Ground deformation analysis device and ground deformation analysis method

Abstract

The present invention is provided with: a time series saliency data acquisition unit (12) for converting time series deformation data indicating ground surface deformation to time series saliency data representative of the saliency of the deformation; a candidate precursor site detection unit (13) that takes a point in time where a saliency index value is shown to be greater than a threshold value to be an abnormal point in time, and detects a predetermined number or more of predetermined unit areas where the abnormal point in time is identical as being candidate precursor sites for a geological disaster; and a precursor site extraction unit (14) for extracting, as being a precursor site for a geological disaster, predetermined unit areas present at locations at an interval of a predetermined distance or less from one another, from among the detected candidate precursor sites. On the basis of, rather than time series deformation data itself, the time series saliency data converted from the time series deformation data, only the predetermined number or more of predetermined unit areas determined to be abnormal points in time at identical points in time are detected as being candidate precursor sites. Furthermore, solely those candidate precursor sites present at locations at an interval of a predetermined distance or less from one another are extracted as being a precursor site. As a result of the foregoing, it is possible to more accurately detect ground deformation conditions and identify precursors of geological disasters.

Description

Technical Field

[0001] The present invention relates to a ground movement analysis device and a ground movement analysis method, and is particularly suitable for use in an apparatus and method for analyzing the situation of ground movement using observation data measured using a synthetic aperture radar.

Background Art

[0002] The properties of the ground on the ground surface are various and may cause ground disasters such as landslides, ground subsidence, slope collapses, sinkholes, embankment collapses, and liquefaction. Ground disasters mainly occur due to earthquakes or weather conditions in many cases. However, some ground disasters may also occur due to ground deterioration or creep behavior. Since ground disasters have a great impact on society, it is important to monitor the ground movement situation, detect precursors of ground disasters, and issue warnings.

[0003] Conventionally, for the purpose of precursor observation of landslides, etc., a technique has been proposed to obtain the amount of ground movement along the ground surface at the measurement target point using a synthetic aperture radar (SAR) mounted on a satellite (see, for example, Patent Documents 1 and 2). Also, a technique has been known in which terrain movement data is obtained using interferometric SAR (InSAR) image data generated by a predetermined interference process (see, for example, Patent Document 3).

[0004] Patent Document 1: Japanese Patent Application Laid-Open No. 2021-56008 Patent Document 2: Japanese Patent No. 3987451 Patent Document 3: Japanese Patent Application Laid-Open No. 9-5433

[0005] In the landslide area detection device described in Patent Document 1, landslide-like ground movement is detected based on at least a combination of landslide characteristics in each of the phase difference data and coherence data obtained by performing interference analysis on the surface data obtained by radar measurement at different timings.

[0006] The method for measuring ground surface deformation described in Patent Document 2 determines the amount of ground surface deformation by performing synthetic aperture interferometry on the ground surface reflected waves of pulsed radio waves emitted from a single synthetic aperture radar mounted on a satellite. Specifically, by performing synthetic aperture interferometry on the reflected waves acquired twice each in the ascending and descending orbits, the distance from the ascending and descending orbits to the measurement target point is determined, and by adding the condition of the direction of ground surface deformation specified by the ground structure, the actual amount of deformation from the past position of the measurement target point is determined.

[0007] In the interferometric synthetic aperture radar image processing apparatus described in Patent Document 3, two SAR reconstructed data are converted into frequency domain data by Fourier transform, and after shaping each frequency characteristic to improve the accuracy of the interference processing output, an inverse Fourier transform is performed to return to the original time domain data, and the time domain data processed in this way is sent to the interference processing unit. The interference processing unit performs interference processing on the two SAR reconstructed data with improved correlation to generate interference fringes containing elevation information, which are output as elevation data. [Overview of the project] [Problems that the invention aims to solve]

[0008] However, while Patent Documents 1 to 3 disclose monitoring of ground deformation, they do not disclose detecting precursors to ground disasters and issuing warnings. In order to issue appropriate warnings, it is necessary to capture ground deformation more accurately.

[0009] This invention was made to solve these problems, and aims to enable more accurate detection of ground deformation and the identification of precursors to ground disasters. [Means for solving the problem]

[0010] To solve the above-mentioned problems, the present invention converts time-series variation data showing fluctuations at predetermined intervals on the ground surface for each predetermined unit area included in the region of interest into time-series severity data representing the severity of fluctuations at predetermined intervals. Based on this time-series severity data, the time point at which the severity index value indicating the severity of fluctuations at predetermined intervals is greater than a threshold is defined as an abnormal time point. A predetermined number or more predetermined unit areas where the abnormal time point is the same are detected as candidates for ground disaster precursor sites. From among the predetermined unit areas detected as candidates for precursor sites, predetermined unit areas located at intervals of predetermined distance from each other are extracted as ground disaster precursor sites. [Effects of the Invention]

[0011] According to the present invention configured as described above, the time point in time and a predetermined unit area where abnormal fluctuations are occurring are detected based not on the time-series fluctuation data itself, which shows fluctuations at predetermined unit time intervals on the ground surface, but on time-series significance data that represents the significance of fluctuations at predetermined unit time intervals. Furthermore, according to the present invention, if the same time point is determined to be an abnormal time point in a predetermined number or more predetermined unit areas, then that predetermined number or more predetermined unit areas are detected as candidate locations for ground disaster precursors. Moreover, from among the detected candidate locations for ground disaster precursors, only predetermined unit areas located at intervals of a predetermined distance from each other are extracted as locations for ground disaster precursors. For this reason, according to the present invention, it is possible to more accurately grasp the situation of ground fluctuations, such as at what time and at what location fluctuations that could be precursors to ground disasters are occurring. [Brief explanation of the drawing]

[0012] [Figure 1] This figure shows the network configuration of the analysis system according to this embodiment. [Figure 2] This is a block diagram showing an example of the functional configuration of a ground deformation analysis device according to this embodiment. [Figure 3] This diagram illustrates the direction of Line of Sight (LoS). [Figure 4]This is a block diagram showing a specific example of the functional configuration of the time-series sampling data acquisition unit according to this embodiment. [Figure 5] This figure shows an example of time-series variation data and time-series spleness data. [Figure 6] This block diagram shows a specific example of the functional configuration of the precursor location candidate detection unit according to this embodiment. [Figure 7] This diagram illustrates the processing details of the precursor location candidate detection unit according to this embodiment. [Figure 8] This diagram illustrates the processing details of the precursor location extraction unit according to this embodiment. [Figure 9] This flowchart shows an example of the operation of the ground deformation analysis device according to this embodiment (an example of the processing procedure for the ground deformation analysis method). [Modes for carrying out the invention]

[0013] Hereinafter, an embodiment of the present invention will be described with reference to the drawings. Figure 1 is a diagram showing the network configuration of the analysis system 100 according to this embodiment. The analysis system 100 comprises a satellite base station 50 that receives satellite data from a satellite S flying in a satellite orbit in space, a database DB that records the satellite data, and a ground deformation analysis device 10 that analyzes the satellite data.

[0014] Satellite S is equipped with a satellite data acquisition unit (for example, an optical sensor and a synthetic aperture radar), and transmits satellite data, including images of the ground taken by the satellite data acquisition unit, to the satellite base station 50.

[0015] The satellite base station 50 receives satellite data from satellite S and records it in a database DB. The database DB is connected to a communication network N such as the Internet. The ground deformation analysis device 10 retrieves the satellite data stored in the database DB via the communication network N.

[0016] FIG. 2 is a block diagram showing a functional configuration example of the ground movement analysis device 10 according to the present embodiment. As shown in FIG. 2, the ground movement analysis device 10 of the present embodiment includes, as functional components, a time-series variation data acquisition unit 11, a time-series significance data acquisition unit 12, a precursor location candidate detection unit 13, and a precursor location extraction unit 14.

[0017] The above functional blocks 11 to 14 can be configured by any of hardware, DSP (Digital Signal Processor), and software. For example, when configured by software, the above functional blocks are configured with a computer's CPU, RAM, ROM, etc., and are realized by the operation of an analysis program stored in a storage medium such as RAM, ROM, hard disk, or semiconductor memory. The analysis program may be stored and provided in the storage medium, or may be provided via a communication network.

[0018] Note that the physical configuration described here is an example and does not necessarily have to be an independent configuration. For example, the ground movement analysis device 10 may include an LSI (Large-Scale Integration) in which a CPU, RAM, and ROM are integrated. Also, in the present embodiment, the ground movement device 10 is described in the case where it is configured by a single computer, but the ground movement analysis device 10 may be realized by a combination of multiple computers.

[0019] The time-series variation data acquisition unit 11 analyzes time-series observation data measured using synthetic aperture radar, which is one of the earth observation technologies, and acquires time-series variation data indicating the variation of the ground surface for each predetermined unit area included in the area of interest at each predetermined unit time. The area of interest refers to the area to be monitored for ground movement. The observation data used in this embodiment is, for example, a time-series SAR image stored in a storage medium (not shown) by daily observation using synthetic aperture radar. Also, the time-series variation data acquired in this embodiment is LoS variation data indicating the magnitude or speed of the variation in the line-of-sight direction when the ground surface is viewed from the satellite for each predetermined unit time.

[0020] The satellite equipped with synthetic aperture radar observes every point on the earth from two directions, the ascending orbit and the descending orbit, from a pre-specified satellite orbit. As a result, a time-series SAR image observed by a satellite whose satellite travel direction is the ascending orbit (hereinafter referred to as an ascending orbit SAR image) and a time-series SAR image observed by a satellite whose satellite travel direction is the descending orbit (hereinafter referred to as a descending orbit SAR image) are input to the time-series variation data acquisition unit 11.

[0021] The time-series variation data acquisition unit 11 analyzes the ascending orbit SAR image and the descending orbit SAR image to obtain data indicating the variation of the ground surface in the line-of-sight direction when the ground surface is viewed from the satellite in the ascending orbit (hereinafter referred to as ascending LoS variation data) and data indicating the variation of the ground surface in the line-of-sight direction when the ground surface is viewed from the satellite in the descending orbit (hereinafter referred to as descending LoS variation data). Here, as the analysis performed by the time-series variation data acquisition unit 11, it is possible to use known InSAR analysis.

[0022] FIG. 3 is a diagram for explaining the LoS direction, which is the line-of-sight direction when the ground surface is viewed from the satellite. The radar wave emitted from the satellite is irradiated in the direction of the incident angles λasc and λdesc shifted from the vertical direction with respect to the ground surface, rather than in the vertical direction. That is, as shown in FIG. 3(a), the ascending orbit The incidence angles λasc and λdesc differ slightly between the satellite in the orbit and the satellite in the southbound orbit shown in Figure 3(b). Radar waves are emitted in that direction. As a result, InSAR analysis determines the amount of deformation in the satellite's Loss of Sight direction, rather than the amount of deformation in the vertical direction of the Earth's surface.

[0023] The time-series variation data acquisition unit 11 replaces the northbound LoS variation data and southbound LoS variation data acquired as described above with variation data for predetermined unit time intervals and predetermined unit area intervals. Here, the time-series variation data acquisition unit 11 performs time axis and spatial axis matching between the northbound LoS variation data and the southbound LoS variation data, and detects the matched northbound LoS variation data and southbound LoS variation data.

[0024] Northbound LoS fluctuation data and southbound LoS fluctuation data may not coincide in the time axis and spatial axis directions. Therefore, the time-series fluctuation data acquisition unit 11 sets a threshold in the time axis direction, and if there is northbound LoS fluctuation data and southbound LoS fluctuation data that can be considered identical within the range of the threshold, it matches and detects them. Similarly, a threshold is set in the spatial axis direction, and if there is northbound LoS fluctuation data and southbound LoS fluctuation data that can be considered identical within the range of the threshold, it matches and detects them. Here, matching refers to the process of determining whether there is a point where northbound LoS fluctuation data and southbound LoS fluctuation data can be considered identical within the range of the threshold set in the time axis direction or the spatial axis direction.

[0025] In this embodiment, as an example of the time axis threshold and spatial axis threshold used when performing matching, it is possible to use thresholds that are generally set for the Sentinel-1 satellite, which is used when measurements requiring strictness and precision are required. Alternatively, the matching conditions may be made more lenient by making at least one of the time axis threshold and the spatial axis threshold larger than the threshold generally set for the Sentinel-1 satellite. For example, the time axis threshold can be set to 11 days, and the spatial axis threshold can be set to a rectangular area of ​​20m square.

[0026] The time-series variation data acquisition unit 11, for example, uses northbound LoS variation data as master data and southbound LoS variation data as slave data, and detects southbound LoS variation data that can be considered identical to the northbound LoS variation data within a time range of 11 days and a distance difference of 20m square. Only when a state that can be considered identical is detected, it adopts the matched northbound LoS variation data and southbound LoS variation data for predetermined time units (for example, every 11 days) and predetermined area units (for example, every 20m square rectangular area). In this case, the predetermined time unit and predetermined area unit are determined based on the northbound LoS variation data used as master data.

[0027] The time-series agility data acquisition unit 12 converts the time-series variation data acquired by the time-series variation data acquisition unit 11 into time-series agility data that represents the agility of the variation for each predetermined unit time. This time-series agility data is data that shows agility index values ​​indicating the agility of the variation at the ground surface for each predetermined unit time. The agility index value is an index value designed so that the value increases as the variation becomes more agility.

[0028] Figure 4 is a block diagram showing a specific example of the functional configuration of the time-series splendor data acquisition unit 12. As shown in Figure 4, the time-series splendor data acquisition unit 12 of this embodiment includes a time-frequency conversion unit 12a, a correction calculation unit 12b, and a frequency-time conversion unit 12c as its specific functional configuration.

[0029] The time-frequency conversion unit 12a converts the time-series variation data acquired by the time-series variation data acquisition unit 11 into frequency domain spectral data. For example, the time-frequency conversion unit 12a obtains frequency spectral data representing the amplitude for each frequency by performing a Fast Fourier Transform (FFT) on the time-series variation data.

[0030] The correction calculation unit 12b performs a correction calculation on the frequency spectrum data converted by the time-frequency conversion unit 12a, based on a reference value calculated by an aggregate function filter using a moving window. For example, the correction calculation unit 12b obtains the spectral residual by subtracting the reference value from the amplitude value or amplitude-related value specified by the frequency spectrum data. The reference value is obtained by aggregating the amplitude values ​​contained in a moving window of a predetermined size using a predetermined aggregate function. Examples of aggregate functions include a function to calculate the mean, a function to calculate the median, or a function to calculate the chi-squared value.

[0031] An example of the calculation performed in the correction calculation unit 12b is shown in the following [Equation 1]. Here, F(X) is frequency spectrum data, Amp(f) is the amplitude value for each frequency, L(f) is the logarithm of the amplitude for each frequency (corresponding to amplitude-related values), H is an aggregate function filter using a moving window, and Ave(L(f)) is a reference value calculated by multiplying the amplitude logarithm L(f) by the aggregate function filter H (in the case of [Equation 1]). (where L(f) is the moving average), and Res(f) is the spectral residual for each frequency. Here, the spectral residual Res(f) is calculated using the amplitude logarithm L(f), but it may also be calculated using the amplitude value Amp(f).

[0032]

number

[0033] The frequency-time conversion unit 12c converts the spectral residual data calculated by the correction calculation unit 12b into time-domain data, i.e., time-series splendor data. The processing performed by the frequency-time conversion unit 12c is the reverse of that performed by the time-frequency conversion unit 12a. Here, the frequency-time conversion unit 12c converts the spectral residual data into time-series splendor data by performing an inverse FFT operation on the spectral residual data.

[0034] For example, the frequency-time conversion unit 12c obtains time-series splendor data S(X) by performing an inverse FFT process according to the following equation [Equation 2]. Here, F -1 represents the inverse FFT process, and Ph(f) represents the phase of the spectral residual.

[0035]

number

[0036] Figure 5 shows an example of time-series fluctuation data input to the time-series splendor data acquisition unit 12 and time-series splendor data output from the time-series splendor data acquisition unit 12. This Figure 5 shows the time-series gradient of data for a certain predetermined unit area at predetermined unit time intervals (every 11 days), with each dot representing data acquired every 11 days. Figure 5(a) is the time-series fluctuation data, and Figure 5(b) is the time-series splendor data.

[0037] Returning to Figure 2, the precursor location candidate detection unit 13 uses the time-series agility data acquisition unit 12 to identify a point in time when the agility index value, which indicates the agility of fluctuations for a predetermined unit time, is greater than a threshold, as an abnormal point in time. It then detects a predetermined number or more predetermined unit areas where the abnormal point in time is the same as candidate locations for ground disaster precursors.

[0038] In this embodiment, instead of immediately detecting a predetermined unit area where an abnormal time point indicating a significantness index value greater than a threshold is detected as a candidate for a precursor location, only if a predetermined number or more of predetermined unit areas with the same abnormal time point exist are detected as candidates for a precursor location. This means that the detection of precursor location candidates is limited to multiple predetermined unit areas that are temporally related.

[0039] Figure 6 is a block diagram showing a specific example of the functional configuration of the precursor location candidate detection unit 13. As shown in Figure 6, the precursor location candidate detection unit 13 of this embodiment includes an abnormal time point detection unit 13a and a simultaneous abnormal area detection unit 13b as its specific functional configuration.

[0040] The abnormal time point detection unit 13a detects abnormal time points in each predetermined unit area where the sampling index value is greater than the threshold, based on the time-series sampling data acquired by the time-series sampling data acquisition unit 12. The processing details of this abnormal time point detection unit 13a will be explained using Figure 7. Figure 7(a) shows the time-series sampling data and threshold for each predetermined unit area, and Figure 7(b) shows the processing results by the abnormal time point detection unit 13a.

[0041] Figure 7(a) shows time-series splendor data S1 to S10 for 10 predetermined unit areas A1 to A10. Each time-series splendor data contains 10 splendor index values ​​for each predetermined unit time. For example, the time-series splendor data S1 for predetermined unit area A1 contains 10 splendor index values ​​S1-01 to S1-10. The time-series splendor data S2 for predetermined unit area A2 contains 10 splendor index values ​​S2-01 to S2-10. Note that here, 10 time-series splendor data S1 to S10 The schematic examples shown, where each contains 10 sampling index values, are for illustrative purposes only. In reality, there are many more time-series sampling data points for a given unit area, and one time-series sampling data point contains many more sampling index values.

[0042] As shown in Figure 7(a), in this embodiment, thresholds Th1 to Th10 are set for each predetermined unit area. Here, the abnormal time point detection unit 13a determines the time series for each predetermined unit area A1 to A10. The time-series spleness data S1 to S10 acquired by the spleness data acquisition unit 12 is used to calculate thresholds Th1 to Th10, and the calculated thresholds Th1 to Th10 are individually compared with the spleness index values ​​S1-01 to S10-10 to detect the time point of an anomaly.

[0043] For example, the abnormal time point detection unit 13a calculates thresholds Th1 to Th10 using the upper quartiles and outliers of the sampling index values ​​S1-01 to S10-10 shown in the time-series sampling data S1 to S10, as shown in [Equation 3] below. In [Equation 3], uq i (i=1~10) indicates the upper quartiles of the sampling index values ​​Si-01~Si-10, IQR i The sampling index values ​​Si-01 to Si-10 represent the interquartile range (the 75th percentile minus the 25th percentile), and sv represents the sensitivity. The sensitivity value sv is set to 1.5, for example, the value commonly used in the definition of outliers. Alternatively, the initial value of sensitivity sv can be set to 1.5, allowing the user to adjust the value as appropriate depending on the topography and structure of the region of interest.

[0044]

number

[0045] The abnormal time point detection unit 13a compares 10 stellarity index values ​​S1-01 to S1-10 with a threshold Th1 individually based on the time-series stellarity data S1 of the first predetermined unit area A1, and detects time points where the stellarity index values ​​S1-01 to S1-10 are greater than the threshold Th1 as abnormal time points. Figure 7(b) shows the results, and the parts marked "abnormal" indicate that an abnormal time point was detected. In the example in Figure 7(b), the third stellarity index value S1-03 is determined to be greater than the threshold Th1, and the time point corresponding to the third stellarity index value S1-03 (2021 / 11 / 28) is an abnormal time point. It is detected as a point.

[0046] Furthermore, the abnormal time point detection unit 13a compares the 10 sampling index values ​​S2-01 to S2-10 and the threshold Th2 individually based on the time-series sampling data S2 of the second predetermined unit area A2, and the sampling index Time points in which values ​​S2-01 to S2-10 are greater than the threshold Th2 are detected as abnormal time points. In the example in Figure 7(b), the first significantness index value S2-01 and the tenth significantness index value S2-10 are determined to be greater than the threshold Th2, and the time point corresponding to the first significantness index value S2-01 (2021 / 11 / 6) The time point corresponding to the 10th significantness index value S2-10 (2022 / 2 / 13) has been detected as an abnormal time point.

[0047] The abnormal time point detection unit 13a detects abnormal time points by performing the same processing on the time-series severity data S3 to S10 for the third and subsequent predetermined unit areas A3 to A10.

[0048] The simultaneous abnormal area detection unit 13b detects, based on the detection results by the abnormal time detection unit 13a as shown in Figure 7(b), any predetermined number of predetermined unit areas that show a significantness index value greater than a threshold at the same time as candidate locations for a ground disaster precursor.

[0049] In other words, the simultaneous point anomaly area detection unit 13b looks at the table shown in Figure 7(b) in the column direction and, if it determines that there are a predetermined number or more predetermined unit areas marked as "abnormal" within the same column (i.e., within the same date), it detects those predetermined number or more predetermined unit areas as candidate precursor locations.

[0050] For example, looking at the column for 2021 / 11 / 6, there are five designated unit areas marked "Abnormal". Therefore, these five designated unit areas A2, A3, A6, A7, and A9 are detected as potential precursor locations. Also, looking at the column for 2022 / 2 / 13, there are six designated unit areas marked as "abnormal". These six designated unit areas A2, A5, A6, A8, A9, and A10 are detected as potential precursor locations. Anomalies were detected on 2021 / 11 / 28 and 2021 / 1 / 22, but more than five locations were detected at the same time. Since no abnormal time points were detected, these are not detected as potential precursor locations.

[0051] Returning to Figure 2, the explanation continues. The precursor location extraction unit 14 extracts predetermined unit areas located within a predetermined distance from each other from among the predetermined unit areas detected as precursor location candidates by the precursor location candidate detection unit 13 as described above, and identifies them as precursor locations for ground disasters. This means that precursor locations are extracted from among multiple precursor location candidates, limited to predetermined unit areas that are spatially related. Alternatively, an area formed to include multiple predetermined unit areas located within a predetermined distance from each other may be extracted as a precursor location for ground disasters.

[0052] Here, the value of the predetermined distance can be set to a different value depending on the SAR image used as input for the InSAR analysis. For example, when performing Sentinel-1 based InSAR analysis that includes both single-look and multi-look processing using the IPTA (Interferometric Point Target) method, the predetermined distance can be set to 44m.

[0053] If a candidate for a precursor location is detected according to the example shown in Figure 7(b), the precursor location extraction unit 14 determines that the five predetermined unit areas A2, A3, A6, A7, and A9 detected as of 2021 / 11 / 6 are relative to each other. The system determines whether or not locations exist within a predetermined distance interval, and extracts predetermined unit areas that are determined to exist within a predetermined distance interval as precursor locations. For example, if three predetermined unit areas A6, A7, and A9 are determined to exist within a predetermined distance interval, these three predetermined unit areas A6, A7, and A9 are extracted as precursor locations.

[0054] Similarly, the precursor location extraction unit 14 identifies six predetermined unit areas detected as of 2022 / 2 / 13. The system determines whether A2, A5, A6, A8, A9, and A10 are located within a predetermined distance from each other, and extracts predetermined unit areas that are determined to be located within a predetermined distance from each other as precursor locations.

[0055] In this way, the precursor location extraction unit 14 extracts predetermined unit areas located within a predetermined distance from each other from among a predetermined number or more predetermined unit areas detected as precursor location candidates by the precursor location candidate detection unit 13 at a given time, as precursor locations for ground disasters at the abnormal time detected by the abnormal time detection unit 13a.

[0056] Figure 8 is a diagram illustrating an example of the processing content of the precursor location extraction unit 14. Figure 8 shows a view of a certain area of ​​the ground surface from above, and the individual locations indicated by the ■ marks are multiple precursor location candidates P1 detected by the precursor location candidate detection unit 13 at a certain point in time. This indicates ~P5. In reality, there are many more predetermined unit areas that are detected as potential precursor locations. While several possibilities exist, Figure 8 is a simplified illustration for the sake of explanation.

[0057] Of the multiple candidate precursor locations P1 to P5 shown in Figure 8, three candidate precursor locations P1 to P3 are located in close proximity to each other, within a predetermined distance. On the other hand, the remaining two candidate precursor locations P4 and P5 are not located within a predetermined distance from any of the others. In this case, the precursor location extraction unit 14 extracts the set of the three candidate precursor locations P1 to P3 or the area PA containing them as a precursor location. The area PA containing the set of the three candidate precursor locations P1 to P3 is, for example, a rectangular region that encloses the three candidate precursor locations P1 to P3 in a tangent manner, as shown in Figure 8. The remaining two candidate precursor locations P4 and P5 are not extracted as precursor locations. In other words, candidate precursor locations P4 and P5, which are unlikely to have a spatial relationship with the other candidate precursor locations, are excluded from consideration as precursor locations.

[0058] Here, the precursor location extraction unit 14 may extract multiple precursor location candidates as precursor locations only if there are a predetermined number (for example, 5) or more precursor location candidates located in close proximity to each other at intervals of a predetermined distance.

[0059] Note that the shape of the precursor location area PA shown in Figure 8 is just an example and is not limited to this. For example, it may be a rectangle, circle, ellipse, or other polygon with the smallest area that encompasses the candidate precursor locations P1 to P3.

[0060] The above describes the processing performed by the precursor location candidate detection unit 13 and the precursor location extraction unit 14 using the time-series splendor data acquired by the time-series splendor data acquisition unit 12. However, precursor locations may be detected over the entire time series, or they may be detected over a portion of the time series. When detecting precursor locations over a portion of the time series, for example, precursor locations may be detected only from the most recent point in the time series, or they may be detected over the time series from the most recent point in the past up to a predetermined point in the past.

[0061] For example, the precursor location candidate detection unit 13 uses the prominentness index values ​​at the most recent point in time shown in the time-series prominentness data (in the example in Figure 7, prominentness index values ​​S1-10, S2-10, ..., S10-10 at the time of 2022 / 2 / 13) to identify a predetermined number of predetermined unit areas where the most recent point in time is an abnormal point in time. Detect it as a potential location.

[0062] In this case, the abnormal time point detection unit 13a determines, for each predetermined unit area, whether the significance index values ​​S1-10, S2-10, ..., S10-10 at the most recent time point are greater than the thresholds Th1 to Th10. For predetermined unit areas where the significantness index value exceeds a threshold, the most recent point in time is detected as an abnormal point in time. If the simultaneous point abnormal area detection unit 13b determines that there are a predetermined number or more abnormal points detected at the most recent point in time, it detects that predetermined number or more predetermined unit areas as candidates for ground disaster precursor locations.

[0063] Furthermore, the precursor location extraction unit 14 extracts predetermined unit areas located within a predetermined distance from each other from a predetermined number or more predetermined unit areas that were detected as precursor location candidates at the most recent time, as precursor locations for ground disasters. In this way, it is possible to constantly monitor whether or not precursors to ground disasters are occurring at the most recent time, and if so, where they are occurring. If it is determined that precursors to ground disasters are occurring, it is possible to issue a predetermined warning.

[0064] Alternatively, the precursor location candidate detection unit 13 detects precursor location candidates using multiple significantity index values ​​from the most recent point in time, as shown in the time-series significantity data, to a predetermined point in the past. The sampling number for determining how far back in time the significantity index values ​​to target may be a fixed value or a variable value that can be arbitrarily set by the user. Precursor location candidates may be detected using significantity index values ​​from any arbitrary period, not limited to significantity index values ​​from a specific period in the past from the most recent point in time.

[0065] Furthermore, by setting environmental conditions such as the season and climate of the area of ​​interest, a certain period may be automatically determined according to the set environmental conditions. Depending on the environmental conditions, there may be patterns of ground deformation (for example, snowmelt in summer or refreezing in winter). Therefore, depending on the environmental conditions, candidate precursor sites may be detected by targeting the notability index values ​​of a certain period corresponding to the deformation pattern specific to that environmental condition.

[0066] Furthermore, detecting precursor locations using multiple agility index values ​​over the entire period or a portion of the period shown in time-series agility data is also effective, for example, when, after a ground disaster has actually occurred, one wants to verify the signs of a ground disaster from the ground deformation conditions prior to the disaster and thereby investigate the cause of the ground disaster.

[0067] Figure 9 is a flowchart showing an example of the operation of the ground deformation analysis device 10 according to this embodiment (an example of the processing procedure for the ground deformation analysis method). First, the time-series deformation data acquisition unit 11 analyzes the time-series observation data measured using synthetic aperture radar and acquires time-series deformation data showing the deformation at predetermined unit time intervals for each predetermined unit area included in the region of interest (step SP1).

[0068] Next, the time-series splendor data acquisition unit 12 converts the time-series fluctuation data acquired by the time-series fluctuation data acquisition unit 11 into time-series splendor data that represents the splendor of fluctuations for each predetermined unit of time (step SP2).

[0069] Next, the precursor location candidate detection unit 13, based on the time-series agility data acquisition unit 12, identifies a point in time when the agility index value, which indicates the agility of fluctuations for a predetermined unit time, is greater than a threshold, as an abnormal point in time, and detects a predetermined number or more predetermined unit areas where the abnormal point in time is the same as candidates for ground disaster precursor locations (step SP3).

[0070] Furthermore, the precursor location extraction unit 14 extracts predetermined unit areas located within a predetermined distance from each other from among the predetermined unit areas detected as precursor location candidates by the precursor location candidate detection unit 13 as precursor locations for ground disasters (step SP4). With this, the processing shown in the flowchart in Figure 9 is completed.

[0071] As explained in detail above, in this embodiment, for each predetermined unit area included in the region of interest, time-series variation data showing the variation at a predetermined unit time interval on the ground surface is converted into time-series saturation data representing the saturation of the variation at a predetermined unit time interval. Based on this time-series saturation data, a point in time when the saturation index value showing the saturation of the variation at a predetermined unit time interval is greater than a threshold is defined as an abnormal point in time. After detecting a predetermined number or more predetermined unit areas where the abnormal point in time is the same as candidate locations for ground disaster precursors, predetermined unit areas located within a predetermined distance interval from each other are extracted as ground disaster precursor locations from among the predetermined unit areas detected as candidate locations for precursors.

[0072] According to this embodiment, instead of time-series variation data showing fluctuations at predetermined unit time intervals on the ground surface, the time point and predetermined unit area where abnormal fluctuations occur are detected based on time-series significance data representing the significance of fluctuations at predetermined unit time intervals. Furthermore, according to this embodiment, if the same time point is determined to be an abnormal time point in a predetermined number or more predetermined unit areas, then that predetermined number or more predetermined unit areas are detected as candidate locations for ground disaster precursors. Moreover, from among the detected candidate locations for ground disaster precursors, only predetermined unit areas located within a predetermined distance from each other are extracted as locations for ground disaster precursors. Therefore, according to this embodiment, it is possible to more accurately grasp the state of ground deformation, such as at what time and at what location fluctuations that could be precursors to ground disasters are occurring.

[0073] This will enable appropriate early warnings of disaster risks based on a more accurate understanding of ground deformation. For example, it will be possible to notify of precursory phenomena occurring before an actual ground disaster occurs. Furthermore, by detecting precursor locations based on the agility index values ​​for a specific period of time-series agility data, according to environmental conditions such as season and climate in the area of ​​interest, it will be possible to more accurately grasp the state of ground deformation in accordance with environmental conditions and provide appropriate warnings.

[0074] Furthermore, in this embodiment, a threshold is calculated for each predetermined unit area using time-series aggression data, and the detection of anomalies is performed by individually comparing the calculated threshold with the aggression index value. This makes it possible to individually set thresholds according to past ground deformation for each predetermined unit area, where the properties of the ground may differ, and to more accurately capture significant deformations for each predetermined unit area.

[0075] In the above embodiment, an example was described in which a precursor location for a ground disaster is detected using time-series agility data from the past to the most recent point in time. However, the present invention is not limited to this. For example, by analyzing the fluctuation trend of agility index values ​​from the entirety of time-series agility data from the past to the most recent point in time, future time-series agility data (extrapolated time-series data) that is expected for a point in time beyond the most recent point in time may be generated, and abnormal points in time and precursor locations may be detected using future agility index values. By doing so, it becomes possible to predict the possibility of a ground disaster occurring in the future at an earlier stage.

[0076] Furthermore, although the above embodiment describes an example in which the precursor location candidate detection unit 13 detects precursor location candidates at predetermined time intervals, the present invention is not limited thereto. For example, after detecting precursor location candidates at predetermined time intervals as described above, the precursor location candidates may be further narrowed down according to predetermined conditions. As an example, only when the precursor location candidates detected at predetermined time intervals are continuous or intermittent with respect to each other within a predetermined time interval may the precursor location extraction unit 14 process only those precursor location candidates that are continuous or intermittent in time.

[0077] Furthermore, although the above embodiment described an example in which different thresholds are set for each predetermined unit area, a fixed threshold common to all predetermined unit areas may also be used. However, as mentioned above, setting different thresholds for each predetermined unit area is preferable because it makes it possible to more accurately capture significant fluctuations for each predetermined unit area.

[0078] Furthermore, the above embodiments are merely examples of how the present invention may be implemented, and the technical scope of the invention should not be interpreted as being limited by them. In other words, the present invention can be implemented in various ways without departing from its gist or its main features. [Explanation of Symbols]

[0079] 10. Ground deformation analysis device 11. Time-series variation data acquisition unit 12. Time-series sampling data acquisition unit 12a Time-frequency conversion section 12b Correction calculation section 12c Frequency-Time Conversion Unit 13. Precursor location candidate detection unit 13a Anomaly detection unit 13b Simultaneous Anomaly Area Detection Unit 14. Precursor Point Extraction Unit

Claims

1. A time-series variation data acquisition unit analyzes time-series observation data measured using Earth observation technology and acquires time-series variation data showing the variation at predetermined unit time intervals on the Earth surface for each predetermined unit area included in the region of interest. A time-frequency conversion unit converts the time-series variation data acquired by the time-series variation data acquisition unit into spectral data in the frequency domain. A correction calculation unit performs a correction calculation on the spectral data in the frequency domain converted by the above time-frequency conversion unit, based on a reference value calculated by an aggregate function filter using a moving window. A frequency-time conversion unit converts the spectral residual data calculated by the correction calculation unit into time-domain data, and acquires the converted data as time-series significance data representing the significance of the variation for each predetermined unit time. Based on the time-series significance data acquired by the frequency-time conversion unit, the unit detects a predetermined number of predetermined unit areas where the significance index value indicating the significance of the fluctuations for each predetermined unit time is greater than a threshold, designating these points as abnormal points, and identifies these abnormal points as candidate locations for ground disaster precursors. The system includes a precursor location extraction unit that extracts predetermined unit areas located within a predetermined distance from each other from among the predetermined unit areas detected as precursor location candidates by the precursor location candidate detection unit as precursor locations for ground disasters. A ground deformation analysis device characterized by the following features.

2. The ground deformation analysis apparatus according to claim 1, characterized in that the above-mentioned precursor location candidate detection unit calculates the threshold value for each predetermined unit area using the time-series significance data acquired by the above-mentioned frequency-time conversion unit.

3. The ground deformation analysis device according to claim 2, characterized in that the above-mentioned precursor location candidate detection unit calculates the threshold using the upper quartiles and outliers of the above-mentioned sampling index values ​​shown in the above-mentioned time-series sampling data.

4. The ground deformation analysis apparatus according to any one of claims 1 to 3, characterized in that the correction calculation unit obtains spectral residuals by subtracting the reference value from the amplitude value or amplitude-related value identified by the spectral data.

5. The above-mentioned precursor location candidate detection unit is: An abnormal time point detection unit detects, based on the time-series stellarity data acquired by the frequency-time conversion unit, an abnormal time point detection unit that detects a time point in each predetermined unit area where the stellarity index value is greater than a threshold, The system also includes a simultaneous point anomaly area detection unit that, based on the detection results by the anomaly time point detection unit, detects a predetermined number or more predetermined unit areas as candidate precursor locations if there are a predetermined number or more predetermined unit areas at the same time in which the significance index value is greater than the threshold. A ground deformation analysis device according to any one of claims 1 to 3, characterized by the features described herein.

6. The ground deformation analysis device according to any one of claims 1 to 3, characterized in that the above-mentioned precursor location candidate detection unit determines whether the significance index value at the most recent point in time shown in the above-mentioned time-series significance data is greater than the threshold for each of the above-mentioned predetermined unit areas, and if the significance index value is greater than the threshold, the most recent point in time is designated as the above-mentioned abnormal point, and if there are a predetermined number or more predetermined unit areas designated as abnormal points, the predetermined number or more predetermined unit areas are detected as precursor location candidates.

7. The ground deformation analysis device according to any one of claims 1 to 3, characterized in that the above-mentioned precursor location candidate detection unit determines for each predetermined unit area whether multiple of the above-mentioned significantity index values ​​for the entire period or a part of the period shown in the above-mentioned time-series significantity data are greater than the above threshold, the time when the significantity index value is greater than the above threshold is defined as the above-mentioned abnormal time, and if there are a predetermined number or more predetermined unit areas where the above-mentioned abnormal time is the same, the predetermined number or more predetermined unit areas are detected as precursor location candidates.

8. The ground deformation analysis apparatus according to any one of claims 1 to 3, characterized in that the above-mentioned precursor location candidate detection unit generates future time-series agility data by analyzing the fluctuation trend of the agility index value based on the above-mentioned time-series agility data acquired by the above-mentioned frequency-time conversion unit, and detects the above-mentioned precursor location candidate using the said future time-series agility data.

9. The first step involves the time-series deformation data acquisition unit of the ground deformation analysis device analyzing time-series observation data measured using earth observation technology and acquiring time-series deformation data showing the deformation at predetermined unit time intervals on the ground surface for each predetermined unit area included in the region of interest. The second step involves the time-frequency conversion unit of the above-mentioned ground deformation analysis device converting the time-series deformation data acquired by the above-mentioned time-series deformation data acquisition unit into spectral data in the frequency domain. A third step in which the correction calculation unit of the above-mentioned ground deformation analysis device performs a correction calculation on the spectral data in the frequency domain converted by the above-mentioned time-frequency conversion unit, based on a reference value calculated by an aggregate function filter using a moving window, The fourth step involves the frequency-time conversion unit of the ground deformation analysis device converting the spectral residual data calculated by the correction calculation unit into time-domain data, and acquiring the converted data as time-series significance data representing the significance of the deformation for each predetermined unit time. A fifth step in which the precursor location candidate detection unit of the above-mentioned ground deformation analysis device detects a predetermined number of predetermined unit areas as precursor locations for ground disasters, based on the time-series significance data acquired by the frequency-time conversion unit, at a point in time when the significance index value indicating the significance of the above-mentioned deformation for each predetermined unit time is greater than a threshold, and at a predetermined number of predetermined unit areas where the abnormal point in time is the same. The ground deformation analysis device has a sixth step in which the precursor location extraction unit extracts predetermined unit areas located within a predetermined distance from each other from among the predetermined unit areas detected as precursor location candidates by the precursor location candidate detection unit as precursor locations for ground disasters. A method for analyzing ground deformation, characterized by the following features.

Images