Object evaluation device, object evaluation method, and program

Abstract

A model acquisition unit (140) acquires an evaluation model. The evaluation model generates, as training data, some among a plurality of pieces of partial data that have been generated as a result of dividing, into a plurality of pieces, three-dimensional data indicating the shape of an object. An evaluation data generation unit (150) generates evaluation data for evaluating the presence or absence of object abnormalities by using, as input data for the evaluation model, at least some of the remainder of the plurality of pieces of partial data. The three-dimensional shape indicated by each of at least two pieces of partial data is the same in a range including a prescribed error. In addition, at least one among these pieces of partial data having the same three-dimensional shape is included in first partial data for generating training data, and at least one other piece of the partial data is included in the input data.

Description

[Technical Field]

[0001] The present invention relates to an object evaluation device, an object evaluation method, and a recording medium. [Background technology]

[0002] Detecting anomalies in objects using 3D data that shows the shape of the object has been studied. In particular, in recent years, detecting these anomalies using models trained by machine learning has been studied.

[0003] For example, Patent Document 1 describes the following method for generating a model. First, an image of a symmetrical portion is extracted from a learning image obtained by capturing an image of an object of the same type as the object to be inspected, and the image of the symmetrical portion is divided into two in the symmetrical direction to generate a divided image. Next, each divided image is inverted to generate an inverted image. Then, an image of one side of the divided image and an inverted image obtained by inverting the image of the other side of the divided image are acquired as training data for one side of the symmetrical portion of the object. Furthermore, an image of the other side of the divided image and an inverted image obtained by inverting the image of one side of the divided image are acquired as training data for the other side of the symmetrical portion of the object. Then, machine learning is performed based on the acquired training data to generate a model to be used for visual inspection of the object. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] Japanese Patent Publication No. 2020-102111 Summary of the Invention [Problem to be solved by the invention]

[0005] In the above-mentioned Patent Document 1, the symmetry of the object is used to increase the amount of training data. However, depending on the type of object, the number of objects having an abnormality may be small. In this case, the method described in Patent Document 1 may result in an insufficient amount of training data having an abnormality, which may result in insufficient improvement in the accuracy of the model.

[0006] One example of the object of the present invention is to provide an object evaluation device, an object evaluation method, and a recording medium that can detect abnormalities with high accuracy even if the object rarely has abnormalities, in consideration of the above-mentioned problems. [Means for solving the problem]

[0007] According to one aspect of the present invention, a model acquisition means acquires an evaluation model generated by dividing three-dimensional data indicating the shape of an object into a plurality of parts and using some of the plurality of partial data as training data; an evaluation data generating means for generating evaluation data for evaluating the presence or absence of an abnormality in the object by using at least a portion of the remaining plurality of partial data as input data for the evaluation model; Equipped with the three-dimensional shapes indicated by the at least two partial data are identical within a range including a predetermined error; An object evaluation device is provided in which at least one of the at least two partial data whose three-dimensional shapes are identical to each other is included in the first partial data for generating the learning data, and at least one other of the at least two partial data is included in the second partial data that becomes the input data.

[0008] According to one aspect of the present invention, a computer Obtain an evaluation model generated by dividing the three-dimensional data representing the shape of the object into multiple parts and using some of the multiple partial data as training data; generating evaluation data for evaluating the presence or absence of an abnormality in the object by using at least a portion of the remaining plurality of partial data as input data for the evaluation model; the three-dimensional shapes indicated by the at least two partial data are identical within a range including a predetermined error; An object evaluation method is provided in which at least one of the at least two partial data whose three-dimensional shapes are identical to each other is included in the first partial data for generating the learning data, and at least one other of the at least two partial data is included in the second partial data that becomes the input data.

[0009] According to one aspect of the present invention, a computer includes: A model acquisition function that acquires an evaluation model generated by dividing the 3D data representing the shape of the object into multiple parts and using some of the multiple partial data as training data; and an evaluation data generation function that generates evaluation data for evaluating the presence or absence of an abnormality in the object by using at least a portion of the remaining plurality of partial data as input data for the evaluation model; It records the program that allows you to the three-dimensional shapes indicated by the at least two partial data are identical within a range including a predetermined error; A computer-readable recording medium is provided in which at least one of the at least two partial data whose three-dimensional shapes are identical to each other is included in first partial data for generating the learning data, and at least one other of the at least two partial data is included in second partial data that becomes the input data. [Effects of the Invention]

[0010] According to one aspect of the present invention, it is possible to provide an object evaluation device, an object evaluation method, and a recording medium that can detect abnormalities with high accuracy even if the object rarely has abnormalities. [Brief explanation of the drawings]

[0011] [Figure 1] 1 is a diagram showing an overview of an object evaluation device according to an embodiment; [Figure 2]FIG. 2 is a diagram illustrating an example of a functional configuration of an object evaluation device. [Figure 3] FIG. 3 is a diagram showing a first example of information stored in a storage unit. [Figure 4] FIG. 10 is a diagram illustrating a second example of information stored in the storage unit. [Figure 5] FIG. 10 is a diagram illustrating a first example of processing performed by a dividing unit. [Figure 6] FIG. 10 is a diagram illustrating a second example of processing performed by the dividing unit. [Figure 7] 10A and 10B are diagrams illustrating a first example of a process performed by an allocation unit. [Figure 8] FIG. 10 is a diagram illustrating a second example of a process performed by the allocation unit. [Figure 9] FIG. 2 is a diagram illustrating an example of a hardware configuration of an object evaluation device. [Figure 10] 4 is a flowchart showing a first example of processing performed by the object evaluation device. [Figure 11] FIG. 10 is a diagram for explaining a second example of processing performed by the object evaluation device. [Figure 12] 10 is a flowchart illustrating an example of processing performed by the object evaluation device. DETAILED DESCRIPTION OF THE INVENTION

[0012] Hereinafter, embodiments of the present invention will be described with reference to the drawings. In all the drawings, like components are designated by like reference numerals, and the description thereof will be omitted as appropriate.

[0013] FIG. 1 is a diagram showing an overview of an object evaluation device 10 according to an embodiment. The object evaluation device 10 includes a model acquisition unit 140 and an evaluation data generation unit 150. The model acquisition unit 140 acquires an evaluation model. The evaluation model is generated by dividing three-dimensional data indicating the shape of the object into multiple pieces, and using some of the multiple partial data as training data. The evaluation data generation unit 150 generates evaluation data for evaluating the presence or absence of an abnormality in the object by using at least some of the remaining multiple partial data as input data for the evaluation model. The three-dimensional shapes indicated by at least two partial data are identical within a range that includes a predetermined error. At least one of these partial data having the same three-dimensional shape is included in first partial data for generating training data, and at least one of the other partial data is included in input data. The second partial data becomes Included in.

[0014] The object evaluation device 10 does not require the preparation of training data in advance. Furthermore, at least one three-dimensional shape of the first partial data for generating the evaluation model and at least one three-dimensional shape of the second partial data for generating the evaluation data are identical within a predetermined error range. Therefore, even if an object rarely has an abnormality, the abnormality can be detected with high accuracy.

[0015] A detailed example of the object evaluation device 10 will be described below.

[0016] 2 is a diagram showing an example of the functional configuration of the object evaluation device 10. In addition to the model acquisition unit 140 and evaluation data generation unit 150 described above, the object evaluation device 10 has a three-dimensional data acquisition unit 110, a division unit 120, and an allocation unit 130, and is used together with a storage unit 20. Note that the storage unit 20 may be a part of the object evaluation device 10.

[0017] The object evaluation device 10 is, for example, a server, but may also be a terminal.

[0018] The storage unit 20 stores three-dimensional data indicating the shape of an object. The three-dimensional data is, for example, a measurement result by LiDAR, i.e., point cloud data, but may also be generated using images of the object.

[0019] The objects are structures such as bridges, overpasses, buildings, and tunnels, but may also be products produced in a factory. When there are multiple types of objects, the storage unit 20 stores 3D data for each of the multiple types of objects. For example, the storage unit 20 stores 3D data for each of multiple bridges (or each of tunnels, buildings, or overpasses). Furthermore, when measurements are taken multiple times for the same object, 3D data is generated each time the object is measured. The storage unit 20 then stores these multiple 3D data for each object.

[0020] The three-dimensional data acquisition unit 110 acquires three-dimensional data from the storage unit 20. When the storage unit 20 stores multiple pieces of three-dimensional data, the three-dimensional data acquisition unit 110 acquires three-dimensional data of the object that is being checked for the presence or absence of an abnormality in this process.

[0021] The dividing unit 120 divides the 3D data acquired by the 3D data acquisition unit 110 into multiple partial data. As described with reference to FIG. 1, the 3D shapes represented by at least two partial data are identical within a predetermined error range. The dividing unit 120 preferably divides the 3D data so that each partial data has at least one other partial data with the same shape. Regarding the predetermined error, if the divided data contains a predetermined shape such as a plane or sphere, it is sufficient that the other data also contains the predetermined shape, for example, a plane or sphere. The number of points, area, and volume may differ slightly between the divided data. Specific examples of the processing performed by the dividing unit 120 will be described later using other figures.

[0022] The dividing unit 120 may use part of the three-dimensional data acquired from the three-dimensional data acquiring unit 110 as the data that will form the partial data. For example, if the three-dimensional data is a bridge, the three-dimensional data acquiring unit 110 may use the parts of the bridge that correspond to the piers as the data that will form the partial data.

[0023] The allocation unit 130 allocates the plurality of partial data generated by the division unit 120 into first partial data for generating an evaluation model and second partial data for generating evaluation data. As described above, for each piece of partial data, there exists other partial data having the same shape. The allocation unit 130 then assigns some of these plurality of partial data to the first partial data and the remaining partial data to the second partial data. As a result, at least one of at least two partial data having the same three-dimensional shape is included in the first partial data, and the remaining of the at least two partial data is included in the second partial data.

[0024] The model acquisition unit 140 generates an evaluation model using the first partial data. The type of machine learning used for the evaluation model is, for example, an autoencoder or a generative adversarial network.

[0025] The model acquisition unit 140 may store the generated evaluation model in the storage unit 20. When the storage unit 20 stores three-dimensional data for each of a plurality of objects, the model acquisition unit 140 stores an evaluation model for each of the plurality of objects.

[0026] The evaluation data generation unit 150 generates evaluation data by inputting the second partial data into the evaluation model generated by the model acquisition unit 140. If an autoencoder is used in the evaluation model, the evaluation data will be data representing a three-dimensional shape. The evaluation data generation unit 150 then uses the difference between the shape represented by the second partial data and the shape represented by the evaluation data to determine whether or not there is an abnormality in the portion of the object corresponding to the second partial data. For example, if this difference is equal to or greater than a reference value, the evaluation data generation unit 150 determines that there is an abnormality in the portion of the object corresponding to the second partial data.

[0027] For example, consider the case where the three-dimensional data is point cloud data. The second partial data S is a set of multiple point clouds, and the evaluation data S' is also a set of multiple point clouds. If S is expressed as the following formula (1) and S^ is expressed as the following formula (2), the evaluation data generation unit 150 calculates the difference using the following formula (3), and if this difference is equal to or greater than a reference value, it determines that there is an abnormality in the part corresponding to this second partial data.

[0028]

number

[0029]

number

[0030]

number

[0031] If the evaluation model is stored in the storage unit 20, the evaluation data generation unit 150 reads out the evaluation model from the storage unit 20. If the storage unit 20 stores evaluation models for multiple objects, the evaluation data generation unit 150 acquires from the storage unit 20 the evaluation model corresponding to the object that is the current processing target.

[0032] 3 is a diagram showing a first example of information stored in the storage unit 20. In the example shown in this figure, the storage unit 20 stores three-dimensional data for each of a plurality of objects.

[0033] 4 is a diagram showing a second example of information stored in the storage unit 20. In the example shown in this figure, the storage unit 20 stores evaluation models for a plurality of objects.

[0034] 5 is a diagram showing a first example of processing performed by the dividing unit 120. In the example shown in this figure, the three-dimensional data has translational symmetry. The dividing unit 120 then divides the three-dimensional data along a plane perpendicular to the axial direction of the translational symmetry (the up-down direction in the example shown in this figure), thereby generating a plurality of partial data. In the example shown in this figure, the three-dimensional data is divided into three partial data, but the three-dimensional data may also be divided into two partial data, or four or more partial data.

[0035] 6 is a diagram showing a second example of processing performed by the dividing unit 120. In the example shown in this figure, the three-dimensional data has rotational symmetry. The dividing unit 120 then divides the three-dimensional data into multiple pieces at the same angle (for example, 360° / n, where n is an integer) using the rotation axis of the rotation symmetry as the center of rotation, thereby generating multiple pieces of partial data. In the example shown in this figure, the three-dimensional data is also divided into three pieces of partial data, but the three-dimensional data may also be divided into two pieces of partial data, or four or more pieces of partial data.

[0036] Furthermore, if the three-dimensional data has mirror symmetry, the dividing section 120 may generate a plurality of partial data by utilizing this mirror symmetry.

[0037] Even if the overall shape of the three-dimensional data is not symmetrical, when the three-dimensional data is divided into multiple parts, at least one part may have symmetry. In this case, the dividing unit 120 may divide the three-dimensional data into multiple parts and then divide each part into multiple parts to generate multiple partial data. The process of dividing the three-dimensional data into multiple parts may be performed manually.

[0038] 7 is a diagram illustrating a first example of processing performed by the allocation unit 130. In the example shown in this figure, the division unit 120 generates two partial data. The shapes indicated by these two partial data are identical to each other. The allocation unit 130 then sets one of the two partial data as first partial data, i.e., partial data for learning, and the other as second partial data, i.e., partial data for evaluation.

[0039] FIG. 8 is a diagram illustrating a second example of processing performed by the allocation unit 130. In the example shown in this figure, the division unit 120 divides the three-dimensional data into a plurality of symmetrical parts, and further divides these plurality of parts to generate a plurality of partial data. The allocation unit 130 then assigns partial data for learning, i.e., first partial data, and partial data for evaluation, i.e., second partial data, to each of the plurality of parts. For example, when a plurality of partial data has been generated from a part corresponding to a bridge pier, the allocation unit 130 assigns this plurality of partial data to the first partial data and the second partial data. Furthermore, when a plurality of partial data has been generated from a part corresponding to a bridge deck, the allocation unit 130 assigns this plurality of partial data to the first partial data and the second partial data.

[0040] 9 is a diagram showing an example of the hardware configuration of the object evaluation device 10. The object evaluation device 10 includes a bus 1010, a processor 1020, a memory 1030, a storage device 1040, an input / output interface 1050, and a network interface 1060.

[0041] The bus 1010 is a data transmission path for transmitting and receiving data among the processor 1020, memory 1030, storage device 1040, input / output interface 1050, and network interface 1060. However, the method of connecting the processor 1020 and the like to each other is not limited to bus connection.

[0042] The processor 1020 is implemented by a CPU (Central Processing Unit), a GPU (Graphics Processing Unit), or the like.

[0043] The memory 1030 is a main storage device realized by a RAM (Random Access Memory) or the like.

[0044] The storage device 1040 is an auxiliary storage device realized by removable media such as a hard disk drive (HDD), a solid state drive (SSD), or a memory card, or a read-only memory (ROM), and has a recording medium. The recording medium of the storage device 1040 stores program modules that realize each function of the object evaluation device 10 (e.g., the three-dimensional data acquisition unit 110, the division unit 120, the allocation unit 130, the model acquisition unit 140, and the evaluation data generation unit 150). The processor 1020 loads each of these program modules into the memory 1030 and executes them, thereby realizing each function corresponding to the program module. The storage device 1040 may also function as the memory unit 20.

[0045] The input / output interface 1050 is an interface for connecting the object evaluation device 10 to various input / output devices.

[0046] The network interface 1060 is an interface for connecting the object evaluation device 10 to a network. This network is, for example, a LAN (Local Area Network) or a WAN (Wide Area Network). The method for connecting the network interface 1060 to the network may be a wireless connection or a wired connection.

[0047] 10 is a flowchart showing a first example of processing performed by the object evaluation device 10. A user of the object evaluation device 10 identifies an object to be evaluated and inputs information indicating this object into the object evaluation device 10. Then, the three-dimensional data acquisition unit 110 reads out the three-dimensional data of this object from the storage unit 20 (step S10).

[0048] The dividing unit 120 then generates a plurality of partial data sets using the three-dimensional data acquired by the three-dimensional data acquiring unit 110 (step S20). The allocating unit 130 then allocates the plurality of partial data sets to first partial data sets and second partial data sets (step S30). The model acquiring unit 140 then generates an evaluation model using the first partial data sets (step S40).

[0049] Thereafter, the evaluation data generation unit 150 generates evaluation data by inputting the second partial data into the evaluation model (step S50). Then, the evaluation data generation unit 150 uses this evaluation data to determine whether or not there is an abnormality in the object at the time the three-dimensional data is generated, and outputs the determination result (step S60). Note that if the object evaluation device 10 is a server, the output destination is a terminal. On the other hand, if the object evaluation device 10 is a terminal, the output destination is a display or a printing device.

[0050] 11 is a diagram illustrating a second example of processing performed by the object evaluation device 10. In this example, the allocation unit 130, the model acquisition unit 140, and the evaluation data generation unit 150 repeatedly generate an evaluation model and evaluation data for at least two pieces of partial data having the same three-dimensional shape, while changing the combination of data included in the first partial data and the second partial data. Then, the evaluation data generation unit 150 identifies an abnormal part of the object using the multiple pieces of evaluation data.

[0051] 11, three pieces of partial data X, Y, and Z are generated. The shapes of these three pieces of partial data are identical to each other. In this case, the allocation unit 130, the model acquisition unit 140, and the evaluation data generation unit 150 generate the evaluation model and the evaluation data six times.

[0052] Specifically, in the first round, the first partial data is X and Y, and the second partial data is Z. In the second round, the first partial data is X and Z, and the second partial data is Y. In the third round, the first partial data is Y and Z, and the second partial data is X. In the fourth round, the first partial data is Z, and the second partial data is X and Y. In the fifth round, the first partial data is Y, and the second partial data is X and Z. In the sixth round, the first partial data is X, and the second partial data is Y and Z.

[0053] Then, the evaluation data generation unit 150 calculates the difference between the shape based on the second partial data and the shape based on the evaluation data each time. For example, Z is evaluated in the first, fifth, and sixth times. The learned data at each time is different. If any of the evaluations determines that Z is abnormal, it is determined that there is a high possibility that there is an abnormality in the part corresponding to Z. Similarly, for Y, it is determined whether there is an abnormality based on the results of the second, fourth, and sixth times. For X, it is determined whether there is an abnormality based on the results of the third, fourth, and fifth times.

[0054] Fig. 12 is a flowchart showing an example of the processing performed by the object evaluation device 10 in the example shown in Fig. 11. The example shown in this figure is similar to the processing described using Fig. 10, except for the following points.

[0055] First, the object evaluation device 10 repeats the processing of steps S30 to S50 a predetermined number of times (step S52). The predetermined number of times is set within a range in which the combination of the first partial data and the second partial data can be changed.

[0056] The evaluation data generation unit 150 then uses the multiple pieces of evaluation data to determine whether or not an abnormality exists in the object, and if an abnormality exists, identifies the location of the abnormality. The evaluation data generation unit 150 then outputs the determination result and the identification result (step S62).

[0057] As described above, according to this embodiment, the dividing unit 120 divides at least a portion of the three-dimensional data of the object into a plurality of partial data. The shapes represented by these partial data are identical within a predetermined error range. The model acquisition unit 140 then generates an evaluation model using some of these partial data. The evaluation data generation unit 150 then generates evaluation data using the remaining partial data. Using this evaluation data, it is possible to determine whether or not there is an abnormality in the portion of the object corresponding to this partial data. Therefore, by using the object evaluation device 10, it is possible to determine whether or not there is an abnormality in the object without having to prepare training data in advance.

[0058] Although the embodiments of the present invention have been described above with reference to the drawings, these are merely examples of the present invention, and various other configurations can also be adopted.

[0059] In addition, in the flowcharts used in the above description, multiple steps (processes) are described in order, but the order of execution of the steps performed in each embodiment is not limited to the order described. In each embodiment, the order of the steps shown in the drawings can be changed to the extent that the content is not affected. Furthermore, the above-mentioned embodiments can be combined to the extent that the content is not contradictory.

[0060] Some or all of the above-described embodiments can be described as, but are not limited to, the following supplementary notes. 1. A model acquisition means for acquiring an evaluation model generated by dividing three-dimensional data representing the shape of an object into a plurality of parts and using some of the plurality of partial data as training data; an evaluation data generating means for generating evaluation data for evaluating the presence or absence of an abnormality in the object by using at least a portion of the remaining plurality of partial data as input data for the evaluation model; Equipped with the three-dimensional shapes indicated by the at least two partial data are identical within a range including a predetermined error; An object evaluation device, wherein at least one of the at least two partial data whose three-dimensional shapes are identical to each other is included in the first partial data for generating the learning data, and at least one other of the at least two partial data is included in the second partial data that becomes the input data. 2. In the object evaluation device described in 1 above, a dividing means for dividing the three-dimensional data into the plurality of partial data; an allocation means for allocating the plurality of partial data to the first partial data and the second partial data; Equipped with The model acquisition means generates the evaluation model using the training data including the first partial data. 3. In the object evaluation device described in 2 above, the allocation means, the model acquisition means, and the evaluation data generation means repeatedly generate the evaluation model and the evaluation data while changing a combination of data included in the first partial data and data included in the second partial data for the at least two partial data having the same three-dimensional shape; The evaluation data generating means identifies an abnormal part of the object using a plurality of pieces of evaluation data. 4. In the object evaluation device according to any one of the above items 1 to 3, The object evaluation device, wherein the object is at least a part of a structure. 5. In the object evaluation device according to any one of 1 to 4 above, The object evaluation device, wherein the evaluation model uses an autoencoder. 6. In the object evaluation device according to any one of 1 to 5 above, the evaluation model is generated for each of the plurality of objects; The model acquisition means acquires the evaluation model corresponding to the object. 7. The computer Obtain an evaluation model generated by dividing the three-dimensional data representing the shape of the object into multiple parts and using some of the multiple partial data as training data; generating evaluation data for evaluating the presence or absence of an abnormality in the object by using at least a portion of the remaining plurality of partial data as input data for the evaluation model; the three-dimensional shapes indicated by the at least two partial data are identical within a range including a predetermined error; An object evaluation method, wherein at least one of the at least two partial data having the same three-dimensional shape is included in the first partial data for generating the learning data, and at least one other of the at least two partial data is included in the second partial data that becomes the input data. 8. In the object evaluation method described in 7 above, The computer Dividing the three-dimensional data into the plurality of partial data; allocating the plurality of partial data to the first partial data and the second partial data; An object evaluation method for generating the evaluation model using the training data including the first partial data. 9. In the object evaluation method described in 8 above, The computer repeating the generation of the evaluation model and the generation of the evaluation data while changing a combination of data included in the first partial data and data included in the second partial data for the at least two partial data having the same three-dimensional shape; An object evaluation method for identifying an abnormal part of the object using a plurality of the evaluation data. 10. In the object evaluation method according to any one of 7 to 9 above, The object evaluation method, wherein the object is at least a part of a structure. 11. In the object evaluation method according to any one of the above items 7 to 10, The object evaluation method, wherein the evaluation model uses an autoencoder. 12. In the object evaluation method according to any one of 7 to 11 above, the evaluation model is generated for each of the plurality of objects; The computer acquires the evaluation model corresponding to the object. 13. On the computer, A model acquisition function that acquires an evaluation model generated by dividing the 3D data representing the shape of the object into multiple parts and using some of the multiple partial data as training data; and an evaluation data generation function that generates evaluation data for evaluating the presence or absence of an abnormality in the object by using at least a portion of the remaining plurality of partial data as input data for the evaluation model; It records the program that allows you to the three-dimensional shapes indicated by the at least two partial data are identical within a range including a predetermined error; A computer-readable recording medium, in which at least one of the at least two partial data whose three-dimensional shapes are identical to each other is included in the first partial data for generating the learning data, and at least one other of the at least two partial data is included in the second partial data that becomes the input data. 14. The recording medium according to claim 13, The program causes the computer to: a division function for dividing the three-dimensional data into the plurality of partial data; an allocation function for allocating the plurality of partial data to the first partial data and the second partial data; Let them have The model acquisition function generates the evaluation model using the training data including the first partial data. 15. The recording medium according to claim 14, the allocation function, the model acquisition function, and the evaluation data generation function repeatedly generate the evaluation model and the evaluation data while changing a combination of data included in the first partial data and data included in the second partial data for the at least two partial data having the same three-dimensional shape; The evaluation data generation function identifies an abnormal part of the object using a plurality of the evaluation data. 16. The recording medium according to any one of the above items 13 to 15, The object is a recording medium that is at least a part of a structure. 17. The recording medium according to any one of the above items 13 to 16, The evaluation model uses an autoencoder. 18. The recording medium according to any one of the above items 13 to 17, the evaluation model is generated for each of the plurality of objects; The model acquisition function acquires the evaluation model corresponding to the object. 19. The program according to any one of items 13 to 18 above. [Explanation of symbols]

[0061] 10. Object evaluation device 20 Memory section 110 3D data acquisition unit 120 Division 130 Allocation Section 140 Model Acquisition Department 150 Evaluation data generation unit

Claims

1. a model acquisition means for acquiring an evaluation model generated by dividing three-dimensional data representing the shape of an object into a plurality of parts and using some of the plurality of partial data as training data; an evaluation data generating means for generating evaluation data for evaluating the presence or absence of an abnormality in the object by using at least a portion of the remaining plurality of partial data as input data for the evaluation model; Equipped with the three-dimensional shapes indicated by the at least two partial data are identical within a range including a predetermined error; An object evaluation device, wherein at least one of the at least two partial data whose three-dimensional shapes are identical to each other is included in the first partial data for generating the learning data, and at least one other of the at least two partial data is included in the second partial data that becomes the input data.

2. 2. The object evaluation device according to claim 1, a dividing means for dividing the three-dimensional data into the plurality of partial data; an allocation means for allocating the plurality of partial data to the first partial data and the second partial data; Equipped with The model acquisition means generates the evaluation model using the training data including the first partial data.

3. 3. The object evaluation device according to claim 2, the allocation means, the model acquisition means, and the evaluation data generation means repeatedly generate the evaluation model and the evaluation data while changing a combination of data included in the first partial data and data included in the second partial data for the at least two partial data having the same three-dimensional shape; The evaluation data generating means identifies an abnormal part of the object using a plurality of pieces of evaluation data.

4. The object evaluation device according to any one of claims 1 to 3, The object evaluation device, wherein the object is at least a part of a structure.

5. The object evaluation device according to any one of claims 1 to 3, The object evaluation device, wherein the evaluation model uses an autoencoder.

6. The object evaluation device according to any one of claims 1 to 3, the evaluation model is generated for each of the plurality of objects; The model acquisition means acquires the evaluation model corresponding to the object.

7. The computer Obtaining an evaluation model generated by dividing three-dimensional data representing the shape of the object into a plurality of parts and using some of the plurality of partial data as training data; generating evaluation data for evaluating the presence or absence of an abnormality in the object by using at least a portion of the remaining plurality of partial data as input data for the evaluation model; the three-dimensional shapes indicated by the at least two partial data are identical within a range including a predetermined error; An object evaluation method, wherein at least one of the at least two partial data whose three-dimensional shapes are identical to each other is included in the first partial data for generating the learning data, and at least one other of the at least two partial data is included in the second partial data that becomes the input data.

8. On the computer, a model acquisition function for acquiring an evaluation model generated by dividing three-dimensional data representing the shape of an object into a plurality of parts and using some of the plurality of partial data as training data; an evaluation data generation function that generates evaluation data for evaluating the presence or absence of an abnormality in the object by using at least a portion of the remaining plurality of partial data as input data for the evaluation model; Let them have the three-dimensional shapes indicated by the at least two partial data are identical within a range including a predetermined error; A program in which at least one of the at least two partial data whose three-dimensional shapes are identical to each other is included in the first partial data for generating the learning data, and at least one other of the at least two partial data is included in the second partial data that becomes the input data.

Images