Information processing system, information processing method and information processing program

Abstract

To correct various position deviations at relatively high accuracy in technology correcting a position deviation between two pieces of satellite images.SOLUTION: A reasoning device 200 comprises a reasoning unit 220 that inputs, to a learned model, a first image serving as a satellite image, and a second image serving as the satellite image which corresponds to the first image, and has a position deviation to the first image occurring, and thereby reasons a displacement map to be used in correcting the position deviation of the first image to the second image. The learned model is a model that is based on machine learning, inputs two pieces of satellite images in which the position deviation occurs each other, and outputs the displacement map to be used in correcting a position deviation of other with respect to one of the two pieces of satellite images.SELECTED DRAWING: Figure 3

Description

[Technical field]

[0001] The present disclosure relates to an information processing system, an information processing method, and an information processing program. [Background technology]

[0002] In recent years, analytical methods using remote sensing data have been proposed as a way to grasp the damage caused by disasters over a wide area. Examples of remote sensing include aircraft and UAVs (drones), as well as optical satellites and synthetic aperture radar satellites (SAR satellites). By overlaying images taken before and after a disaster, it is possible to extract areas where significant changes have occurred. Of these, optical satellite images are used as a means of monitoring various objects on the earth's surface. In addition, SAR satellites emit microwaves onto the earth's surface as the flying object moves, receive echoes reflected from the surface, and generate SAR images by signal processing the received echoes. Radar images obtained by SAR have amplitude and phase in each pixel. SAR is an abbreviation for Synthetic Aperture Radar. In SAR, radar images equivalent to those obtained by using an antenna with a large aperture are obtained by artificially synthesizing an aperture.

[0003] There is a process for estimating pixel offsets for two images captured by remote sensing in order to accurately align the two images, and then converting the images based on the estimation results to eliminate the pixel position shift between the two images. Non-Patent Document 1 discloses a technique for optimizing parameters of a polynomial that performs deformation, deforming an original image using the optimized polynomial, and overlaying the two images to output an image in which the positional deviation between the two images has been corrected. When optimizing the parameters, the parameters of the mapping function are optimized so that the difference between the coordinate difference calculated from the two images using an area correlation method and the coordinate difference calculated using a mapping function represented by the affine transform is minimized. Image 1 is deformed using the optimized parameters, and the deformed image 1 is overlaid on image 2 to eliminate the pixel positional deviation between the two images. [Prior art documents] [Non-patent literature]

[0004] [Non-Patent Document 1] Tobita, Rosen, "High Precision Image Matching for Interferometric SAR", Journal of Geodetic Society, Vol. 45, No. 4, 1999, pp. 297-314 Summary of the Invention [Problem to be solved by the invention]

[0005] The technology disclosed in Non-Patent Document 1 has a problem that it cannot handle complex deformations that cannot be expressed by a first-order affine transformation function because it employs a first-order affine transformation function. In addition, the technology has a problem that the accuracy of the mapping function varies depending on the selection of corresponding points, and the variation does not occur uniformly within the image, resulting in some areas where the positional shift is small and other areas where there is little change. The present disclosure aims to provide a technique for correcting positional deviations between two satellite images, which allows for correction of various types of positional deviations with relatively high accuracy. [Means for solving the problem]

[0006] The information processing system according to the present disclosure includes: An inference unit that infers a displacement map used to correct the displacement of one of the two satellite images relative to the second image by inputting a first image, which is a satellite image, and a second image, which is a satellite image corresponding to the first image and is displaced from the first image, to a trained model that is a model based on machine learning and that inputs two satellite images that are displaced from each other and outputs a displacement map used to correct the displacement of the other of the two satellite images relative to the first image. An inference device comprising: Equipped with. Effect of the Invention

[0007] According to the present disclosure, a disparity map is inferred using a trained model based on machine learning. Here, the inferred disparity map is used to correct misalignment between two satellite images. Thus, according to the present disclosure, in a technique for correcting misalignment between two satellite images, various misalignments can be corrected with relatively high accuracy. [Brief description of the drawings]

[0008] [Figure 1] 1 is a diagram illustrating an overview of an information processing system 90 according to a first embodiment. [Diagram 2] FIG. 1 shows an example of the configuration of a learning device 100 according to a first embodiment. [Diagram 3] FIG. 2 shows an example of the configuration of an inference device 200 according to the first embodiment. [Figure 4] FIG. 2 is a diagram showing an example of the hardware configuration of the learning device 100 according to the first embodiment. [Diagram 5] FIG. 2 is a diagram for explaining the operation of the learning device 100 according to the first embodiment. [Figure 6] 1A and 1B are diagrams illustrating a displacement map, where FIG. 1A shows a specific example of a displacement map, and FIG. 1B shows a specific example of image deformation. [Figure 7] FIG. 2 is a diagram for explaining the operation of the inference device 200 according to the first embodiment. [Figure 8] FIG. 1 is a diagram for explaining a problem with the conventional technology. [Figure 9] FIG. 13 shows an example of the configuration of a learning device 100 according to a modification of the first embodiment. [Figure 10] FIG. 13 shows an example of the configuration of a learning device 100 according to a modification of the first embodiment. [Figure 11] FIG. 13 shows an example of the configuration of an inference device 200 according to a modification of the first embodiment. [Figure 12] FIG. 13 is a diagram showing an example of a hardware configuration of a learning device 100 according to a modification of the first embodiment. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0009] In the description of the embodiments and the drawings, the same elements and corresponding elements are given the same reference numerals. Descriptions of elements given the same reference numerals are omitted or simplified as appropriate. Arrows in the drawings primarily indicate data flow or processing flow. In addition, "part" may be read as "circuit," "step," "procedure," "processing," or "circuitry" as appropriate.

[0010] Embodiment 1 Hereinafter, the present embodiment will be described in detail with reference to the drawings.

[0011] ***Configuration Description*** FIG. 1 is a diagram for explaining an overview of an information processing system 90. The information processing system 90 is a system for carrying out AI (Artificial Intelligence) learning and AI utilization, and includes a learning device 100 and an inference device 200. The information processing system 90 corresponds to a framework for estimating a displacement map capable of expressing any deformation for correcting the positional deviation between two satellite images. The learning device 100 and the inference device 200 may be configured integrally. The satellite image is, for example, a Synthetic Aperture Radar (SAR) image or an optical satellite image. The learning device 100 executes a process in a learning phase. In the learning phase, a learned model is generated by supervised learning. The inference device 200 executes processing in the utilization phase. In the utilization phase, by utilizing the trained model, the displacement is estimated from two different images, and the positional deviation of one of the images is corrected.

[0012] An overview of the processing of the information processing system 90 will now be described. First, the learning device 100 acquires an image. In this specification, an image is basically data showing a satellite image. The satellite image may be an image actually taken by an artificial satellite, or may be an image generated by a simulator of radar image analysis. Next, the learning device 100 performs preprocessing on the image, and sets a pair of the image on which the preprocessing has been performed and an image corresponding to the image on which the preprocessing has been performed as learning data. The preprocessing is a process of deforming an image using a displacement map. As a specific example, the image corresponding to the image on which the preprocessing has been performed is an image that is different from both the image before being deformed by the displacement map, or the image before and after the preprocessing, and is an image in which a positional shift occurs between the image on which the preprocessing has been performed. The displacement map may be a displacement map that is generated randomly, or may be a displacement map that is generated according to a rule. Next, the learning device 100 generates a trained model by executing a learning program using the learning data. The trained model is a model based on machine learning, which receives two satellite images with a positional shift from each other as input and outputs a displacement map used to correct the positional shift of one of the two satellite images relative to the other. As a specific example, the two input images are images taken at different times. The two images may be images taken by the same satellite or may be images taken by different satellites. Next, inference apparatus 200 inputs the two images to the trained model to infer a displacement map that is used to correct the misalignment of the two images.

[0013] Fig. 2 shows an example of the configuration of a learning device 100 according to this embodiment. As shown in Fig. 2, the learning device 100 includes a displacement map generating unit 110, an image deformation unit 120, and a model generating unit 130. An image 1 is an input image to the learning device 100.

[0014] The disparity map generator 110 generates a disparity map that corresponds to the image 1 and is used to deform the image 1. The image 1 corresponds to an input image. The input image is a satellite image input to the learning device 100. The disparity map generator 110 may generate the disparity map using a polynomial transformation.

[0015] The image deformation unit 120 generates a deformed image by deforming the image 1 using the displacement map generated by the displacement map generation unit 110. That is, the image deformation unit 120 generates the other of the two satellite images using the image 1 and the generated displacement map. Note that the image 1 does not refer to a specific image, and typically there are multiple images 1. The image deformation unit 120 generates a deformed image corresponding to each image 1. The image deformation unit 120 may also generate a deformed image using a different displacement map for each image 1.

[0016] The model generating unit 130 generates a trained model by learning the correspondence between two satellite images and a displacement map corresponding to the positional deviation of one of the two satellite images relative to the other. One of the two satellite images may be an input image. Specifically, the model generation unit 130 generates a trained model using a pair of image 1 and the deformed image, and the displacement map generated by the displacement map generation unit 110, and sends the generated trained model to the trained model storage unit 300. Note that another image in which a positional shift occurs between the deformed image and image 1 may be used instead of image 1. The trained model may be a model that outputs a displacement map used to correct either image 1 or the deformed image. The trained model storage unit 300 is a component of the information processing system 90, and stores the trained model.

[0017] 3 shows an example of the configuration of inference device 200 according to this embodiment. As shown in FIG. 3, inference device 200 includes data acquisition section 210, inference section 220, image transformation section 230, and image synthesis section 240.

[0018] The data acquisition unit 210 acquires two images. Hereinafter, one of the two images will be referred to as image 1, and the other as image 2. It is assumed that a positional shift occurs between image 1 and image 2. Image 1 acquired by the data acquisition unit 210 is also called the first image. Image 2 acquired by the data acquisition unit 210 is also called the second image. The first image is a satellite image. The second image is a satellite image corresponding to the first image, and is an image that has a positional shift with respect to the first image.

[0019] The inference unit 220 acquires a trained model from the trained model storage unit 300, and inputs image 1 and image 2 to the acquired trained model to infer a displacement map corresponding to the positional deviation of image 1 relative to image 2. The displacement map is used to correct the positional deviation of the first image relative to the second image.

[0020] The image deformation unit 230 generates a corrected image by correcting the positional shift of the image 1 based on the displacement map inferred by the inference unit 220 .

[0021] The image synthesis unit 240 overlaps the corrected image with Image 2 to eliminate the misalignment between the two images.

[0022] 4 shows an example of the hardware configuration of learning device 100 according to this embodiment. Learning device 100 is composed of a computer. Learning device 100 may be composed of multiple computers.

[0023] As shown in the figure, the learning device 100 is a computer equipped with hardware such as a processor 11, a memory 12, an auxiliary storage device 13, an input / output IF (Interface) 14, and a communication device 15. These pieces of hardware are appropriately connected via signal lines 19.

[0024] The processor 11 is an integrated circuit (IC) that performs arithmetic processing and controls the hardware of the computer. Specific examples of the processor 11 include a central processing unit (CPU), a digital signal processor (DSP), and a graphics processing unit (GPU). The learning device 100 may include a plurality of processors that replace the processor 11. The plurality of processors share the role of the processor 11.

[0025] The memory 12 is typically a volatile storage device, and a specific example is a RAM (Random Access Memory). The memory 12 is also called a primary storage device or a main memory. Data stored in the memory 12 is saved in the auxiliary storage device 13 as necessary.

[0026] The auxiliary storage device 13 is typically a non-volatile storage device, and specific examples thereof include a ROM (Read Only Memory), an HDD (Hard Disk Drive), or a flash memory. Data stored in the auxiliary storage device 13 is loaded into the memory 12 as necessary. The memory 12 and the auxiliary storage device 13 may be integrated into one unit.

[0027] The input / output IF 14 is a port to which an input device and an output device are connected. A specific example of the input / output IF 14 is a Universal Serial Bus (USB) terminal. Specific examples of the input device are a keyboard and a mouse. A specific example of the output device is a display.

[0028] The communication device 15 is a receiver and a transmitter, and is, for example, a communication chip or a network interface card (NIC).

[0029] Each part of the learning device 100 may use the input / output IF 14 and the communication device 15 as appropriate when communicating with other devices.

[0030] Auxiliary storage device 13 stores an information processing program. The information processing program is a program that causes a computer to realize the functions of each unit of learning device 100. The information processing program is loaded into memory 12 and executed by processor 11. The functions of each unit of learning device 100 are realized by software.

[0031] Data used when executing the information processing program and data obtained by executing the information processing program are appropriately stored in a storage device. Each part of the learning device 100 uses a storage device as appropriate. As a specific example, the storage device is composed of at least one of the memory 12, the auxiliary storage device 13, a register in the processor 11, and a cache memory in the processor 11. Note that the terms "data" and "information" may have the same meaning. The storage device may be independent of the computer. The functions of the memory 12 and the auxiliary storage device 13 may be realized by other storage devices.

[0032] The information processing program may be recorded in a computer-readable non-volatile recording medium. Specific examples of the non-volatile recording medium include an optical disk and a flash memory. The information processing program may be provided as a program product. The hardware configuration of inference device 200 may be similar to the hardware configuration of learning device 100. The trained model storage unit 300 may be any type of storage device.

[0033] ***Explanation of Operation*** The operation procedure of each device included in the information processing system 90 corresponds to an information processing method. Also, the program that realizes the operation of each device included in the information processing system 90 corresponds to an information processing program.

[0034] 5 is a diagram showing an example of the operation of learning device 100. The operation of learning device 100 will be described with reference to FIG.

[0035] (Step S101) The disparity map generator 110 generates a disparity map based on random numbers. The displacement map is a displacement map showing 10 pixels above as an example.

[0036] A specific example of the method for generating the displacement map is a method for selecting several pixels in the original image, generating random numbers for each selected pixel so that the amount of displacement falls within ±several pixels, and performing polynomial transformation based on the generated random numbers. Specific examples of the polynomial transformation include transformation using a thin plate spline, affine transformation, or parallel translation. By imposing restrictions when generating the displacement map in this way, a natural displacement map with a realistic amount of displacement can be generated, and the inference accuracy is expected to improve. Note that a displacement map corresponding to a case where a relatively large positional deviation occurs, such as before and after a landslide or before and after a volcanic eruption, may be generated. Another possible method for generating a displacement map is to estimate the magnitude of the displacement between two images using a 3D (3-dimensional) model and satellite information, and generate a displacement map based on the estimated magnitude of the displacement. This method has the advantage of not requiring a large number of satellite images to be prepared as learning data.

[0037] (Step S102) The image deformation unit 120 generates a deformed image by deforming the image 1 using the displacement map generated in step S101.

[0038] Fig. 6 is a diagram explaining a displacement map. A displacement map is a map in which, for each pixel of an image, the direction of displacement is expressed by hue and the amount of displacement is expressed by saturation. Note that, although the displacement map is expressed in black and white in Fig. 6(a), in reality, the displacement map is expressed in a rich variety of colors. FIG. 6(a) shows a displacement map in which pixel 1 is displaced 3 pixels to the right and pixel 2 is moved 4 pixels up and to the left. Figure 6(b) shows how an image is deformed using the displacement map shown in Figure 6(a). The original image is deformed for each coordinate based on the displacement map, and the deformed image is generated by shaping it through interpolation.

[0039] (Step S103) The model generation unit 130 learns the learned model by learning the correspondence between the pair of the deformed image and image 2 generated in step S102, or the deformed image and image 1, and the displacement map generated in step S101. The input displacement map corresponds to the correct answer.

[0040] 7 is a diagram showing an example of the operation of inference device 200. The operation of inference device 200 will be described with reference to FIG.

[0041] (Step S201) The inference unit 220 acquires a learned model from the learned model memory unit 300, and infers a displacement map by inputting image 1 and image 2 acquired by the data acquisition unit 210 to the acquired learned model. As a specific example, image 2 is an image that is shifted 5 pixels to the right and 5 pixels up relative to image 1.

[0042] (Step S202) The image deformation unit 230 generates a corrected image by deforming the image 1 using the displacement map inferred in step S201.

[0043] (Step S203) The image synthesis unit 240 synthesizes the corrected image generated in step S202 with Image 2.

[0044] ***Explanation of the Effects of the First Embodiment*** Fig. 8 is a diagram for explaining the conventional method. In the conventional method, the parameters of a polynomial are optimized based on the amount of deviation between several corresponding points, and the image is corrected by a polynomial using the optimized parameters. Therefore, the conventional method has a problem that when the displacement is discontinuous as shown in the lower right of Fig. 8, the polynomial cannot express such deformation and therefore cannot perform appropriate correction. On the other hand, according to the present embodiment, the correction is performed based on a displacement map generated by machine learning. Here, the displacement map can indicate the amount of displacement in pixel units. Therefore, according to the present embodiment, even if the displacement is discontinuous, the correction can be performed appropriately.

[0045] According to this embodiment, the displacement amount at each coordinate can be adjusted by using a displacement map. Therefore, according to this embodiment, an image can be generated in which the amount of positional displacement for each pixel in the image is small. Also, compared to the conventional method in which adjustment is performed by one polynomial for one image, this embodiment can handle complex deformations that cannot be expressed by a polynomial. Furthermore, when images before and after deformation are input to the learning device 100, there is no need to prepare image pairs corresponding to two time periods. Therefore, in this case, there is an effect that the amount of data set that can be secured increases. Furthermore, when the input to the learning device 100 is an image other than the image before transformation and an image after transformation, it is possible to perform inference that takes into account changes or disturbances in two time periods corresponding to the two input images. Therefore, in this case, it is possible to perform inference with relatively high accuracy.

[0046] ***Other configurations*** <Variation 1> By using the altitude information, the learning device 100 is expected to generate a more natural displacement map, and the inference accuracy is expected to improve. A specific example of the altitude information is a DEM (Digital Elevation Model). This modification makes it possible to deal with position deviations caused by altitude differences on the ground surface. There are two possible methods for using altitude information: Method 1: When generating a displacement map, altitude information is used as input to adjust the amount of displacement. Method 2: Altitude information is also input when training the model.

[0047] FIG. 9 shows an example of the configuration of a learning device 100 that realizes the method 1. The displacement map generating unit 110 generates a displacement map based on the altitude information. Specifically, the displacement map generating unit 110 generates a displacement map using altitude information indicating the altitude of the shooting range of the image 1.

[0048] FIG. 10 shows an example of the configuration of a learning device 100 that realizes the second method. The disparity map generator 110 generates the disparity map in the same manner as in method 1. The model generation unit 130 also receives altitude information during model learning. That is, the model generation unit 130 generates a learned model using altitude information indicating the altitude of the shooting range of the image 1. Regardless of the configuration of learning device 100, altitude information must be provided as input to inference device 200 at the time of inference, along with the two images.

[0049] FIG. 11 shows an example of the configuration of an inference device 200 according to this modification. The inference unit 220 inputs altitude information to the trained model when inferring the displacement map. That is, the inference unit 220 inputs information indicating the altitude of the shooting range of the first image to the trained model when inferring the displacement map corresponding to the first image.

[0050] <Variation 2> FIG. 12 shows an example of the hardware configuration of the learning device 100 according to this modified example. The learning device 100 includes a processing circuit 18 in place of the processor 11 , the processor 11 and memory 12 , the processor 11 and auxiliary storage device 13 , or the processor 11 , memory 12 , and auxiliary storage device 13 . The processing circuitry 18 is hardware that realizes at least a portion of each unit of the learning device 100. The processing circuitry 18 may be dedicated hardware, or may be a processor that executes programs stored in the memory 12 .

[0051] When processing circuitry 18 is dedicated hardware, processing circuitry 18 may be, for example, a single circuit, a composite circuit, a programmed processor, a parallel programmed processor, an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array), or a combination thereof. The learning device 100 may include a plurality of processing circuits that replace the processing circuit 18. The plurality of processing circuits share the role of the processing circuit 18.

[0052] In the learning device 100, some functions may be realized by dedicated hardware, and the remaining functions may be realized by software or firmware.

[0053] Processing circuitry 18 is illustratively implemented in hardware, software, firmware, or a combination thereof. The processor 11, memory 12, auxiliary storage device 13, and processing circuit 18 are collectively referred to as the “processing circuitry.” In other words, the functions of each of the functional components of the learning device 100 are realized by the processing circuitry. The hardware configuration of inference device 200 may also be similar to that of this modified example.

[0054] ***Other embodiments*** Although the first embodiment has been described, it is possible to combine a plurality of parts of the present embodiment. Alternatively, it is possible to partially implement the present embodiment. In addition, the present embodiment may be modified in various ways as necessary, and may be implemented in any combination, either as a whole or in part. The above-described embodiment is essentially a preferred example, and is not intended to limit the scope of the present disclosure, its applications, and uses. The described procedure may be modified as appropriate.

[0055] Various aspects of the present disclosure are summarized below as appendices.

[0056] (Appendix 1) An inference unit that infers a displacement map used to correct the displacement of one of the two satellite images relative to the second image by inputting a first image, which is a satellite image, and a second image, which is a satellite image corresponding to the first image and is displaced from the first image, to a trained model that is a model based on machine learning and that inputs two satellite images that are displaced from each other and outputs a displacement map used to correct the displacement of the other of the two satellite images relative to the first image. An inference device comprising: An information processing system comprising:

[0057] (Appendix 2) The inference device further comprises: an image transformation unit that generates a corrected image by correcting the positional shift of the first image based on the inferred displacement map; an image synthesis unit that synthesizes the corrected image and the second image; 2. An information processing system according to claim 1, comprising:

[0058] (Appendix 3) The information processing system further comprises: A learning device including a model generation unit that generates the trained model by learning a correspondence relationship between the two satellite images and a displacement map corresponding to a positional deviation of one of the two satellite images relative to the other. 3. The information processing system according to claim 1 or 2, comprising:

[0059] (Appendix 4) One of the two satellite images is an input image that is a satellite image input to the learning device, The learning device further comprises: a displacement map generator for generating a displacement map corresponding to the input image; an image transformation unit that uses the input image and the generated displacement map to generate the other of the two satellite images; 4. The information processing system according to claim 3, comprising:

[0060] (Appendix 5) 5. The information processing system according to claim 4, wherein the displacement map generation unit generates the displacement map using a polynomial transformation.

[0061] (Appendix 6) 6. The information processing system according to claim 4, wherein the displacement map generation unit generates the displacement map using altitude information indicating an altitude of a shooting range of the input image.

[0062] (Appendix 7) 7. The information processing system according to claim 3, wherein the model generation unit generates the trained model using the elevation information.

[0063] (Appendix 8) The information processing system according to claim 6 or 7, wherein the inference unit inputs information indicating an altitude of a shooting range of the first image to the trained model when inferring a displacement map corresponding to the first image. [Explanation of symbols]

[0064] 11 processor, 12 memory, 13 auxiliary storage device, 14 input / output IF, 15 communication device, 18 processing circuit, 19 signal line, 90 information processing system, 100 learning device, 110 displacement map generation unit, 120 image deformation unit, 130 model generation unit, 200 inference device, 210 data acquisition unit, 220 inference unit, 230 image deformation unit, 240 image synthesis unit, 300 learned model storage unit.

Claims

1. An inference unit that infers a displacement map used to correct the displacement of the first image relative to the second image by inputting a first image, which is a satellite image, and a second image, which is a satellite image corresponding to the first image and is displaced from the first image, to a trained model that is a model based on machine learning and that inputs two satellite images that are displaced from each other and outputs a displacement map used to correct the displacement of one of the two satellite images relative to the other. An inference device comprising: An information processing system comprising:

2. The inference device further comprises: an image transformation unit that generates a corrected image by correcting the positional deviation of the first image based on the inferred displacement map; an image synthesis unit that synthesizes the corrected image and the second image; The information processing system according to claim 1 .

3. The information processing system further comprises: A model generation unit that generates the trained model by learning a correspondence relationship between the two satellite images and a displacement map corresponding to a positional deviation of one of the two satellite images relative to the other. A learning device comprising: The information processing system according to claim 1 or 2, comprising:

4. One of the two satellite images is an input image that is a satellite image input to the learning device, The learning device further comprises: a displacement map generator for generating a displacement map corresponding to the input image; an image transformation unit that uses the input image and the generated displacement map to generate the other of the two satellite images; The information processing system according to claim 3 .

5. The information processing system according to claim 4 , wherein the displacement map generator generates the displacement map using a polynomial transformation.

6. The information processing system according to claim 4 , wherein the displacement map generating unit generates the displacement map by using altitude information indicating an altitude of a shooting range of the input image.

7. The information processing system according to claim 6 , wherein the model generation unit generates the trained model using the altitude information.

8. The information processing system according to claim 6 , wherein the inference unit inputs information indicating an altitude of a shooting range of the first image to the trained model when inferring a displacement map corresponding to the first image.

9. An information processing method in which a computer inputs a first image, which is a satellite image, and a second image, which is a satellite image corresponding to the first image and is misaligned with respect to the first image, into a trained model, which is a model based on machine learning and inputs two satellite images that are misaligned with each other, and outputs a displacement map used to correct the positional misalignment of one of the two satellite images relative to the other.The computer infers a displacement map used to correct the positional misalignment of the first image relative to the second image by inputting a first image, which is a satellite image, and a second image, which is a satellite image corresponding to the first image and is misaligned with respect to the first image.

10. An inference process for inferring a displacement map used to correct the displacement of the first image relative to the second image by inputting a first image, which is a satellite image, and a second image, which is a satellite image corresponding to the first image and is displaced from the first image, to a trained model that is a model based on machine learning and that inputs two satellite images that are displaced from each other and outputs a displacement map used to correct the displacement of one of the two satellite images relative to the other. An information processing program that causes an inference device, which is a computer, to execute the above.

Images