Image processing method, image processing device, and image processing program

Abstract

To make it possible to estimate a two-dimensional distribution which is not limited to a horizontal direction when estimating, from an image showing a distribution of one physical quantity related to an earth structure, a distribution of another physical quantity different from one physical quantity by image conversion.SOLUTION: An image processing method includes: a step (S12) of converting physical quantities related to an earth structure (S) to be analyzed into discrete values according to characteristics of the physical quantities; a step (S13) of generating a pre-conversion image (I1) with the same shape as a cross section of the earth structure (S) showing a distribution of the discrete values within the earth structure (S); and a step (S14) of converting the pre-conversion image (I1) into a rectangular input image (I2).SELECTED DRAWING: Figure 2

Description

[Technical Field]

[0001] The present invention relates to an image processing method, an image processing device, and an image processing program. [Background technology]

[0002] For example, as described in the following non-patent document 1, a method has been proposed in which, when an image showing the distribution of hydraulic conductivity within an earth structure is input, a generator is constructed using a generative adversarial network to generate an image showing the distribution of pressure head within the earth structure, and the generator is used to estimate the distribution of pressure head within the earth structure. [Prior art documents] [Non-patent literature]

[0003] [Non-Patent Document 1] Alexander Y. Sun,Discovering state-parameter mappings in subsurface models using generative adversarial networks,Geophysical Research Letters,October 2018 Summary of the Invention [Problem to be solved by the invention]

[0004] However, the conventional technology described in the above-mentioned Non-Patent Document 1 was only able to estimate the pressure head distribution in the horizontal direction. The present invention aims to enable estimation of a two-dimensional distribution that is not limited to the horizontal direction when estimating the distribution of another physical quantity different from a given physical quantity related to an earth structure by image conversion from an image showing the distribution of the given physical quantity. [Means for solving the problem]

[0005] An image processing method according to one embodiment of the present invention includes the steps of converting physical quantities related to an earth structure to be analyzed into discrete values ​​corresponding to the characteristics of the physical quantities, generating a pre-conversion image of the same shape as the cross-section of the earth structure, which shows the distribution of the discrete values ​​within the earth structure, and geometrically converting the pre-conversion image into a rectangular input image.

[0006] An image processing device according to another aspect of the present invention comprises a first conversion unit that converts physical quantities related to an earth structure to be analyzed into discrete values ​​corresponding to the characteristics of the physical quantities, a generation unit that generates a pre-conversion image of the same shape as a cross-section of the earth structure, showing the distribution of the discrete values ​​within the earth structure, and a second conversion unit that geometrically converts the pre-conversion image into a rectangular input image.

[0007] Each aspect of the present invention may be realized by a computer. In this case, the image processing program that causes the computer to operate as each part (software element) of the image processing device to realize the image processing device on the computer, and the computer-readable recording medium on which it is recorded, also fall within the scope of the present invention. [Effects of the Invention]

[0008] According to one aspect of the present invention, when estimating the distribution of another physical quantity different from a distribution of one physical quantity related to an earth structure by image conversion from an image showing the distribution of the one physical quantity, it is possible to estimate a two-dimensional distribution that is not limited to the horizontal direction. [Brief explanation of the drawings]

[0009] [Figure 1] 1 is a diagram showing an example of an image to be processed by an image processing method according to an embodiment of the first invention; [Figure 2] 1 is a flowchart showing the flow of the method. [Figure 3] FIG. 10 is a diagram showing an example of the content performed in the method. [Figure 4] FIG. 10 is a diagram showing an example of the content performed in the method. [Figure 5]FIG. 10 is a diagram showing the difference between the content performed in the method and the content performed in a method different from the method. [Figure 6] FIG. 10 is a diagram showing an example of the content performed in the method. [Figure 7] FIG. 10 is a block diagram showing a schematic configuration of an image processing apparatus according to an embodiment of a second invention. [Figure 8] 10 is a flowchart showing the flow of a generator construction method according to an embodiment of the third invention. [Figure 9] FIG. 10 is a diagram showing an example of the content performed in the method. [Figure 10] FIG. 10 is a diagram illustrating the operation of a generator constructed by the same method. [Figure 11] FIG. 10 is a block diagram showing a schematic configuration of a generator construction device according to an embodiment of the fourth invention. [Figure 12] 10 is a flowchart showing the flow of an estimation method according to an embodiment of the fifth invention. [Figure 13] FIG. 10 is a diagram showing an example of the content performed in the method. [Figure 14] FIG. 10 is a block diagram showing a schematic configuration of an estimation device according to an embodiment of the sixth invention. [Figure 15] 10 is a histogram showing the distribution of evaluation values ​​(SSIM) of images for output obtained using the methods or devices according to the first to sixth embodiments of the present invention. [Figure 16] 10 is a histogram showing the distribution of evaluation values ​​(errors) of images for output obtained using the methods or devices according to the embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0010] <First embodiment of the invention> First, an embodiment of the first invention (image processing method) will be described.

[0011] [Image to be processed] The image to be processed by the image processing method according to this embodiment shows the distribution of physical quantities related to an earth structure S to be analyzed. The earth structure S to be analyzed includes all structures constructed of earth, including embankments (dams). The cross-sectional shape of the earth structure S is generally non-rectangular. When the earth structure S is a dam, its cross-sectional shape is trapezoidal. Furthermore, the earth structure S has its own unique permeability distribution depending on the properties of the soil and sand used. That is, the earth structure S may have a uniform permeability coefficient or may have two or more regions with different permeability coefficients. Below, we will explain using an example image showing the cross-section of a dam S having a first portion P1, a second portion P2, and a third portion P3, as shown in Figure 1. The second portion P2 forms the center of the dam and has the lowest permeability coefficient. The first portion P1 forms the upstream side of the dam and has a higher permeability coefficient than the second portion P2. The third portion P3 forms the downstream side of the dam and has a higher permeability coefficient than the first portion P1. The above-mentioned permeability coefficient distribution is an example, and the permeability coefficient distribution of the earth structure S that is the processing target of this image processing method is not limited to the above.

[0012] [Image processing method flow] As shown in FIG. 2, the image processing method according to this embodiment includes a first acquisition step S11, a first conversion step S12, a generation step S13, a second conversion step S14, and a first output step S15.

[0013] (First acquisition step) In the first acquisition step S11, physical quantities related to the earth structure S to be analyzed are acquired. The acquisition of these physical quantities can be performed using, for example, a first acquisition unit 131 (see FIG. 7) of an image processing device 1 (second invention) to be described later.

[0014] (First conversion step) After acquiring the physical quantities, the process proceeds to a first conversion step S12. In the first conversion step S12, the physical quantities acquired in the first acquisition step S11 are converted into discrete values ​​according to the characteristics of the physical quantities. The physical quantities in this embodiment are the hydraulic conductivity or the pressure head.

[0015] When the physical quantity is the hydraulic conductivity, in the first conversion step S12, the hydraulic conductivity is converted into gradation such that the brightness or saturation changes by one step every time the hydraulic conductivity is multiplied by 10, as shown in the upper part of FIG. -1 ~10 -8 Permeability is a physical quantity that can take on a wide range of values, on the order of cm / s. For this reason, if the permeability coefficient were simply converted into gradation, the difference in color tone would be too small, making it difficult to distinguish differences in permeability coefficient values. However, by using the above configuration, even small differences in value can be made to appear with relatively large differences in color tone. This makes it possible to obtain an input image I2 (described in detail below) that can be used to generate an estimated image I6 (described in detail below) that accurately reflects the distribution of permeability coefficients.

[0016] On the other hand, if the physical quantity is pressure head, in the first conversion step S12, as shown in the lower part of Figure 3, the saturation of one hue increases as the pressure head value increases in the positive direction, and the saturation of another hue increases as the pressure head value increases in the negative direction. By doing this, the area (line L indicating the infiltration surface) corresponding to the part that is neither saturated nor unsaturated (pressure head is 0) is displayed in white, with the lowest saturation. This makes it possible to obtain an input image I2 from which an estimated image I6 can be generated that accurately reflects the position of line L indicating the infiltration surface within the earth structure S.

[0017] The above-described conversion of discrete values ​​can be performed using, for example, a first conversion unit 132 (see FIG. 7) of the image processing device 1, which will be described later.

[0018] (Generation step) After the physical quantities are converted into discrete values, the process proceeds to generation step S13 as shown in Figure 2. In generation step S13, a pre-conversion image I1 is generated, which has the same shape as the cross-section of the earth structure S and shows the distribution of discrete values ​​within the earth structure S. The pre-conversion image I1 is an image which has the same shape as the cross-section of the earth structure S and shows the distribution of discrete values ​​within the earth structure S.

[0019] The generation step S13 according to this embodiment includes a partitioning step and an extraction step. In the initial partitioning step, the pre-conversion image I1 is partitioned into a plurality of micro-regions. In the partitioning step according to this embodiment, the pre-conversion image I1 is partitioned so that the micro-regions at the top of the pre-conversion image I1 are smaller than the micro-regions at the bottom. Specifically, the size of the micro-regions is gradually changed so that the micro-regions become smaller toward the top of the pre-conversion image I1. After the pre-conversion image I1 has been partitioned into a plurality of micro-regions, the process proceeds to the extraction step. In the extraction step, one pixel p is extracted from each micro-region. In this way, a group of pixels arranged so that they become denser toward the top of the pre-conversion image I1 is obtained, as shown on the left side of Figure 4, for example.

[0020] The pre-conversion image I1 can be generated using, for example, a generation unit 133 (see FIG. 7) of the image processing device 1, which will be described later.

[0021] (Second conversion step) After generating the pre-conversion image I1, the process proceeds to the second conversion step S14 as shown in FIG. 2. In the second conversion step S14, the pre-conversion image I1 is geometrically converted into a rectangular input image I2. Specifically, multiple pixels p arranged in the pre-conversion image I1 (trapezoid) as shown on the left side of FIG. 4 are each moved so that they are distributed throughout the rectangle as shown on the right side of FIG. 4. The distribution may be completely uniform, or there may be some unevenness. Then, in the second conversion step S14 according to this embodiment, after the pixels p have been moved, interpolation is performed between pixels p and p. As described above, the arrangement of the pixels p extracted in the extraction step becomes denser toward the top of the pre-conversion image I1. Therefore, the pixels p at the top of the pre-conversion image I1 are moved more significantly. This allows for the acquisition of an input image I2 in which the entire image is uniformly interpolated. In particular, if the interval between pixels p and p becomes too wide, it may become unclear whether the line L indicating the infiltration surface between pixels p and p is convex upward or downward, and the curved portion of the line L indicating the infiltration surface may not be accurately interpolated (it may be necessary to perform linear interpolation as shown in the left side of Figure 5). However, by doing as described above, the interval between pixels p and p does not become too wide. Therefore, as shown in the right side of Figure 5, the curved portion of the line L indicating the infiltration surface may be accurately interpolated. Note that in the second conversion step S14, the pre-conversion image I1 (trapezoid) may be geometrically converted (to aggregate the pixels) into a rectangular input image I2 that is smaller than the pre-conversion image I1.

[0022] In addition, if the earth structure S has two or more regions with different hydraulic conductivity, as shown in FIG. 1, and the physical quantity indicated by the pre-transformed image I1 is the hydraulic conductivity, the pre-transformed image I1 may be geometrically transformed in the second transformation step S14 so that the area of ​​the region with a relatively high hydraulic conductivity in the input image I2 is larger than the area of ​​the region with a relatively high hydraulic conductivity in the pre-transformed image I1. For example, as shown on the left side of FIG. 6, if the area of ​​region R1 indicating the first region P1 in the input image I2 is smaller than the areas of regions R2 and R3 indicating the second region P2 and third region P3, which have lower hydraulic conductivity than the first region P1, the pre-transformed image I1 may be geometrically transformed so that the regions R1 to R3 are congruent, as shown on the right side of FIG. 6. It is known that the pressure head within the earth structure S changes rapidly at the boundary of the region with a high hydraulic conductivity. Therefore, if the region with a relatively high hydraulic conductivity in the input image I2 is narrow, the generator G, described below, may not be able to capture the change in pressure head. However, by doing as described above, it becomes easier for the generator G to capture changes in pressure head. Note that, during the geometric transformation, the extent to which the area of ​​the region showing the portion with a relatively high hydraulic conductivity is increased can be changed as appropriate depending on the value of the hydraulic conductivity.

[0023] The above geometric transformation can be performed, for example, by using a second transformation unit 134 (see FIG. 7) described later of the image processing device 1. The second transformation step S14 may be performed before the first transformation step S12.

[0024] (First output step) After the geometric transformation, the process proceeds to a first output step S15. In the first output step S15, the input image I2 is output (put into a form usable for learning and estimation). The output can be performed using, for example, a first output unit 12 and a first output control unit 135 (see FIG. 7) of the image processing device 1, which will be described later. Note that if there is no need to output the input image I2 (for example, if the image processing method includes a generated image output step S22 and an analysis step S23 of a generator construction method, which will be described later, or an estimation step S32 of an estimation method, and the input image I2 generated in the second conversion step S14 is used as is in the generated image output step S22, the analysis step S23, or the estimation step S32), then the first output step S15 does not need to be performed.

[0025] [Effects of image processing methods] According to the image processing method of the present embodiment described above, by converting physical quantities into discrete values ​​and rectangularizing the image through geometric transformation, particularly by converting the image into discrete values, the color tone of the image is adjusted to capture the characteristics of the image in the construction of the generator described below. Therefore, from an image showing the distribution of one physical quantity related to an earth structure S, an input image I2 can be obtained, which is used to estimate the distribution of another physical quantity different from the one physical quantity through image transformation (a method in which the input image I2 is input into the generator G and an estimated image I6 is generated). Using this input image I2, it becomes possible to estimate two-dimensional distributions, including not only horizontal but also vertical directions. Furthermore, if the input image I2 represents a hydraulic conductivity distribution, it becomes possible to estimate a pressure head distribution that includes not only saturated regions but also unsaturated regions.

[0026] <Second embodiment of the invention> Next, an embodiment of the second invention (image processing device 1) will be described.

[0027] [Configuration of image processing device] As shown in FIG. 7, the image processing device 1 includes a first input unit 11, a first output unit 12, and a first control unit 13.

[0028] [First input section] The first input unit 11 is composed of at least one of a communication module that receives data and signals from other devices (e.g., a measuring device that measures physical quantities, a storage device that stores physical quantities, etc.), a terminal that is connected to other devices, and a drive that reads information from a recording medium.

[0029] [First output section] The first output unit 12 is composed of at least one of a communication module that transmits data and signals to other devices (e.g., the generator construction device 2 and the estimation device 3 described below), a terminal that connects to other devices, and a drive that writes information to a recording medium.

[0030] [First control section] The first control unit 13 according to this embodiment includes a first acquisition unit 131 , a first conversion unit 132 , a generation unit 133 , a second conversion unit 134 , and a first output control unit 135 .

[0031] (First Acquisition Department) The first acquisition unit 131 controls the first input unit 11 to acquire physical quantities related to the earth structure S to be analyzed. As a result, if the first input unit 11 is a communication module or a terminal, the first input unit 11 receives data on the physical quantities from another device. Also, if the first input unit 11 is a drive, the first input unit 11 reads the data on the physical quantities from a recording medium.

[0032] (First conversion unit) The first conversion unit 132 converts the physical quantity acquired by the first acquisition unit 131 into a discrete value according to the characteristics of the physical quantity. Specifically, the first conversion unit 132 executes the same operations as those described in the first conversion step S12 (see FIG. 2) in the image processing method (first invention) described above.

[0033] (Generation part) The generation unit 133 generates a pre-conversion image I1 having the same shape as the cross-section of the earth structure S, which shows the distribution of discrete values ​​within the earth structure S. The generation unit 133 according to this embodiment includes a partitioning unit 133a and an extraction unit 133b. The partitioning unit 133a partitions the pre-conversion image I1 into a plurality of micro-regions. Specifically, the partitioning unit 133a performs the same operations as those described in the partitioning step of the image processing method above. The extraction unit 133b extracts pixels p one by one from each micro-region. Specifically, the extraction unit 133b performs the same operations as those described in the extraction step of the image processing method above.

[0034] (Second conversion part) The second conversion unit 134 geometrically converts the pre-conversion image I1 into a rectangular input image I2. Specifically, the second conversion unit 134 executes the same operations as those described in the second conversion step S14 (see FIG. 2) in the image processing method described above.

[0035] (First output control unit) The first output control unit 135 controls the first output unit 12 to output the input image I2. Furthermore, the first output control unit 135 according to this embodiment also controls the first output unit 12 to output the pre-conversion image I1. As a result, if the first output unit 12 is a communication module or a terminal, the first output unit 12 transmits data of the pre-conversion image I1 and the input image I2 to another device. Furthermore, if the first output unit 12 is a drive, the first output unit 12 writes data of the pre-conversion image I1 and the input image I2 to a recording medium. Note that if there is no need to output the pre-conversion image I1 and the input image I2 (for example, if the image processing device 1 includes a generated image output unit 242 or an analysis unit 243 of the generator construction device 2, or an estimation unit 342 of the estimation device 3, which will be described later, and the input image I2 generated by the second conversion unit 134 is used as is by the generated image output unit 242, the analysis unit 243, or the estimation unit 342), the first control unit 13 does not need to include the first output control unit 135.

[0036] [Effects of image processing device] As described above, the image processing device 1 according to this embodiment, like the image processing method described above, converts physical quantities into discrete values ​​and rectangularizes the image through geometric transformation. In particular, the discrete value conversion transforms the image color tones to capture the image's characteristics in the construction of a generator, as described below. This allows an input image I2 to be obtained from an image showing the distribution of one physical quantity related to an earth structure S, which can be used to estimate the distribution of another physical quantity through image transformation. Using this input image I2, it is possible to estimate two-dimensional distributions, including not only horizontal but also vertical. Furthermore, if the input image I2 represents a hydraulic conductivity distribution, it is possible to estimate a pressure head distribution that includes not only saturated regions but also unsaturated regions.

[0037] <Third embodiment of the invention> Next, an embodiment of the third invention (generator construction method and generator G) will be described.

[0038] [Generator construction method] The generator construction method according to this embodiment includes, for example, as shown in FIG. 8, a second acquisition step S21, a generated image output step S22, an analysis step S23, a classification result output step S24, a judgment step S25, a learning step S26, and a second output step S27.

[0039] (Second acquisition step) In the initial second acquisition step S21, an input image I2 is acquired. Furthermore, in the second acquisition step S21 according to this embodiment, a pre-transformation image I1 is also acquired. The acquisition of the input image I2 and the pre-transformation image I1 can be performed, for example, using a second acquisition unit 241 (see FIG. 11 ) of a generator construction device 2 (fourth invention) described later. The input image I2 and the pre-transformation image I1 may be generated by the image processing device 1, or may be generated by a device other than the image processing device 1. Furthermore, if there is no need to acquire the input image I2 and the pre-transformation image I1 (for example, if the generator construction method includes a generation step S13 and a second conversion step S14 of the image processing method, and the pre-transformation image I1 generated in the generation step S13 and the input image I2 generated in the second conversion step S14 are used as is in a generated image output step S22 or an analysis step S23 described later), the second acquisition step S21 may not be performed.

[0040] (Generated image output step) After acquiring the input image I2, the process proceeds to the generated image output step S22. In the generated image output step S22, as shown in FIG. 9, the input image I2 acquired in the second acquisition step S21 is input to a generator before or in the middle of training (hereinafter referred to as the untrained generator g) or to a trained generator G, and a generated image I3 showing the distribution of the second physical quantity is output. The first physical quantity is one of the permeability coefficient and the pressure head, and the second physical quantity is the other. When the first physical quantity is the permeability coefficient, the generated image I3 showing the distribution of the pressure head can be obtained. On the other hand, when the first physical quantity is the pressure head, the generated image I3 showing the distribution of the permeability coefficient can be obtained. The output of the generated image I3 can be performed, for example, using the generated image output unit 242 (see FIG. 11) of the generator construction device 2 described later.

[0041] (Analysis step) After outputting the generated image I3, the process proceeds to analysis step S23. In analysis step S23, the pre-conversion image I1 is analyzed to obtain a correct image I5 showing the distribution of the second physical quantity. The correct image I5 is obtained by geometrically transforming a pre-conversion analysis image I4 (see FIG. 9) obtained by analyzing the pre-conversion image I1 so that it becomes rectangular. The analysis of this pre-conversion image I1 can be performed, for example, using an analysis unit 243 (see FIG. 11) of the generator construction device 2 described later. The analysis of the pre-conversion image I1 can also be performed by a conventional method. Note that analysis step S23 may be performed before generated image output step S22, or may be performed in parallel with generated image output step S22.

[0042] (Classification result output step) After obtaining the generated image I3 and the supervised image I5, the process proceeds to a classification result output step S24. In the classification result output step S24, the generated image I3 and the supervised image I5 are input to a classifier D, which outputs a classification result as to the authenticity of the generated image I3. This classification result can be output using, for example, a classification result output unit 244 (see FIG. 11) of the generator construction device 2, which will be described later. Note that in the classification result output step S24, depending on the structure of the untrained generator g, the pair of the generated image I3 and the input image I2, and the pair of the supervised image I5 and the input image I2 may be input to the classifier D.

[0043] (Decision step) After the classification result is output, the process proceeds to judgment step S25. In judgment step S25, the classification result of the authenticity of the generated image I3 is judged. This classification result judgment can be performed, for example, using a judgment unit 245 (see FIG. 11) of the generator construction device 2 described later. Note that in judgment step S25, items other than authenticity (such as whether or not learning of the untrained generator g has been performed a predetermined number of times) may also be judged. Note that if there is no need to judge the classification result (for example, if the classification result is output only in the case of either "true (genuine)" or "fake (counterfeit)" in classification result output step S24), this judgment step S25 does not need to be performed.

[0044] (Learning Steps) If the classification result is determined to be "fake" in the above-mentioned determination step S25 (step S25: NO), the process proceeds to the training step S26 as shown in FIG. 8. In the training step S26, at least an untrained generator g is trained (the untrained generator g and the untrained discriminator d may also be trained separately). In this embodiment, the untrained generator g trained in the training step S26 and the generator G constructed by training are U-Nets. In this way, as shown in FIG. 10, for example, at least one layer of the encoder of the untrained generator g or the generator G is skip-connected to the corresponding layer of the decoder. Then, when the encoder performs max pooling, position information of the feature extracted in any layer is transmitted to the corresponding layer of the decoder via the skip connection. This makes it possible to obtain a generated image I3 in which local features are maintained. The training of this untrained generator g can be performed, for example, using the training unit 246 (see FIG. 11) of the generator construction device 2 described later.

[0045] After the learning step S26 is performed, the process returns to step S22 using the trained untrained generator g or generator G, as shown in Fig. 8. That is, the generated image output step S22, the classification result output step S24, the judgment step S25, and the learning step S26 are repeated until the classification result is judged to be "true" in the judgment step S25. Each time these steps are repeated, the estimation accuracy of the untrained generator g improves.

[0046] (Second output step) If it is determined in the above determination step S25 that the classification result is "true" (step S25: YES), the process proceeds to the second output step S27. In the second output step S27, a generator G is output. This generator G can be output using, for example, the second output unit 22 and the second output control unit 247 (see FIG. 11) of the image processing device 1, which will be described later. Note that if there is no need to output the generator G (for example, if the generator construction method includes an estimation step S32 of the estimation method, which will be described later, and the generator G constructed in the above learning step S26 is used as is in the estimation step S32), this second output step S27 does not have to be performed.

[0047] [Generator construction method and other details] In addition, when the first physical quantity is pressure head and the second physical quantity is permeability coefficient, the generator construction method may further include a step of obtaining the permeability coefficient of at least a portion of the earth structure S as prior information before performing the learning step S26. Then, in the learning step S26, the prior information may be provided to the untrained generator g when training the untrained generator g. When attempting to obtain a generated image I3 showing the distribution of permeability coefficients from an input image I2 showing the distribution of pressure heads, it is known to be more difficult to obtain a generated image I3 with high accuracy than when obtaining a generated image I3 showing the distribution of pressure heads from an input image I2 showing the distribution of permeability coefficients. However, by providing prior information to the untrained generator g when constructing the generator G, the accuracy of the generated image I3 (estimated image I6) can be improved.

[0048] [Generator] The generator G (trained model) constructed by the generator construction method according to this embodiment is constructed using a generative adversarial network. Generator G is constructed by inputting input image I2 to untrained generator g, outputting generated image I3, inputting generated image I3 and ground truth image I5 to classifier D, outputting a classification result of the authenticity of generated image I3, and training at least untrained generator g if the classification result is a fake. When input image I2 is input, generator G constructed in this way generates estimated image I6. Estimated image I6 refers to generated image I3, which is generated by the trained generator G and is not classified as a fake by classifier D.

[0049] [Effects of generator construction method] According to the generator construction method of this embodiment described above, by using an input image I2 in which physical quantities have been converted into discrete values ​​and whose shape has been geometrically transformed into a rectangle, and by repeating (using a generative adversarial network) the generated image output step S22, the classification result output step S24, and the learning step S26, the accuracy of the generator G is improved and an estimated image I6 showing a two-dimensional distribution that is not limited to the horizontal direction but also includes the vertical direction can be obtained. Furthermore, if the input image I2 shows a hydraulic conductivity distribution, an estimated image I6 showing a pressure head distribution that includes not only saturated regions but also unsaturated regions can be obtained.

[0050] <Fourth embodiment of the invention> Next, an embodiment of the fourth invention (generator construction device 2) will be described.

[0051] [Configuration of generator construction device] As shown in FIG. 11, the generator construction device 2 according to this embodiment includes a second input unit 21, a second output unit 22, a storage unit 23, and a second control unit 24.

[0052] [Second input section] The second input unit 21 is composed of at least one of a communication module that receives data and signals from other devices (for example, the image processing device 1 described above), a terminal that connects to other devices, and a drive that reads information from a recording medium.

[0053] [Second output unit] The second output unit 22 is composed of at least one of a communication module that transmits data and signals to other devices (such as the estimation device 3 described below), a terminal that connects to other devices, and a drive that writes information to a recording medium.

[0054] [Storage section] The storage unit 23 stores an unlearned generator g. The storage unit 23 is also configured to be able to store a learned generator G.

[0055] [Second control section] The second control unit 24 includes a second acquisition unit 241, a generated image output unit 242, an analysis unit 243, a classification result output unit 244, a judgment unit 245, a learning unit 246, and a second output control unit 247.

[0056] (Second Acquisition Department) The second acquisition unit 241 controls the second input unit 21 to acquire the input image I2. As a result, if the second input unit 21 is a communication module or a terminal, the second input unit 21 receives data for the input image I2 from another device. If the second input unit 21 is a drive, the second input unit 21 reads data for the input image I2 from a recording medium. Note that if there is no need to acquire the input image I2 (for example, if the generator construction device 2 is equipped with the second conversion unit 134 of the image processing device 1, and the input image I2 generated by the second conversion unit 134 is used as is by the generated image output unit 242 or the analysis unit 243 described later), the second control unit 24 does not need to be equipped with this second acquisition unit 241.

[0057] (Generated image output section) The generated image output unit 242 inputs the input image I2 acquired by the second acquisition unit 241 to the untrained generator g or generator G, and outputs a generated image I3. Specifically, the generated image output unit 242 executes the same operations as those described in the generated image output step S22 (see FIG. 8) in the generator construction method (third invention).

[0058] (Analysis Department) The analysis unit 243 analyzes the pre-conversion image I1 and obtains the correct image I5. Specifically, the analysis unit 243 executes the same operations as those described in the analysis step S23 (see FIG. 8) in the above-described generator construction method.

[0059] (Classification result output section) The classification result output unit 244 inputs the generated image I3 and the correct image I5 to the classifier D and outputs a classification result of the authenticity of the generated image I3. Specifically, the classification result output unit 244 executes the same operations as those described in the classification result output step S24 (see FIG. 8) in the above generator construction method.

[0060] (Judgment Department) The judgment unit 245 judges the result of the identification of the authenticity of the generated image I3. Specifically, the identification result output unit 244 executes the same operations as those described in the judgment step S25 (see FIG. 8) in the above-described generator construction method. Note that if there is no need to judge the identification result (for example, if the identification result output unit 244 outputs the identification result only when it is either "true" or "fake"), the second control unit 24 may not be provided with this judgment unit 245.

[0061] (Study Department) The learning unit 246 trains at least the untrained generator g when the determination unit 245 determines that the classification result is fake. Specifically, the learning unit 246 executes the same operations as those described in the learning step S26 (see FIG. 8) in the generator construction method.

[0062] (Second output control unit) The second output control unit 247 controls the second output unit 22 to output the generator G. As a result, if the second output unit 22 is a communication module or a terminal, the second output unit 22 transmits the data of the generator G to another device. Also, if the second output unit 22 is a drive, the second output unit 22 writes the data of the generator G to a recording medium. Note that if there is no need to output the generator G (for example, if the generator construction device 2 is equipped with an estimation unit 342 of the estimation device 3 described below, and the estimation unit 342 uses the generator G constructed by the learning unit 246 as is), the second control unit 24 does not need to be equipped with the second output control unit 247.

[0063] [Generator construction device and others] In addition, when the first physical quantity is a pressure head and the second physical quantity is a permeability coefficient, the generator construction device 2 may further include a prior information acquisition unit that acquires the permeability coefficient of at least a portion of the earth structure S in advance as prior information. The learning unit may be configured to provide the prior information to the unlearned generator g when training the unlearned generator g.

[0064] [Effects of the generator construction device] According to the generator construction device 2 of this embodiment described above, the generated image output unit 242, the classification result output unit 244, the judgment unit 245, and the learning unit 246 use an input image I2 in which physical quantities have been converted into discrete values ​​and whose shape has been geometrically transformed into a rectangle, and repeat processing (using a generative adversarial network). As with the above generator construction method, this improves the accuracy of the generator G and makes it possible to obtain an estimated image I6 that shows a two-dimensional distribution that is not limited to the horizontal direction but also includes the vertical direction. Furthermore, if the input image I2 shows a hydraulic conductivity distribution, it becomes possible to obtain an estimated image I6 that shows a pressure head distribution that includes not only saturated regions but also unsaturated regions.

[0065] <Fifth embodiment of the invention> Next, an embodiment of the fifth invention (estimation method) will be described.

[0066] [Flow of estimation method] As shown in FIG. 12, the estimation method according to this embodiment includes a third acquisition step S31, an estimation step S32, a third conversion step S33, and a third output step S34.

[0067] (Third acquisition step) In the initial third acquisition step S31, an input image I2 is acquired. The acquisition of this input image I2 can be performed using, for example, a third acquisition unit 341 (see FIG. 14) of an estimation device 3 (sixth invention) described later. The input image I2 may be generated by the image processing device 1 described above, or may be generated by a device other than the image processing device 1 described above. Furthermore, in the third acquisition step S31, a generator G may be acquired. Furthermore, if there is no need to acquire the input image I2 (for example, if the estimation method includes the second conversion step S14 of the image processing method, and the input image I2 generated in the second conversion step S14 is used as is in the estimation step S32 described later), this second acquisition step S21 may not be performed.

[0068] (Estimation step) After acquiring the input image I2, the process proceeds to estimation step S32. In estimation step S32, the input image I2 acquired in the third acquisition step S31 is input to a generator G to generate an estimated image I6. The estimated image I6 can be generated using, for example, an estimation unit 342 (see FIG. 14) of the estimation device 3, which will be described later.

[0069] (Third conversion step) After generating the estimated image I6, the process proceeds to the third conversion step S33. In the third conversion step S33, as shown in FIG. 13, the rectangular estimated image I6 is geometrically converted into an output image I7 of the cross-sectional shape (trapezoid) of the earth structure S. The multiple pixels p arranged within the rectangle are each moved so that they are uniformly distributed within a shape (trapezoid) congruent with the pre-conversion image I1. In other words, this third conversion step S33 performs the reverse of what is performed in the second conversion step of the first invention (image processing method) described above. This geometric conversion can be performed, for example, using the third conversion unit 343 (see FIG. 14) of the estimation device 3 described below.

[0070] (Third output step) After generating the estimated image I6, the process proceeds to a third output step S34 as shown in Fig. 12. In the third output step S34, an output image I7 is output. This output image I7 can be output using, for example, a third output unit 32 and a third output control unit 344 (see Fig. 14) of the estimation device 3, which will be described later.

[0071] [Effects of the estimation method] According to the estimation method of this embodiment described above, by using a generator G constructed using a generative adversarial network, it is possible to obtain an estimated image I6 (a generated image I3 that is not identified as a fake by a classifier D) that shows a two-dimensional distribution that is not limited to the horizontal direction but also includes the vertical direction. Furthermore, when the input image I2 shows a hydraulic conductivity distribution, it is possible to obtain an estimated image I6 that shows a pressure head distribution that includes not only saturated regions but also unsaturated regions.

[0072] <Sixth embodiment of the invention> Next, an embodiment of a sixth aspect of the invention (estimation device) will be described.

[0073] [Configuration of the estimation device] As shown in FIG. 14, the estimation device 3 according to this embodiment includes a third input unit 31, a third output unit 32, a storage unit 33, and a third control unit .

[0074] [Third input section] The third input unit 31 is composed of at least one of a communication module that receives data and signals from other devices (e.g., the image processing device 1, the generator construction device 2, etc.), a terminal that connects to other devices, and a drive that reads information from a recording medium.

[0075] [Third output section] The third output unit 32 is composed of at least one of a communication module that transmits data and signals to other devices (such as a display device that displays the output image I7, a storage device that can store data for the output image I7, etc.), a terminal that connects to other devices, a drive that writes information to a recording medium, and a display that displays images.

[0076] [Storage section] The storage unit 33 stores a generator G. The generator G may be constructed by the generator construction device 2 or may be constructed by another device.

[0077] [Third control section] The third control unit 34 includes a third acquisition unit 341 , an estimation unit 342 , a third conversion unit 343 , and a third output control unit 344 .

[0078] (Third Acquisition Department) The third acquisition unit 341 controls the second input unit 21 to acquire the input image I2. As a result, if the second input unit 21 is a communication module or a terminal, the second input unit 21 receives data for the input image I2 from another device. If the second input unit 21 is a drive, the second input unit 21 reads data for the input image I2 from a recording medium. The third acquisition unit 341 may be configured to acquire the generator G. If there is no need to acquire the input image I2 (for example, if the estimation device 3 includes the second conversion unit 134 of the image processing device 1, and the input image I2 generated by the second conversion unit 134 is used as is by the estimation unit 342 described later), the third control unit 34 may not be provided with the third acquisition unit 341.

[0079] (Estimation Department) The estimation unit 342 generates an estimated image I6 by inputting the input image I2 acquired by the third acquisition unit 341 to a generator G stored in the storage unit 33 or acquired by the third acquisition unit 341. Specifically, the estimation unit 342 executes the same operations as those described in the estimation step S32 (see FIG. 12) in the above estimation method (fifth invention).

[0080] (Third conversion part) The third conversion unit 343 geometrically converts the estimated image I6 into an output image I7 of the cross-sectional shape (trapezoid) of the earth structure S. Specifically, the third conversion unit 343 executes the same operations as those described in the third conversion step S33 (see FIG. 12) in the above estimation method.

[0081] (Third output control unit) The third output control unit 344 controls the third output unit 32 to output the output image I7. As a result, if the third output unit 32 is a communication module or a terminal, the third output unit 32 transmits the data of the output image I7 to another device. If the third output unit 32 is a drive, the third output unit 32 writes the data of the output image I7 to a recording medium. If the third output unit 32 is a display, the third output unit 32 displays the output image I7.

[0082] [Effects of the estimation device] According to the estimation device 3 of this embodiment described above, by using the generator G constructed using a generative adversarial network, it is possible to obtain an estimated image I6 (a generated image I3 that is not identified as a fake by the classifier D) that shows a two-dimensional distribution that is not limited to the horizontal direction but also includes the vertical direction, just like the above estimation method. Furthermore, when the input image I2 shows a hydraulic conductivity distribution, it is possible to obtain an estimated image I6 that shows a pressure head distribution that includes not only saturated regions but also unsaturated regions.

[0083] <First, third, fifth inventions, etc.> The image processing method may include at least any of the steps in the generator construction method and the estimation method. The generator construction method may include at least any of the steps in the image processing method and the estimation method. The estimation method may include at least any of the steps in the image processing method and the generator construction method.

[0084] <Second, fourth, sixth inventions, etc.> Furthermore, the image processing device 1 may include at least one of the control blocks in the generator construction device 2 and the estimation device 3. Furthermore, the generator construction device 2 may include at least one of the control blocks in the image processing device 1 and the estimation device 3. Furthermore, the estimation device 3 may include at least one of the control blocks in the image processing device 1 and the generator construction device 2.

[0085] Furthermore, the functions of at least one of the image processing device 1, generator construction device 2, and estimation device 3 can be realized by a program (image processing program, generator construction program, estimation program) for causing a computer to function as the device, and a program for causing a computer to function as each control block of the device (particularly each part included in at least one of the first control unit 13, the second control unit 24, and the third control unit 34). In this case, the device includes a computer having at least one control device (e.g., a processor) and at least one storage device (e.g., a memory) as hardware for executing the program. The control device and storage device execute the information processing program, thereby realizing the functions described in each of the above embodiments. The program may be non-transitory and may be recorded on one or more computer-readable recording media. The recording media may or may not be included in the device. In the latter case, the program may be supplied to the device via any wired or wireless transmission medium.

[0086] Furthermore, some or all of the functions of the control blocks can be realized by logic circuits. For example, an integrated circuit in which a logic circuit that functions as each of the control blocks is formed is also included in the scope of the present invention. In addition, the functions of the control blocks can also be realized by, for example, a quantum computer.

[0087] The present invention is not limited to the above-described embodiments, and various modifications are possible within the scope of the claims. Embodiments obtained by appropriately combining the technical means disclosed in different embodiments are also included in the technical scope of the present invention. [Example]

[0088] Next, examples of the first to sixth aspects of the invention will be described.

[0089] First, by the image processing method or using the image processing device 1, 15 3 An input image I2 showing the distribution of hydraulic conductivity was generated. Next, the above 15 3 The input images I2 and the 15 3 A set of correct images I5 was used as training data, and a generator G was constructed by the above generator construction method or using the above generator construction device 2. Next, the input image I2 showing the hydraulic conductivity distribution was input to the constructed generator G, and an estimated image I6 showing the pressure head distribution was generated, which was then geometrically transformed into the output image I7 (forward). Next, the SSIM (similarity to the pixel value of the correct image I5) of each pixel p of the output image I7 was calculated.

[0090] Furthermore, by the image processing method or using the image processing device 1, 15 3 An input image I2 showing the pressure head distribution was generated. Next, the above 15 3 The training input images I2 and the 15 3 A set of correct images I5 was used as training data, and a generator G was constructed by the above generator construction method or using the above generator construction device 2. Next, 343 verification images (input images for prediction I2) showing the pressure head distribution were input to the constructed generator G, and an estimated image I6 showing the hydraulic conductivity distribution was generated, which was then geometrically transformed into an output image I7 (inversion). Next, the SSIM of each output image I7 was calculated.

[0091] Next, using the calculated SSIM, a histogram (overlapping forward and inversion images) was created, as shown in FIG. 15, with SSIM on the horizontal axis and frequency (number of images) on the vertical axis. From the histogram, it can be seen that the SSIM of most output images I7 (forward) is 0.98 or higher. This indicates that the output image I7 (forward) according to this embodiment has a level of accuracy (similarity to the correct image I5) that makes it indistinguishable from the correct image I5 using classifier D of the generative adversarial network. On the other hand, it can be seen that the SSIM of many output images I7 (inversion) is less than 0.98.

[0092] Next, the above 15 3 Input image I2 showing the pressure head distribution of 15 3 Using the set of correct images I5 as training data, another generator G was constructed by the above generator construction method or using the above generator construction device 2. At this time, prior information obtained in advance (the hydraulic conductivity of the downstream portion of the dam body (third portion P3) corresponding to each input image I2) was provided to the untrained generator g. Next, the input image I2 showing the pressure head distribution was input to another generator G, and an estimated image I6 showing the hydraulic conductivity distribution was generated, which was then geometrically transformed into the output image I7 (inversion). Next, the error of the output image I7 relative to the correct image I5 was calculated.

[0093] Next, using the calculated errors, a histogram (with and without prior information at the time of learning superimposed) was created, with the error on the horizontal axis and the frequency (number of images) on the vertical axis, as shown in Figure 16. From the histogram, it was found that the error of most of the output images I7 (with prior information) was 10 1It can be seen that the following is true: This shows that by using prior information when constructing the generator G, the accuracy of the output image I7 (inversion) according to this embodiment is greatly improved. [Explanation of symbols]

[0094] 1. Image processing device 11 First input section 12 First output section 13 First Control Section 131 First Acquisition Department 132 First Conversion Unit 133 Generation part 134 Second Conversion Unit 135 First output control section 2 Generator construction device 21 Second input section 22 Second output section 23 Memory section 24 Second Control Section 241 Second Acquisition Department 242 Generated image output unit 243 Analysis Department 244 Classification result output unit 245 Judgment Department 246 Learning Department 247 Second Output Control Unit 3 Estimation device 31 Third input section 32 Third output section 33 Storage section 34 Third Control Section 341 Third Acquisition Department 342 Estimation Department 343 Third Conversion Unit 344 Third Output Control Unit D Discriminator d Untrained classifier G generator g Untrained generator I1 Image before conversion R1 first site region R2: Region showing the second site R3: Region showing the third site I2 Input images I3 generated image I4 analysis image before conversion I5 Correct Image I6 Estimated Image I7 output image L Line indicating the invasion surface S earth structure P1 first part P2 second part P3 third part p pixels

Claims

1. A step of converting physical quantities relating to an earth structure to be analyzed, which has a trapezoidal cross section with its upper base shorter than its lower base, into discrete values ​​according to the characteristics of the physical quantities; generating a pre-transformation image having the same shape as a cross section of the earth structure, the pre-transformation image showing the distribution of the discrete values ​​within the earth structure; geometrically transforming the pre-transformed image into a rectangular input image; An image processing method comprising:

2. The physical quantity is a hydraulic conductivity or a pressure head. The image processing method according to claim 1 .

3. When the physical quantity is a hydraulic conductivity, In the step of converting into the discrete value, the permeability coefficient is converted into a gradation such that the lightness or saturation changes by one step every time the coefficient is multiplied by 10. The image processing method according to claim 2 .

4. When the physical quantity is a pressure head, In the step of converting into discrete values, As the pressure head value increases in the positive direction, the saturation of one hue increases. As the pressure head value increases in the negative direction, the saturation of other hues increases. The image processing method according to claim 2 .

5. The cross-sectional shape of the earth structure is trapezoidal, The step of generating a pre-transformed image includes: Dividing the pre-transformed image into a plurality of small regions; extracting pixels one by one from each of the micro-regions; and In the step of geometrically transforming, the plurality of pixels arranged in the pre-transformed image are respectively moved so as to be uniformly distributed within a rectangle; In the step of dividing the image into minute regions, the image before transformation is divided so that the upper minute region of the image before transformation is smaller than the lower minute region. The image processing method according to claim 1 .

6. When the soil structure has two or more portions with different permeability coefficients and the physical quantity is the permeability coefficient, In the step of geometrically transforming, the pre-transformed image is geometrically transformed so that an area of ​​a region showing a relatively high hydraulic conductivity in the input image is larger than an area of ​​a region showing a relatively high hydraulic conductivity in the pre-transformed image. The image processing method according to claim 1 .

7. A first conversion unit that converts physical quantities related to an earth structure to be analyzed, which has a trapezoidal cross section with its upper base shorter than its lower base, into discrete values ​​according to the characteristics of the physical quantities; a generation unit that generates a pre-conversion image having the same shape as a cross section of the earth structure, which image shows the distribution of the discrete values ​​within the earth structure; a second conversion unit that geometrically converts the pre-conversion image into a rectangular input image; An image processing device comprising:

8. 8. An image processing program for causing a computer to function as the image processing device according to claim 7, the image processing program causing a computer to function as the first conversion unit, the generation unit, and the second conversion unit.

Images