · WO
Patent Type
Applications
Current Assignee / Owner
Filing Date
2023-05-23
Publication Date
2024-11-28

AI Technical Summary

Technical Problem

Current deterioration prediction methods for infrastructure equipment lack accuracy due to the inability to consider location, materials, and time elapsed since construction, leading to potential increases in Life Cycle Cost through unanticipated maintenance changes.

Method used

A predictor creation device that uses captured images of infrastructure equipment, incorporating attribute information such as environmental, material, and elapsed time data to create a predictor that accurately forecasts deterioration, enhancing prediction accuracy by accounting for these factors.

Benefits of technology

The solution improves the accuracy of deterioration predictions, allowing for more informed maintenance planning and reduced Life Cycle Costs by considering the specific conditions and attributes of the infrastructure equipment.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure JP2023019174_28112024_PF_FP_ABST
    Figure JP2023019174_28112024_PF_FP_ABST
Patent Text Reader

Abstract

Provided is a predictor creation device (200) according to the present disclosure comprising: a data creation unit (120) that creates, as learning data, a pair of input data including a captured image of a structure to be learned at one point in time and attribute information pertaining to the attribute of the structure to be learned, and correct answer data, which is a captured image of the structure to be learned at a point in time after the one point in time; and a learning unit (130) that, through learning using the created learning data, uses a captured image obtained by capturing the structure as an input, and creates a predictor that predicts the deterioration of the structure.
Need to check novelty before this filing date? Find Prior Art

Description

Predictor creation device, prediction device and program

[0001] The present disclosure relates to a predictor creating device, a prediction device, and a program.

[0002] In recent infrastructure maintenance and management efforts, efforts are underway to optimize the life cycle cost (LCC) of infrastructure facilities. LCC refers to the costs incurred over the life cycle of a building, and includes construction costs, operation and maintenance costs such as inspection and maintenance costs, repair and renewal costs, and demolition and disposal costs. Efforts to optimize LCC include reducing inspection costs by minimizing the number of facility inspections that were previously required, and reducing excessive repair costs by extending repair intervals.

[0003] In carrying out the above initiatives, deterioration prediction methods are used to predict future equipment conditions (deterioration) based on current equipment conditions as a tool for proposing optimal inspection and repair times or for building consensus within facility management companies. Deterioration prediction methods are effective for making decisions based on future times, but if sufficient prediction accuracy is not ensured, unexpected changes to maintenance plans or equipment repairs may occur, which could lead to an increase in LCC. Therefore, there is a need to improve the accuracy of deterioration prediction methods.

[0004] Non-Patent Document 1 describes a technology that uses image recognition to predict the future state of a prediction object from images of the prediction object captured in a time series. As shown in FIG. 11 , this technology includes a learning phase using captured images of the prediction object captured in a time series from a past time t-n to a time t (present), and a prediction phase using captured images at time t. In the learning phase, as shown in FIG. 11 , a predictor that predicts the future state of the prediction object from an image of the prediction object is created by learning using pairs of input data, which is an image at a certain time, and ground truth data, which is an image at a time following the certain time, as training data. In the prediction phase, as shown in FIG. 11 , an image of the prediction object at time t is input into the predictor, and the image at time t+1 output in response to the input is again input into the predictor, and this process is repeated to predict the future state of the prediction object.

[0005] Xingjian Shi, Zhourong Chen, Hao Wang, DitYan Yeung, Wai-kin Wong, Wang-chun Woo: Convolutional LSTM Network: A MachineLearning Approach for Precipitation Nowcasting, arXiv:1506.04214v2, 2015

[0006] For infrastructure facilities, factors such as the facility's location, materials, and the time elapsed since construction are important factors for improving the accuracy of predictions of the progression of deterioration. In the technology described in Non-Patent Document 1, a predictor is trained and predictions are made using only photographed images of the object to be predicted. Therefore, it has been difficult to improve the accuracy of deterioration predictions that take into account the factors described above.

[0007] In consideration of the above-mentioned problems, an object of the present disclosure is to provide a predictor creation device, a prediction device, and a program that can improve the accuracy of deterioration prediction of a structure to be predicted.

[0008] In order to solve the above problem, the predictor creation device according to the present disclosure is a predictor creation device that takes as input an image of a structure and creates a predictor that predicts deterioration of the structure, and includes: a data creation unit that creates, as learning data, pairs of input data that includes an image of the structure to be learned at a certain point in time and attribute information related to the attributes of the structure to be learned, and ground truth data that is an image of the structure to be learned at a point in time later than the certain point in time; and a learning unit that creates the predictor by learning using the created learning data.

[0009] In order to solve the above problem, the predictor creation device according to the present disclosure is a predictor creation device that takes as input an image of a structure and creates a predictor that predicts deterioration of the structure, and includes: a data creation unit that creates pairs of input data including an image of the structure to be learned at a certain point in time and an image of the structure to be learned at a point in time earlier than the certain point in time, and ground truth data that is an image of the structure to be learned at a point in time later than the certain point in time, as learning data; and a learning unit that creates the predictor by learning using the created learning data.

[0010] In order to solve the above problem, the prediction device according to the present disclosure is a prediction device that predicts deterioration of a structure to be predicted from an image of the structure to be predicted, and includes: a data creation unit that creates input data including an image of the structure to be predicted at a certain point in time and attribute information related to attributes of the structure to be predicted; and a prediction unit that inputs the created input data into a predictor created by the above-mentioned predictor creation device, and predicts deterioration of the structure to be predicted at a point in time after the certain point in time.

[0011] In order to solve the above problem, a program according to the present disclosure causes a computer to operate as the above-described predictor creation device or prediction device.

[0012] According to the predictor creation device, prediction device, and program of the present disclosure, it is possible to improve the accuracy of deterioration prediction of the structure to be predicted.

[0013] 7 is a diagram illustrating an example configuration of a prediction system according to a first embodiment of the present disclosure. FIG. 1 is a diagram for explaining learning data created by a data creation unit shown in FIG. 1. FIG. 2 is a diagram for explaining creation of a predictor in the prediction system shown in FIG. 1 and prediction of deterioration using the predictor. FIG. 3 is a flowchart illustrating an example of operation in a learning phase of the prediction system shown in FIG. 1. FIG. 4 is a flowchart illustrating an example of operation in a prediction phase of the prediction system shown in FIG. 1. FIG. 5 is a diagram illustrating an example of distribution of captured images of a structure to be learned according to the number of years elapsed since the construction of the structure. FIG. 6 is a diagram illustrating an example configuration of a prediction system according to a second embodiment of the present disclosure. FIG. 7 is a diagram for explaining learning data created by a data creation unit shown in FIG. 7. FIG. 8 is a diagram for explaining creation of a predictor in the prediction system shown in FIG. 1 and prediction of deterioration using the predictor. FIG. 9 is a diagram illustrating an example configuration of a computer that functions as a prediction system according to the present disclosure. FIG. 10 is a diagram for explaining creation of a conventional predictor and prediction of deterioration using the predictor.

[0014] Hereinafter, embodiments of the present disclosure will be described with reference to the drawings.

[0015] 1 is a diagram illustrating an example configuration of a prediction system 100 according to a first embodiment of the present disclosure. The prediction system 100 according to this embodiment receives photographed images of a structure such as infrastructure equipment as input, creates a predictor that predicts deterioration of the structure, and uses the created predictor to predict deterioration of the structure based on photographed images of the structure whose deterioration is to be predicted.

[0016] As shown in FIG. 1, the prediction system 100 according to this embodiment includes an image input unit 110 , a data creation unit 120 , a learning unit 130 , a prediction unit 140 , and an image output unit 150 .

[0017] The image input unit 110 receives input of photographed images of a structure. In the learning phase in which a predictor is created, the image input unit 110 receives input of photographed images of the structure to be learned in chronological order. In the prediction phase in which the predictor is used to predict deterioration of the structure to be predicted, the image input unit 110 receives input of photographed images of the structure to be predicted. The structure to be learned and the structure to be predicted may be the same or different. The image input unit 110 outputs the input photographed images of the structure to the data creation unit 120.

[0018] In the learning phase, the data creation unit 120 creates, as learning data, pairs (data sets) of input data including an image of a structure to be learned and attribute information of the structure to be learned at a certain time point, and supervised answer data, which is an image of the structure to be learned at a time point later than the certain time point. As described above, in the learning phase, images of the structure to be learned, which are captured in chronological order, are input. For example, for each time point at which the structure is captured, the data creation unit 120 creates, as learning data, pairs of input data including the structure to be learned and attribute information of the structure to be learned at that time point, and supervised answer data, which is an image of the structure to be learned at a time point later than the certain time point (e.g., the image captured one time later in the chronological order). The data creation unit 120 outputs the created learning data to the learning unit 130.

[0019] In the prediction phase, the data creation unit 120 creates input data including a photographed image of the structure to be predicted at a certain point in time and attribute information of the structure to be predicted. The data creation unit 120 outputs the created input data to the prediction unit 140.

[0020] Here, the attribute information of a structure is information related to the attributes of the structure. The attribute information includes at least one of environmental information related to the environment in which the structure is installed, material information related to the materials used to make the structure, and historical information related to the time elapsed since the construction of the structure. The data creation unit 120 receives this information via, for example, an input unit (not shown in FIG. 1 ).

[0021] For example, if a structure is made of steel, the environmental information includes information on the amount of precipitation, the pH of raindrops, the amount of snow, the temperature, humidity, and location (e.g., whether it is along the coast) at the location where the structure is installed. Also, the material information includes information on the type of steel, the presence or absence of corrosion protection, and the type of corrosion protection.

[0022] For example, if a structure is made of concrete, the environmental information includes information on the amount of precipitation, the pH of raindrops, the location (e.g., whether it is on the coast or not), the load applied to the structure, etc. The material information includes information on the type of aggregate, the mixing ratio of the material, the moisture content, etc.

[0023] The data creation unit 120 may include in the environmental information the history of disasters (earthquakes, wind and flood damage, etc.) that have occurred in the location where the structure is installed.

[0024] The elapsed time information is, for example, information indicating the number of years that have passed since the structure was installed.

[0025] Including environmental information in the attribute information makes it possible to train a predictor that takes into account the degree of deterioration due to the environment in which the structure is installed, and to make predictions using that predictor.In addition, including material information in the attribute information makes it possible to train a predictor that takes into account the degree of deterioration due to the material of the structure, and to make predictions using that predictor.In addition, including historical information in the attribute information makes it possible to train a predictor that takes into account the degree of deterioration due to the time elapsed since the structure was installed, and to make predictions using that predictor.

[0026] As shown in FIG. 2 , the data creation unit 120 may assign attribute information as a tensor expressed as a multidimensional array. For example, the data creation unit 120 may assign attribute information to each pixel of the captured image. In this case, the data creation unit 120 may assign material information corresponding to the material of the structure at each pixel as attribute information to each pixel of the captured image. Note that, for example, if the attribute information is information that does not change for each pixel of the captured image, such as environmental information or historical information, the data creation unit 120 may assign the same information to all pixels. By assigning attribute information to each pixel of the captured image, more accurate prediction is possible when the degree of deterioration varies depending on the pixel region.

[0027] 1 again, in the learning phase, the learning unit 130 creates a predictor that predicts deterioration of a structure by inputting photographed images of the structure through learning using the learning data created by the data creation unit 120. There are various methods for such learning, and detailed explanations will be omitted.

[0028] In the prediction phase, the prediction unit 140 inputs input data, created by the data creation unit 120, including a photographed image of the structure to be predicted at a certain point in time and attribute information of the structure to be predicted, to the predictor created by the learning unit 130, and predicts deterioration of the structure to be predicted at a point in time after the certain point in time. The prediction unit 140 outputs the predicted result of deterioration of the structure to the image output unit 150.

[0029] The image output unit 150 outputs a predicted image showing the prediction result of the prediction unit 140 by displaying it on a display device included in the prediction system 100 or a display device connected to the prediction system 100 .

[0030] In the prediction system 100 according to this embodiment, as shown in Fig. 3 , in the learning phase, a predictor is created using, as training data, pairs of input data including photographed images of a structure to be learned and attribute information of the structure to be learned at a certain point in time, and ground truth data, which is photographed images of the structure to be learned at a point in time after the certain point in time. Also, in the prediction system 100 according to this embodiment, as shown in Fig. 3 , in the prediction phase, input data including photographed images of a structure to be learned at a certain point in time and attribute information of the structure to be learned are input to the predictor to predict deterioration of the structure to be learned at a point in time after the certain point in time. This makes it possible to train the predictor and make predictions using the predictor, taking into account the environment in which the structure is installed, the materials constituting the structure, the time elapsed since the structure was installed, and so on, thereby achieving higher accuracy in predicting deterioration of the structure compared to when only photographed images of the structure are used.

[0031] 1, the prediction system 100 may be configured as a predictor creation device 200 that takes an image of a structure as an input and creates a predictor that predicts deterioration of the structure, and a prediction device 300 that predicts deterioration of a prediction target structure from an image of the structure. In other words, the configuration for creating the predictor and the configuration for predicting deterioration using the predictor may be provided separately.

[0032] In this case, predictor creation device 200 includes image input unit 110, data creation unit 120 as a first data creation unit, and learning unit 130. Data creation unit 120 constituting predictor creation device 200 creates, as learning data, pairs of input data including an image of a structure to be learned at a certain point in time and attribute information related to the attributes of the structure to be learned, and ground truth data which is an image of the structure to be learned at a point in time later than the certain point in time.

[0033] Prediction device 300 also includes image input unit 110, data creation unit 120 as a second data creation unit, prediction unit 140, and image output unit 150. Data creation unit 120 constituting prediction device 300 creates input data including a photographed image of a structure to be predicted at a certain point in time and attribute information relating to the attributes of the structure to be predicted.

[0034] 1 shows an example in which the image input unit 110 and the data creation unit 120 are shared by the predictor creation device 200 and the prediction device 300, but this is not limiting. The predictor creation device 200 and the prediction device 300 may each be provided with the image input unit 110 and the data creation unit 120.

[0035] Next, the operation of the prediction system 100 according to this embodiment will be described separately for the operation in the learning phase (i.e., the operation of the predictor creating device 200) and the operation in the prediction phase (i.e., the operation of the prediction device 300).

[0036] FIG. 4 is a flowchart showing an example of the operation of the prediction system 100 in the learning phase, and is a diagram for explaining the predictor generating method by the predictor generating device 200.

[0037] The captured images of the structure to be learned are input to the image input unit 110 (step S11). As described above, the captured images of the structure to be learned are images of the structure captured in time series.

[0038] The data creation unit 120 uses the input photographed images of the structure to be learned to create learning data to be used for learning the predictor (step S12). Specifically, the data creation unit 120 creates, as the learning data, pairs of input data including a photographed image of the structure to be learned at a certain point in time and attribute information of the structure to be learned, and ground truth data which is a photographed image of the structure to be learned at a point in time later than the certain point in time.

[0039] The learning unit 130 creates a predictor by learning using the created learning data (step S13).

[0040] FIG. 5 is a flowchart showing an example of the operation of the prediction system 100 in the prediction phase, and is a diagram for explaining the prediction method performed by the prediction device 300.

[0041] The image input unit 110 receives a captured image of a structure to be predicted (step S21).

[0042] The data creation unit 120 uses the input photographed image of the structure to be learned to create input data to be input to the predictor (step S22). Specifically, the data creation unit 120 creates input data including a photographed image of the structure to be predicted at a certain point in time and attribute information of the structure to be predicted.

[0043] The prediction unit 140 inputs the input data created by the data creation unit 120 into a predictor, and predicts deterioration of the structure to be predicted at a time point later than the given time point (step S23).

[0044] The image output unit 150 outputs a predicted image indicating the prediction result of the prediction unit 140 (step S24).

[0045] As described above, the predictor creation device 200 according to this embodiment includes a data creation unit 120 and a learning unit 130. The data creation unit 120 creates, as learning data, pairs of input data including an image of a structure to be learned at a certain point in time and attribute information related to the attributes of the structure to be learned, and ground truth data which is an image of the structure to be learned at a point in time later than the certain point in time. The learning unit 130 creates a predictor by learning using the created learning data.

[0046] The prediction device 300 according to this embodiment also includes a data creation unit 120 and a prediction unit 140. The data creation unit 120 creates input data including a photographed image of a structure to be predicted at a certain point in time and attribute information related to attributes of the structure to be predicted. The prediction unit 140 inputs the created input data to a predictor created by the predictor creation device 200, and predicts deterioration of the structure to be predicted at a point in time after the certain point in time.

[0047] By using attribute information of a structure in addition to photographed images of the structure to create a predictor and predict deterioration using the predictor, it becomes possible to predict deterioration with greater accuracy, taking into account the degree of deterioration according to the attributes of the structure.

[0048] For example, when the attribute information is historical information, as shown in FIG. 6, there may be a case where the number of captured images of the structure to be learned is small within a specific range of the number of years since the structure was constructed (between 30 and 40 years in the example shown in FIG. 6). In this case, the prediction accuracy for that specific range will be reduced. Therefore, for that specific range, the data creation unit 120 may add to the input data the prediction results of a previous deterioration prediction model that inputs the number of years since the structure was constructed and outputs the degree of deterioration of the structure according to that number of years.

[0049] That is, the data creation unit 120 may add to the input data, for a specific range of variables corresponding to the attributes of the structure to be learned, the prediction results of a previous deterioration prediction model that predicts deterioration of the structure based on the variables. Here, the specific range is, for example, a range in which the prediction accuracy is lower than in other ranges due to a lack of learning data, or a range in which the prediction accuracy is equal to or lower than a predetermined threshold.

[0050] For the range where a decrease in prediction accuracy is expected, the decrease in prediction accuracy in that range can be suppressed by adding the prediction results of the existing deterioration prediction model to the input data. Note that the variables according to attributes are not limited to the number of years since the construction of the structure described above.

[0051] Second Embodiment FIG. 7 is a diagram illustrating an example of the configuration of a prediction system 100A according to a second embodiment of the present disclosure.

[0052] As shown in FIG. 7 , a prediction system 100A according to this embodiment includes an image input unit 110, a data creation unit 120A, a learning unit 130A, a prediction unit 140A, and an image output unit 150. The prediction system 100A according to this embodiment differs from the prediction system 100 according to the first embodiment in that the data creation unit 120, the learning unit 130, and the prediction unit 140 are replaced by a data creation unit 120A, a learning unit 130A, and a prediction unit 140A, respectively. The image input unit 110, the data creation unit 120A, and the learning unit 130A constitute a predictor creation device 200A. The image input unit 110, the data creation unit 120A, the prediction unit 140A, and the image output unit 150 constitute a prediction device 300A.

[0053] In the learning phase, the data creation unit 120A creates pairs of learning data, each pair consisting of input data including an image of the structure to be learned at a certain time point and an image of the structure to be learned at a time point earlier than the certain time point (one time point earlier, two time points earlier, etc.), and ground truth data, which is an image of the structure to be learned at a time point later than the certain time point. The data creation unit 120A outputs the created learning data to the learning unit 130A.

[0054] Furthermore, in the prediction phase, the data creation unit 120A creates input data including an image of the structure to be predicted at a certain time point and an image of the structure to be predicted at a time point prior to the certain time point (one time point prior, two time points prior, etc.) The data creation unit 120A outputs the created learning data to the prediction unit 140A.

[0055] The data creating unit 120A may add an image captured at a time point prior to a certain time point as a tensor expressed as a multidimensional array, as shown in Fig. 8. The data creating unit 120A may also include attribute information of the structure in the input data.

[0056] In the learning phase, the learning unit 130A creates a predictor by learning using the learning data created by the data creation unit 120A.

[0057] In the prediction phase, the prediction unit 140A inputs input data created by the data creation unit 120A, including photographed images of the structure to be predicted at a certain point in time and photographed images of the structure to be predicted at a point in time earlier than the certain point in time, into the predictor created by the learning unit 130A, and predicts deterioration of the structure to be predicted at a point in time later than the certain point in time.

[0058] In the prediction system 100A according to this embodiment, as shown in FIG. 9 , in the learning phase, a predictor is created using pairs of input data including photographed images of the structure to be learned at a certain time point and photographed images of the structure to be learned at a time point earlier than the certain time point (images of the structure to be learned before the certain time point), and ground truth data that is photographed images of the structure to be learned at a time point later than the certain time point. Also, in the prediction system 100A according to this embodiment, as shown in FIG. 3 , in the prediction phase, input data including photographed images of the structure to be predicted at a certain time point and photographed images of the structure to be learned at a time point earlier than the certain time point is input to the predictor to predict deterioration of the structure to be predicted at a time point later than the certain time point. This enables the predictor to be trained taking into account the deterioration trend of the structure, and predictions using the predictor can be made with higher accuracy than when photographed images of the structure alone are used.

[0059] The above-described prediction systems 100, 100A (predictor generating devices 200, 200A and prediction devices 300, 300A) can each be realized by a computer 10 shown in FIG. 10 . A program for causing the computer 10 to function as the prediction systems 100, 100A (predictor generating devices 200, 200A and prediction devices 300, 300A) may be provided. The program may be stored in a storage medium or provided via a network. FIG. 10 is a block diagram showing a schematic configuration of a computer 10 functioning as the prediction systems 100, 100A (predictor generating devices 200, 200A and prediction devices 300, 300A). The computer 10 may be a general-purpose computer, a dedicated computer, a workstation, a personal computer (PC), an electronic notepad, or the like. The program instructions may be program code, code segments, or the like for executing necessary tasks.

[0060] 10, a computer 10 includes a processor 11, a ROM (Read Only Memory) 12, a RAM (Random Access Memory) 13, a storage 14, an input unit 15, a display unit 16, and a communication interface (I / F) 17. Each component is connected to each other via a bus 19 so as to be able to communicate with each other. The processor 11 is specifically a CPU (Central Processing Unit), an MPU (Micro Processing Unit), a GPU (Graphics Processing Unit), a DSP (Digital Signal Processor), a SoC (System on a Chip), or the like, and may be configured by multiple processors of the same type or different types.

[0061] The processor 11 is a control unit that controls each component and performs various arithmetic operations. That is, the processor 11 reads a program from the ROM 12 or the storage 14 and executes the program using the RAM 13 as a work area. The processor 11 controls each component and performs various arithmetic operations in accordance with the program stored in the ROM 12 or the storage 14. In this embodiment, the ROM 12 or the storage 14 stores a program for operating the computer 10 as the prediction system 100, 100A (predictor creation devices 200, 200A and prediction devices 300, 300A) according to the present disclosure. The program is read and executed by the processor 11 to realize each component of the prediction system 100, 100A (predictor creation devices 200, 200A and prediction devices 300, 300A).

[0062] The program may be provided in a form stored on a non-transitory storage medium such as a CD-ROM (Compact Disk Read Only Memory), a DVD-ROM (Digital Versatile Disk Read Only Memory), a USB (Universal Serial Bus) memory, etc. The program may also be provided in a form downloaded from an external device via a network.

[0063] The ROM 12 stores various programs and various data. The RAM 13 temporarily stores programs or data as a working area. The storage 14 is configured with an HDD (Hard Disk Drive) or an SSD (Solid State Drive) and stores various programs including the operating system and various data.

[0064] The input unit 15 includes a pointing device such as a mouse and a keyboard, and is used to input various types of information.

[0065] The display unit 16 is, for example, a liquid crystal display, and displays various information. The display unit 16 may be a touch panel type and function as the input unit 15.

[0066] The communication interface 17 is an interface for communicating with other devices, for example, an interface for a LAN.

[0067] The following additional notes are provided regarding the above-described embodiments.

[0068] [Supplementary Item 1] A predictor creation device that takes an image of a structure as input and creates a predictor that predicts deterioration of the structure, comprising: a first control unit; the first control unit creates, as learning data, pairs of input data that includes an image of the structure to be learned at a certain point in time and attribute information related to attributes of the structure to be learned, and ground truth data that is an image of the structure to be learned at a point in time later than the certain point in time; and creates the predictor by learning using the created learning data.

[0069] [Supplementary Item 2] In the predictor creation device described in Supplementary Item 1, the attribute information includes at least one of environmental information regarding the environment in which the structure to be learned is installed, material information regarding the materials of the structure to be learned, and progress information regarding the time elapsed since the construction of the structure to be learned.

[0070] [Supplementary Item 3] The predictor creation device according to Supplementary Item 2, wherein the environmental information includes information relating to a history of disasters occurring in a location where the structure to be learned is installed.

[0071] [Supplementary Item 4] The predictor creating device according to any one of Supplementary Items 1 to 3, wherein the first control unit assigns the attribute information to each pixel of the captured image.

[0072] [Supplementary Item 5] In the predictor creation device described in any one of Supplementary Items 1 to 4, the first control unit adds, to the input data, a prediction result of a deterioration prediction model that predicts deterioration of the structure based on the variables for a specific range within a range of variables according to the attributes of the structure to be learned.

[0073] [Supplementary Item 6] A predictor creation device that takes an image of a structure as input and creates a predictor that predicts deterioration of the structure, comprising: a second control unit; the second control unit creates, as learning data, pairs of input data including an image of the structure to be learned at a certain point in time and an image of the structure to be learned at a point in time earlier than the certain point in time, and ground truth data that is an image of the structure to be learned at a point in time later than the certain point in time; and creates the predictor by learning using the created learning data.

[0074] [Supplementary Item 7] A prediction device that predicts deterioration of a structure to be predicted from an image of the structure to be predicted, the prediction device comprising: a third control unit that creates input data including an image of the structure to be predicted at a certain point in time and attribute information related to attributes of the structure to be predicted; and inputs the created input data into a predictor created by the predictor creation device described in any one of Supplementary Items 1 to 5, thereby predicting deterioration of the structure to be predicted at a point in time after the certain point in time.

[0075] [Supplementary Item 8] A prediction device that predicts deterioration of a structure to be predicted from an image of the structure to be predicted, the prediction device comprising: a fourth control unit, the fourth control unit creating input data including an image of the structure to be predicted at a certain point in time and an image of the structure to be predicted at a point in time earlier than the certain point in time, and inputting the created input data into a predictor created by the predictor creation device described in Supplementary Item 6, to predict deterioration of the structure to be predicted at a point in time later than the certain point in time.

[0076] [Supplementary Item 9] A prediction system comprising: a predictor creation device that takes an image of a structure as an input and creates a predictor that predicts deterioration of the structure; and a prediction device that predicts deterioration of a structure to be predicted from an image of the structure to be predicted, wherein the predictor creation device creates, as learning data, pairs of input data including an image of the structure to be learned at a certain point in time and attribute information related to the attributes of the structure to be learned, and ground truth data that is an image of the structure to be learned at a point in time later than the certain point in time; creates the predictor by learning using the created learning data; and the prediction device creates input data including an image of the structure to be predicted at a certain point in time and attribute information related to the attributes of the structure to be predicted; and inputs the created input data into a predictor created by the predictor creation device described in any one of Supplementary Items 1 to 5, to predict deterioration of the structure to be predicted at a point in time later than the certain point in time.

[0077] [Supplementary Item 10] A prediction system comprising: a predictor creation device that receives an image of a structure as an input and creates a predictor that predicts deterioration of the structure; and a prediction device that predicts deterioration of a structure to be predicted from an image of the structure to be predicted, wherein the predictor creation device creates, as learning data, pairs of input data including an image of the structure to be learned at a certain point in time and an image of the structure to be learned at a point in time earlier than the certain point in time, and ground truth data that is an image of the structure to be learned at a point in time later than the certain point in time; and creates the predictor by learning using the created learning data; and the prediction device creates input data including an image of the structure to be predicted at a certain point in time and an image of the structure to be predicted at a point in time earlier than the certain point in time; and inputs the created input data into a predictor created by the predictor creation device described in Supplementary Item 6, to predict deterioration of the structure to be predicted at a point in time later than the certain point in time.

[0078] [Supplementary Item 11] A predictor creation method by a predictor creation device that takes an image of a structure as input and creates a predictor that predicts deterioration of the structure, the predictor creation method comprising: creating, as learning data, pairs of input data including an image of the structure to be learned at a certain point in time and attribute information of the structure to be learned, and ground truth data that is an image of the structure to be learned at a point in time after the certain point in time; and creating the predictor by learning using the created learning data.

[0079] [Supplementary Item 12] A predictor creation method by a predictor creation device that takes an image of a structure as input and creates a predictor that predicts deterioration of the structure, the predictor creation method comprising: creating, as learning data, pairs of input data including an image of the structure to be learned at a certain point in time and an image of the structure to be learned at a point in time earlier than the certain point in time, and ground truth data that is an image of the structure to be learned at a point in time later than the certain point in time; and creating the predictor by learning using the created learning data.

[0080] [Supplementary Item 13] A prediction method using a prediction device that predicts deterioration of a structure to be predicted from an image of the structure to be predicted, the prediction method comprising: creating input data including an image of the structure to be predicted at a certain point in time and attribute information of the structure to be predicted; and inputting the created input data into a predictor created by the predictor creation method described in Supplementary Item 11, to predict deterioration of the structure to be predicted at a point in time after the certain point in time.

[0081] [Supplementary Item 14] A prediction method using a prediction device that predicts deterioration of a structure to be predicted from an image of the structure to be predicted, the prediction method comprising: creating input data including an image of the structure to be predicted at a certain point in time and an image of the structure to be predicted at a point in time earlier than the certain point in time; and inputting the created input data into a predictor created by the predictor creation method described in Supplementary Item 12, to predict deterioration of the structure to be predicted at a point in time later than the certain point in time.

[0082] [Supplementary Item 15] A non-transitory storage medium storing a program executable by a computer, the program causing the computer to operate as the predictor creation device according to any one of Supplementary Items 1 to 6 or the prediction device according to any one of Supplementary Items 7 to 8.

[0083] Although the above-described embodiments have been described as typical examples, it will be apparent to those skilled in the art that many modifications and substitutions can be made within the spirit and scope of the present disclosure. Therefore, the present invention should not be interpreted as being limited by the above-described embodiments, and various modifications and alterations are possible without departing from the scope of the claims. For example, multiple building blocks shown in the block diagrams of the embodiments can be combined into one, or one building block can be divided.

[0084] REFERENCE SIGNS LIST 10 Computer 11 Processor 12 ROM 13 RAM 14 Storage 15 Input unit 16 Display unit 17 Communication I / F 19 Path 100, 100A Prediction system 110 Image input unit 120, 120A Data creation unit 130, 130A Learning unit 140, 140A Prediction unit 150 Image output unit

Claims

1. A predictor creation device that takes an image of a structure as input and creates a predictor that predicts deterioration of the structure, comprising: a data creation unit that creates pairs of learning data consisting of input data including an image of the structure to be trained at a certain point in time and attribute information related to the attributes of the structure to be trained, and ground truth data that is an image of the structure to be trained at a point in time after the certain point in time; and a learning unit that creates the predictor by learning using the created learning data.

2. A predictor creation device as described in claim 1, wherein the attribute information includes at least one of environmental information regarding the environment in which the structure to be studied is installed, material information regarding the materials of the structure to be studied, and historical information regarding the time elapsed since the construction of the structure to be studied.

3. A predictor creation device as described in claim 2, wherein the environmental information includes information regarding the history of disasters occurring in the location where the structure to be learned is installed.

4. A predictor generating device according to claim 1, wherein the data generating section assigns the attribute information to each pixel of the captured image.

5. A predictor creation device as described in claim 1, wherein the data creation unit adds to the input data the prediction results of a deterioration prediction model that predicts deterioration of the structure based on the variables for a specific range within the range of variables corresponding to the attributes of the structure to be learned.

6. A predictor creation device that takes as input an image of a structure and creates a predictor that predicts deterioration of the structure, comprising: a data creation unit that creates pairs of input data including an image of the structure to be trained at a certain point in time and an image of the structure to be trained at a time earlier than the certain point in time, and ground truth data that is an image of the structure to be trained at a time later than the certain point in time, as training data; and a learning unit that creates the predictor by learning using the created training data.

7. A prediction device that predicts deterioration of a structure to be predicted from an image of the structure to be predicted, comprising: a data creation unit that creates input data including an image of the structure to be predicted at a certain point in time and attribute information related to attributes of the structure to be predicted; and a prediction unit that inputs the created input data to a predictor created by the predictor creation device described in claim 1, and predicts deterioration of the structure to be predicted at a point in time after the certain point in time.

8. A program for causing a computer to operate as the predictor generating device according to claim 1 or 6 or the prediction device according to claim 7.

Citation Information

Patent Citations

  • Imaging apparatus and imaging method

    JP2020091702A

  • Image processing apparatus, image processing method, and program

    JP2023020343A