Detection device
The detection device uses a GAN to generate false scattering data for comparing with actual data to detect underground anomalies, addressing the limitation of existing SAR technologies in identifying ground subsidence causes.
Patent Information
- Application Number
- PCT/JP2025/016924
- Authority / Receiving Office
- WO · WO
- Patent Type
- Applications
- Current Assignee / Owner
- Filing Date
- 2025-05-08
- Publication Date
- 2026-11-12
AI Technical Summary
Existing technologies, such as those using synthetic aperture radar (SAR) from satellites, can measure ground displacement but fail to detect underground abnormalities like voids or ground loosening that cause subsidence.
A detection device utilizing a Generative Adversarial Network (GAN) to generate false scattering data representing normal ground conditions, which is compared with actual scattering data to identify differences and detect underground anomalies.
Enables accurate detection of underground abnormalities by generating false scattering data to highlight deviations from normal conditions, thereby identifying cavities or ground loosening.
Smart Images

Figure JP2025016924_12112026_PF_FP_ABST
Abstract
Description
Detection device
[0001] This disclosure relates to a detection device.
[0002] Underground spaces house various infrastructure facilities such as sewers, telecommunications, power lines, and railway tunnels, which support daily life and economic activities. However, these infrastructure facilities are now several decades old and are deteriorating. As a result, for example, holes have formed in aging sewer pipes, allowing soil and sediment to flow into the pipes, creating cavities underground and leading to ground subsidence.
[0003] Non-patent document 1 describes a technique for measuring the displacement of ground structures using measurement data from a synthetic aperture radar (SAR) mounted on a satellite. This technique allows for the measurement of the ground elevation of the observed object with an accuracy of several centimeters.
[0004] Yamanaka et al., "Detection of Ground Subsidence by Interferometric SAE Time Series Analysis," Bulletin of the Geospatial Information Authority of Japan, No. 124, 2013.
[0005] To prevent the aforementioned subsidence accidents, there is a need for technology to detect underground abnormalities that cause subsidence, such as voids in the ground or loosening of the ground. The technology described in Non-Patent Literature 1 can measure the amount of displacement after ground movement, but it cannot detect underground abnormalities.
[0006] In light of the problems described above, the purpose of this disclosure is to provide a detection device that can detect abnormalities in the ground.
[0007] A detection device according to one embodiment is a detection device for detecting anomalies underground in a target area, comprising: a generation unit that generates false scattering data indicating that there are no anomalies underground in the target area from actual scattering data showing at least one of surface scattering, corner scattering, and volume scattering obtained from measurement data of a synthetic aperture radar that radiates electromagnetic waves onto the ground of the target area and measures the reflected waves reflected by the electromagnetic waves from the ground; and an identification unit that detects anomalies underground in the target area based on the difference between the actual scattering data and the false scattering data.
[0008] According to this disclosure, it is possible to detect abnormalities in the ground.
[0009] This figure shows an example of the configuration of a detection device according to one embodiment of this disclosure. This figure is for explaining the learning of the generator and discriminator shown in Figure 1. This is a flowchart showing an example of the operation of the detection device shown in Figure 1.
[0010] Embodiments of this disclosure will be described below with reference to the drawings.
[0011] Figure 1 shows an example of the configuration of a detection device 10 according to one embodiment of the present disclosure. The detection device 10 according to the present disclosure uses measurement data of the target area by synthetic aperture radar (SAR) to detect underground abnormalities in the target area that cause ground subsidence, such as voids in the ground or loosening of the ground. SAR is a satellite system mounted on a satellite or the like that emits electromagnetic waves to the ground surface and measures the reflected waves that are reflected from the ground surface, thereby enabling the understanding of underground conditions.
[0012] As shown in Figure 1, the detection device 10 according to this embodiment includes an input unit 11, an output unit 12, a storage unit 13, and a control unit 14.
[0013] The input unit 11 includes at least one input interface. The input interface is, for example, a physical key, a capacitive key, a pointing device, a touchscreen integrated with a display, or a microphone. The input unit 11 accepts, for example, operations to input data used for the operation of the detection device 10. The input unit 11 receives, for example, scattering data. Scattering data is data indicating at least one of surface scattering, corner scattering, and volume scattering, obtained from SAR measurement data obtained by radiating electromagnetic waves onto the ground in a target area and measuring the reflected waves reflected by the ground. The scattering data input to the input unit 11 is data obtained from actual SAR measurement data and will be referred to as "actual scattering data" below.
[0014] Some satellites equipped with SAR (Surface Array) can arbitrarily set the direction of input and output electromagnetic waves. If the vertical and horizontal components of the emitted electromagnetic waves are V1 and H1, respectively, and the vertical and horizontal components of the reflected waves are V2 and H2, respectively, then the SAR can acquire four patterns of spatial data (images) as measurement data: polarization V1V2, polarization V1H2, polarization V2H1, and polarization H1H2. By performing polarization decomposition, such as Pauli decomposition, on this measurement data, images showing surface scattering (= polarization V1V2 + polarization H1H2), corner scattering (= polarization H1H1 - polarization V1V2), and volume scattering (= polarization H1V2 + polarization V1H2) can be obtained. The obtained images are then subjected to noise reduction by multi-look processing, alignment of images by correlation processing, and addition of positional information and distortion correction by Geocoding and Radiometric Calibration, and these images are input as scattering data. In addition to Pauli decomposition, other polarization decomposition methods such as Cloude-Pottier Decomposition, Freeman-Durden Decomposition, and Yamaguchi Decomposition may also be used.
[0015] Surface scattering is the scattering of electromagnetic waves on the ground, and its value increases with increasing unevenness in the ground. Therefore, when the ground settles due to underground cavities or loosening of the soil, and unevenness such as cracks occurs, the value increases. Thus, abnormalities in the ground (underground cavities or loosening of the soil) can be detected by surface scattering.
[0016] Corner scattering is a type of scattering that occurs when electromagnetic waves are reflected twice by underground cavities. The value increases when underground cavities are present. Therefore, anomalies (underground cavities) can be detected using corner scattering.
[0017] Volume scattering is a type of scattering that occurs when electromagnetic waves that have penetrated the ground to some extent are reflected in complex ways within the ground. The value of volume scattering increases when the density of the ground decreases, such as when soil flows into underground pipelines and the ground becomes loose. Therefore, abnormalities in the ground (ground loosening) can be detected by volume scattering.
[0018] Images showing surface scattering, corner scattering, and volume scattering are each acquired as grayscale images, representing the values corresponding to each scattering in the SAR measurement data using a gradation from 0 to 255, with a minimum value of 0 and a maximum value of 255. The acquired grayscale images are then normalized to values between 0 and 1.
[0019] The input unit 11 may receive a single-layer matrix as scattering data, which is a matrix representation of images showing either normalized surface scattering, corner scattering, or volume scattering. The input unit 11 may also receive a two-layer matrix, which is a matrix representation of superimposed images showing two of the normalized surface scattering, corner scattering, and volume scattering. The input unit 11 may also receive a three-layer matrix, which is a matrix representation of superimposed images showing normalized surface scattering, corner scattering, and volume scattering. When at least one of surface scattering, corner scattering, and volume scattering represented in matrix form is input to the input unit 11, the scattering data is obtained from measurement data at a location corresponding to the road. Measurement data from reflections from unpaved ground and buildings may not accurately reflect surface scattering, corner scattering, and volume scattering. Therefore, it is desirable to obtain the scattering data from measurement data at a location corresponding to the road.
[0020] Furthermore, the input unit 11 may receive, as scattering data, difference data before and after a predetermined time has elapsed for at least one of the normalized surface scattering, corner scattering, and volume scattering. For example, the input unit 11 may receive, as scattering data, the difference data between an image (matrix) showing the current surface scattering and an image (matrix) showing the surface scattering from a predetermined time earlier (for example, several days ago). The input unit 11 may also receive difference data before and after a predetermined time has elapsed for corner scattering and volume scattering, similar to surface scattering. When difference data for multiple scattering types is input as scattering data, the elapsed time for taking the difference shall be the same for each scattering type. Note that elements for which the difference before and after the predetermined time has elapsed shall be set to 0.
[0021] By using the difference data of scattering before and after a predetermined time period as scattering data, it is possible to observe the increase in surface scattering, corner scattering, and volume scattering values when an anomaly occurs in the ground, thereby eliminating the influence of differences in the road surface from which the measurement data was acquired.
[0022] The input unit 11 may be connected to the detection device 10 as an external input device, instead of being provided within the detection device 10. Any connection method can be used, such as USB, HDMI®, or Bluetooth®. "USB" is an abbreviation for Universal Serial Bus. "HDMI®" is an abbreviation for High-Definition Multimedia Interface.
[0023] The output unit 12 includes at least one output interface. The output interface is, for example, a display. The display is, for example, an LCD or an organic EL display. "LCD" is an abbreviation for Liquid Crystal Display. "EL" is an abbreviation for Electro Luminescence. The output unit 12 outputs (displays) data obtained by the operation of the detection device 10. For example, the output unit 12 displays the detection results of anomalies in the ground of the target area by the detection device 10 superimposed on a map of the target area. This makes it easier to check the status of the laying of facilities such as sewer pipes underground and the status of houses around the point where an anomaly was detected, and makes it easier to determine whether a detailed investigation of the point where an anomaly was detected is necessary.
[0024] The output unit 12 may be connected to the detection device 10 as an external output device, instead of being provided within the detection device 10. Any connection method can be used, such as USB, HDMI®, or Bluetooth®.
[0025] The memory unit 13 includes at least one semiconductor memory, at least one magnetic memory, at least one optical memory, or any combination thereof. The semiconductor memory is, for example, RAM, ROM, or flash memory. "RAM" is an abbreviation for Random Access Memory. "ROM" is an abbreviation for Read Only Memory. The RAM is, for example, SRAM or DRAM. "SRAM" is an abbreviation for Static Random Access Memory. "DRAM" is an abbreviation for Dynamic Random Access Memory. The ROM is, for example, EEPROM. "EEPROM" is an abbreviation for Electrically Erasable Programmable Read Only Memory. The flash memory is, for example, SSD. "SSD" is an abbreviation for Solid-State Drive. The magnetic memory is, for example, HDD. "HDD" is an abbreviation for Hard Disk Drive. The storage unit 13 functions, for example, as a main memory, auxiliary memory, or cache memory. The storage unit 13 stores information used for the operation of the detection device 10, which is input via the input unit 11, and information obtained by the operation of the detection device 10.
[0026] The control unit 14 includes at least one processor, at least one programmable circuit, at least one dedicated circuit, or any combination thereof. The processor is a general-purpose processor such as a CPU (Central Processing Unit) or GPU (Graphics Processing Unit), or a dedicated processor specialized for a specific process. The programmable circuit is, for example, an FPGA (Field-Programmable Gate Array). The dedicated circuit is, for example, an ASIC (Application Specific Integrated Circuit).
[0027] The control unit 14 controls each part of the detection device 10 and executes processes related to the operation of the detection device 10. As shown in Figure 1, the control unit 14 functions as a generation unit 141 and an identification unit 142.
[0028] The generation unit 141 receives actual scattering data via the input unit 11. From the input actual scattering data, the generation unit 141 generates false scattering data representing a state where there are no abnormalities in the ground of the target area. The generation unit 141 is equipped with a generator 141a, which generates the false scattering data.
[0029] The generator 141a generates false scattering data with a single layer matrix if the actual scattering data is a single layer matrix. Similarly, the generator 141a generates false scattering data with a two-layer matrix if the actual scattering data is a two-layer matrix, and generates false scattering data with a three-layer matrix if the actual scattering data is a three-layer matrix.
[0030] The generator 141a is constructed using machine learning. Figure 2 is a diagram illustrating the training of the generator 141a and the discriminator 142a, which will be described later. The generator 141a and the discriminator 142a are constructed using a deep learning method called a Generative Adversarial Network (GAN).
[0031] The generator 141a receives scattering data (1-layer matrix, 2-layer matrix, or 3-layer matrix) as input. If the input scattering data is a 1-layer matrix, the generator 141a generates false scattering data of a 1-layer matrix; if the input scattering data is a 2-layer matrix, it generates false scattering data of a 2-layer matrix; and if the input scattering data is a 3-layer matrix, it generates false scattering data of a 3-layer matrix. The generator 141a outputs the generated false scattering data to the classifier 142a.
[0032] The classifier 142a receives false scattering data generated by the generator 141a. The classifier 142a also receives scattering data acquired at the same location but at a different time (for example, more than 24 hours after the scattering data input to the generator 141a was acquired) as judgment data. The scattering data input to the classifier 142a (judgment data) and the scattering data input to the generator 141a are scattering data that shows the same scattering. Furthermore, the scattering data input to the generator 141a and the scattering data input to the classifier 142a (judgment data) are data acquired at a time when no abnormalities were detected in the ground.
[0033] Furthermore, if at least one difference data of surface scattering, corner scattering, and volume scattering is input to the input unit 11, the same difference data is also input to the generator 141a and the classifier 142a. The time interval for calculating the difference is the same for the difference data input to the input unit 11 and the difference data input to the generator 141a and the classifier 142a.
[0034] The classifier 142a learns to distinguish between the false scattered data generated by the generator 141a and the scattered data input as judgment data. Specifically, the classifier 142a uses the Residual loss L shown in the following equations (1) and (2). R (z) and Discrimination loss L D Calculate (z). Residual loss L R (z) is the sum of the brightness differences between the judgment data and the false scattering data. Also, the Discrimination loss L D (z) is the sum of the differences in features between the judgment data and the false scattering data.
[0035]
[0036] In equations (1) and (2), x is the decision data, G(z) is the false scattering data, and f() is the feature output in the intermediate layer of the classifier 142a.
[0037] The discriminator 142a calculates the Residual loss LR (z) and Discrimination loss L D (z), the Anomaly-score shown in the following equation (3) is generated. Anomaly-score=(1−λ)L R (z)+λL D (z) Equation (3)
[0038] In equation (3), λ is a constant between 0 and 1.
[0039] The discriminator 142a is trained so that the Anomaly-score is minimized. The discriminator 142a feeds back the authenticity discrimination result of the fake scatter data generated by the generator 141a to the generator 141a. The generator 141a newly generates fake scatter data according to the authenticity discrimination result of the fake scatter data fed back from the discriminator 142a.
[0040] The above-described training of the generator 141a and the discriminator 142a is repeated until the state where the fake scatter data generated by the generator 141a cannot be distinguished from the determination data, that is, the authenticity discrimination of the fake scatter data by the discriminator 142a approaches a probability of 50%. In this way, the generator 141a and the discriminator 142a are constructed by training through repetition of the generation of fake scatter data by the generator 141a according to the discrimination result of the discriminator 142a, and the authenticity discrimination of the fake scatter data by the discriminator 142a.
[0041] The above-described training of the generator 141a and the discriminator 142a is referred to as non-defective product training. Scatter data in a normal state with no anomaly is input to the discriminator 142a, and the generator 141a and the discriminator 142a are trained so that the difference between the scatter data and the fake scatter data generated by the generator 141a is minimized. Therefore, the generator 141a is trained to generate scatter data in a normal state with no anomaly in the ground from the input scatter data.
[0042] Referring again to FIG. 1, the identification unit 142 receives actual scattering data via the input unit 11. The identification unit 142 also receives fake scattering data generated by the generation unit 141. The identification unit 142 detects an abnormality in the underground of the target area based on the difference between the actual scattering data and the fake scattering data. Specifically, when the difference between the actual scattering data and the fake scattering data is equal to or greater than a predetermined threshold, the identification unit 142 identifies that there is an abnormality in the underground of the target area. Further, when the difference between the actual scattering data and the fake scattering data is less than the predetermined threshold, the identification unit 142 identifies that there is no abnormality in the underground of the target area.
[0043] As shown in FIG. 1, the identification unit 142 includes a discriminator 142a. As described with reference to FIG. 2, the discriminator 142a is the one used for learning of the generator 141a. The discriminator 142a detects an abnormality in the underground of the target area based on the difference between the actual scattering data and the fake scattering data. Specifically, the discriminator 142a calculates the Anomaly-score shown in the above-described equation (3), and if the calculated Anomaly-score is equal to or greater than a predetermined threshold, it identifies that there is an abnormality in the underground of the target area.
[0044] As described above, the generator 141a generates scattering data in a normal state where there is no abnormality in the underground. Therefore, when the actual scattering data is data acquired in a state where there is an abnormality in the underground, the difference between the actual scattering data and the fake scattering data generated by the generator 141a becomes large. Accordingly, the identification unit 142 can identify the presence or absence of an abnormality in the underground of the target area based on the difference between the actual scattering data and the fake scattering data.
[0045] Note that in the present embodiment, the description has been given using an example in which the identification unit 142 (identifier 142a) detects an underground anomaly in the target area based on the Anomaly-score, but the present disclosure is not limited thereto. The identification unit 142 may detect an underground anomaly in the target area based on an index different from the Anomaly-score. Examples of such an index include Peak Signal-to-Noise Ratio (PSNR) and Structural Similarity Index (SSIM).
[0046] PSNR is an index that captures changes in luminance, and can be calculated by the following equation (4).
[0047]
[0048] In equation (4), MAX is the maximum value of the pixel values. Further, MSE can be calculated by the following equation (5).
[0049]
[0050] In equation (5), I(x,y) is the pixel value of the (x,y) pixel of the actual scattering data (image representing each scattering). I(x,y) is the pixel value of the (x,y) pixel of the fake scattering data (image representing each scattering).
[0051] If there is an anomaly underground, the intensity of the scattering data increases. Therefore, by using PSNR, which is an index that captures changes in luminance, the underground anomaly can be detected.
[0052] SSIM is an index calculated for each small region (window) in an image, and can be calculated by the following equation (6).
[0053]
[0054] In equation (6), a represents the actual scattering data, and b represents the fake scattering data. Further, μ is the average pixel value within the window, and σ is the standard deviation of the pixel values within the window. Further, σ a,b is the covariance between the actual scattering data and the fake scattering data. Further, C 1 , C2 It is a constant.
[0055] Because SSIM smooths the image using a window, it becomes easier to detect changes when there are anomalies in the ground that alter the overall scattered data.
[0056] Furthermore, the identification unit 142 may combine the above-described multiple indicators to identify whether or not there are any abnormalities in the ground of the target area.
[0057] The function of the detection device 10 is realized by executing the program according to this embodiment on the processor acting as the control unit 14. In other words, the function of the detection device 10 is realized by software. The program causes the computer to perform the operations of the detection device 10, thereby causing the computer to function as the detection device 10. That is, the computer functions as the detection device 10 by performing the operations of the detection device 10 according to the program.
[0058] The program can be stored on a non-temporary computer-readable medium. Examples of non-temporary computer-readable mediums include flash memory, magnetic recording devices, optical discs, magneto-optical recording media, or ROM. The program can be distributed, for example, by selling, transferring, or leasing portable media such as SD (Secure Digital) cards, DVDs (Digital Versatile Discs), or CD-ROMs (Compact Disc Read Only Memory) on which the program is stored. The program may also be distributed by storing it in server storage and transferring it from the server to other computers. The program may also be provided as a program product.
[0059] A computer, for example, stores a program stored on a portable medium or a program transferred from a server in its main memory. Then, the computer reads the program stored in the main memory with its processor and executes the processing according to the read program. The computer may also read the program directly from the portable medium and execute the processing according to the program. The computer may also execute the processing according to the received program sequentially each time a program is transferred to it from a server. Processing may also be performed by a so-called ASP (Application Service Provider) type service, which does not transfer programs from the server to the computer, but realizes its function only through execution instructions and result retrieval. A program includes information used for processing by an electronic computer that is equivalent to a program. For example, data that is not a direct instruction to the computer but has the nature of defining the computer's processing falls under "equivalent to a program".
[0060] Some or all of the functions of the detection device 10 may be implemented by a programmable circuit or a dedicated circuit as the control unit 14. In other words, some or all of the functions of the detection device 10 may be implemented by hardware.
[0061] Next, the operation of the detection device 10 according to this embodiment will be described. Figure 3 is a flowchart showing an example of the operation of the detection device 10 according to this embodiment, and is a diagram for explaining the detection method performed by the detection device 10 according to this embodiment.
[0062] The input unit 11 receives input of actual scattering data that shows at least one of surface scattering, corner scattering, and volume scattering obtained from SAR measurement data (step S11).
[0063] The control unit 14 (generation unit 141) generates false scattering data from the actual input scattering data, representing a state where there are no abnormalities in the ground of the target area (step S12). Specifically, the control unit 14 generates false scattering data using the generator 141a constructed by learning as described with reference to Figure 2.
[0064] Then, the control unit 14 (identification unit 142) detects anomalies in the ground of the target area based on the difference between the actual scattered data and the false scattered data (step S13). Specifically, the control unit 14 uses the classifier 142a, which was constructed by learning as described with reference to Figure 2, to detect anomalies in the ground of the target area based on the difference between the actual scattered data and the false scattered data.
[0065] As described above, the detection device 10 according to this embodiment includes a control unit 14. The control unit 14 (generation unit 141) generates false scattering data representing a state where there is no abnormality in the ground of the target area from actual scattering data showing at least one of surface scattering, corner scattering, and volume scattering, obtained from measurement data of a synthetic aperture radar that radiates electromagnetic waves onto the ground of the target area and measures the reflected waves when the electromagnetic waves are reflected by the ground. The control unit 14 (identification unit 142) also detects abnormalities in the ground of the target area based on the difference between the actual scattering data and the false scattering data.
[0066] The control unit 14 (generation unit 141) generates false scattering data from the actual scattering data, representing a state where there are no abnormalities in the ground of the target area. Therefore, if there are abnormalities in the ground of the target area, the difference between the actual scattering data and the false scattering data becomes large. Accordingly, according to this disclosure, abnormalities in the ground of the target area can be detected based on the difference between the actual scattering data and the false scattering data.
[0067] The following additional information is disclosed regarding the embodiments described above.
[0068] [Addendum 1] A detection device for detecting anomalies underground in a target area, comprising a control unit, wherein the control unit emits electromagnetic waves onto the ground in the target area and measures the reflected waves reflected by the ground. The control unit generates false scattering data representing a state where there are no anomalies underground in the target area from actual scattering data showing at least one of surface scattering, corner scattering, and volume scattering obtained from measurement data of a synthetic aperture radar, and is configured to detect anomalies underground in the target area based on the difference between the actual scattering data and the false scattering data.
[0069] [Appendix Item 2] The detection device described in Appendix Item 1, wherein the control unit generates the false scattering data using a generator, and calculates the difference between the actual scattering data and the false scattering data using a classifier, and the generator and the classifier are constructed by learning through repeated generation of the false scattering data by the generator according to the identification result of the classifier, and identification of the truth value of the false scattering data by the classifier.
[0070] [Appendix 3] A detection device according to Appendix 1 or 2, wherein the scattering data is obtained from the measurement data at a location corresponding to a road in the target area.
[0071] [Appendix 4] A detection device according to Appendix 1 or 2, wherein the scattering data is difference data before and after a predetermined time has elapsed for at least one of the surface scattering, corner scattering, and volume scattering.
[0072] [Appendix 5] A detection method performed by a detection device for detecting anomalies underground in a target area, comprising: generating false scattering data representing a state where there are no anomalies underground in the target area from actual scattering data showing at least one of surface scattering, corner scattering, and volume scattering obtained from measurement data of a synthetic aperture radar that radiates electromagnetic waves onto the ground of the target area and measures the reflected waves reflected by the electromagnetic waves from the ground; and detecting an anomaly underground in the target area based on the difference between the actual scattering data and the false scattering data.
[0073] [Appendix 6] A non-temporary storage medium storing a program executable by a computer, the non-temporary storage medium storing a program that causes the computer to function as a detection device as described in any one of the appendix items 1 to 4.
[0074] Although the embodiments described above are representative examples, it will be apparent to those skilled in the art that many modifications and substitutions are possible within the spirit and scope of this disclosure. Therefore, the present invention should not be construed as being limited by the embodiments described above, and various modifications or changes are possible without departing from the claims. For example, it is possible to combine multiple component blocks shown in the configuration diagram of the embodiments into one, or to divide one component block.
[0075] 10 Detection device 11 Input unit 12 Output unit 13 Storage unit 14 Control unit 141 Generation unit 142 Identification unit 141a Generator 142a Identifier
Claims
1. A detection device for detecting anomalies underground in a target area, comprising: a generation unit that generates false scattering data indicating no anomalies underground in the target area from actual scattering data showing at least one of surface scattering, corner scattering, and volume scattering obtained from measurement data of a synthetic aperture radar that radiates electromagnetic waves onto the ground in the target area and measures the reflected waves reflected by the electromagnetic waves from the ground; and an identification unit that detects anomalies underground in the target area based on the difference between the actual scattering data and the false scattering data.
2. A detection device according to claim 1, wherein the generation unit generates false scattering data using a generator, the identification unit calculates the difference between the actual scattering data and the false scattering data using a classifier, and the generator and the classifier are constructed by learning through repeated generation of false scattering data by the generator according to the identification result of the classifier and identification of the truth value of the false scattering data by the classifier.
3. A detection device according to claim 1, wherein the scattering data is obtained from the measurement data at a location corresponding to a road in the target area.
4. A detection device according to claim 1, wherein the scattering data is difference data before and after a predetermined time has elapsed for at least one of the surface scattering, corner scattering, and volume scattering.
