Landslide disaster area detection device and landslide disaster area detection method
The device uses remote sensing and SAR images to enhance landslide disaster area detection in urban areas by analyzing vegetation changes and backscattering coefficient variations, achieving the same accuracy as in mountainous regions through integrated image analysis and morphological transformations.
Patent Information
- Application Number
- JP2024500139
- Authority / Receiving Office
- JP · JP
- Patent Type
- Patents
- Current Assignee / Owner
- Filing Date
- 2022-05-19
- Publication Date
- 2024-12-06
- Estimated Expiration
- 2042-05-19
AI Technical Summary
Existing methods for detecting landslide disaster areas in urban areas suffer from reduced accuracy due to diverse reflection spectrum changes and the presence of non-vegetated features like buildings and roads, making it difficult to achieve the same detection precision as in mountainous areas.
A landslide disaster area detection device that utilizes remote sensing images and synthetic aperture radar (SAR) images to detect vegetation changes, calculate slope angles, and analyze backscattering coefficient variations, combined with morphological transformations, to accurately identify landslide areas in both urban and mountainous regions.
Enables accurate detection of landslide disaster areas in urban environments with the same precision as in mountainous areas by integrating remote sensing and SAR image analysis with morphological transformations, effectively overcoming the challenges posed by non-vegetated urban features.
Smart Images

Figure 0007595804000003 
Figure 0007595804000004 
Figure 0007595804000005
Abstract
Description
[Technical field]
[0001] The present disclosure relates to a landslide disaster area detection device and a landslide disaster area detection method. [Background technology]
[0002] The damage area detection method described in Non-Patent Document 1 for detecting disasters caused by the movement of soil and sand, such as landslides, mudslides, and debris flows (hereinafter referred to as "landslide disasters"), utilizes the characteristic that a large change occurs in the distribution of vegetation during a landslide, etc., and detects areas damaged by a landslide, etc., based on changes in the vegetation index (NDVI: Normalized Difference Vegetation Index) on optical images, for example, changes in reflectance spectral characteristics, before and after the occurrence of a landslide disaster. [Prior art documents] [Non-patent literature]
[0003] [Non-Patent Document 1] Co-seismic landslide detection using ALOS satellite image in the Mianyuan River Basin, China, C. Denget. Et. Al., 2017 IEEE 9th ICCSN, May 2017. Summary of the Invention [Problem to be solved by the invention]
[0004] However, with the above-mentioned method, in areas that were not vegetated before the landslide disaster, such as urban areas with a variety of features such as buildings and roads, the changes in the reflectance spectral characteristics before and after the landslide disaster in the optical images also become diverse, and there is a problem that the detection accuracy is lower than in areas where the changes in the reflectance spectral characteristics are uniform, such as mountainous areas.
[0005] An object of the present disclosure is to provide a landslide disaster area detection device that is capable of detecting landslide disaster areas in urban areas with the same accuracy as that of detecting landslide disaster areas in mountainous regions. [Means for solving the problem]
[0006] In order to solve the above-mentioned problems, a landslide disaster area detection device according to the present disclosure includes a first acquisition unit that acquires remote sensing images before and after the occurrence of a landslide disaster, a first detection unit that detects a vegetated area from the remote sensing images, a second acquisition unit that acquires a digital elevation model of the area where the landslide disaster occurred, a first calculation unit that calculates a slope angle of a target point from the digital elevation model, and a second calculation unit that detects a slope angle of a target point in a mountainous area when the target point in the remote sensing image was a vegetated area before the occurrence of the landslide disaster, was no longer a vegetated area after the occurrence of the landslide disaster, and has a slope angle that exceeds a predetermined slope angle threshold. the first determination unit determining that the landslide disaster area is in a mountainous area; a third acquisition unit obtaining SAR images before and after the occurrence of the landslide; a second calculation unit calculating the variation in the backscattering coefficient in a local area from the SAR images; a third calculation unit calculating the difference in the variation before and after the occurrence of the landslide; a second determination unit limiting the landslide disaster area to the periphery of the mountainous area; a third determination unit determining that the landslide disaster area is in an urban area when the difference exceeds a predetermined threshold and the area is not a vegetated area both before and after the occurrence of the landslide; and a generation unit generating a landslide disaster area distribution based on the landslide disaster area in the mountainous area and the landslide disaster area in the urban area. Effect of the Invention
[0007] According to the landslide disaster area detection device according to the present disclosure, it is possible to detect landslide disaster areas in urban areas with the same accuracy as that of detecting landslide disaster areas in mountainous regions. [Brief description of the drawings]
[0008] [Figure 1] FIG. 2 is a functional block diagram of the landslide disaster area detection device CK of the first embodiment. [Diagram 2]1 shows the hardware configuration of a landslide disaster area detection device CK according to the first embodiment. [Diagram 3] 1 is a flowchart (part 1) showing the operation of the landslide disaster area detection device CK of the first embodiment. [Figure 4] 13 is a flowchart (part 2) showing the operation of the landslide disaster area detection device CK of the first embodiment. [Diagram 5] 1 shows the vicinity of the landslide disaster area SAJ in the mountainous region of embodiment 1. [Figure 6] Fig. 6A shows the relationship (part 1) between the vegetation points SC, difference evaluation value HCs, threshold evaluation value HCth, and landslide disaster area SIJ in urban areas in the first embodiment. Fig. 6B shows the relationship (part 2) between the vegetation points SC, difference evaluation value HCs, threshold evaluation value HCth, and landslide disaster area SIJ in urban areas in the first embodiment. Fig. 6C shows the relationship (part 3) between the vegetation points SC, difference evaluation value HCs, threshold evaluation value HCth, and landslide disaster area SIJ in urban areas in the first embodiment. Fig. 6D shows the relationship (part 4) between the vegetation points SC, difference evaluation value HCs, threshold evaluation value HCth, and landslide disaster area SIJ in urban areas in the first embodiment. [Figure 7] FIG. 11 is a functional block diagram of a landslide disaster area detection device CK according to the second embodiment. [Figure 8] 11 is a flowchart (part 1) showing the operation of the landslide disaster area detection device CK of the second embodiment. [Figure 9] 13 is a flowchart (part 2) showing the operation of the landslide disaster area detection device CK of the second embodiment. [Figure 10] Morphological transformation (part 1) is shown using a kernel KN created from a slope angle KK obtained from the digital elevation model SHM of the second embodiment. [Figure 11] The morphological transformation (part 2) is shown using a kernel KN created from a slope angle KK obtained from the digital elevation model SHM of the second embodiment. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS
[0009] An embodiment of a landslide disaster area detection device according to the present disclosure will be described.
[0010] Embodiment 1. First Embodiment The landslide disaster area detection device CK of the first embodiment will be described.
[0011] <Functions of the First Embodiment> 1 is a functional block diagram of a landslide disaster area detection device CK of embodiment 1. The functions of the landslide disaster area detection device CK of embodiment 1 will be described with reference to FIG.
[0012] The landslide disaster area detection device CK of embodiment 1 includes an acquisition unit SH, a detection unit KE, a calculation unit SA, a determination unit HA, and a generation unit SE, as shown in FIG. 1, in order to detect landslide disaster areas, which are areas where landslide disasters have occurred, in both mountainous areas and urban areas, using Synthetic Aperture Radar (SAR) images.
[0013] The term "urban area" does not mean only urban areas in the narrow sense, but in a broad sense means the general area where man-made buildings are concentrated, such as urban areas, residential areas, and industrial areas.
[0014] The acquisition unit SH corresponds to the "first acquisition unit," the "second acquisition unit," and the "third acquisition unit."
[0015] The detection unit KE corresponds to the "detection unit."
[0016] The calculation unit SA corresponds to the "first calculation unit," the "second calculation unit," and the "third calculation unit."
[0017] The judgment unit HA corresponds to the "first judgment unit," the "second judgment unit," and the "third judgment unit."
[0018] The generating unit SE corresponds to the "generating unit."
[0019] The acquisition unit SH has the following functions. As a first acquisition section, acquiring a remote sensing image RG1 before the occurrence of the landslide disaster and a remote sensing image RG2 after the occurrence of the landslide disaster. The second part is to obtain the digital elevation model (SHM) of the landslide-affected areas. The third acquisition part is to acquire SAR image SG1 before the landslide disaster occurs, and SAR image SG2 after the landslide disaster occurs.
[0020] Both the remote sensing image RG and the SAR image SG are images of an object measured from a position distant from the object, without touching the object.
[0021] The detection unit KE has the following functions. -Detecting vegetation areas SR from remote sensing images RG1 and RG2.
[0022] The calculation unit SA has the following functions. The first calculation step is to calculate the slope angle KK of the target point TC from the digital elevation model SHM. The second calculation part calculates the variance of the backscattering coefficient in the local region KR from the SAR images SG1 and SG2. - Calculate the difference in variance BA before and after the occurrence of a landslide disaster.
[0023] The determination unit HA has the following functions. The first step is to determine whether the area is a mountainous landslide disaster area (SAJ). - The second part of the judgment should be limited to the area surrounding the landslide disaster area SAJ in mountainous areas. The third judgment part is to judge whether the area is a landslide disaster area SIJ in an urban area.
[0024] The generation unit SE has the following functions. · To generate landslide disaster area distribution JB. Hardware Configuration of the First Embodiment FIG. 2 shows a hardware configuration of the landslide disaster area detection device CK of the first embodiment.
[0025] In order to perform the above-mentioned functions, the landslide disaster area detection device CK of embodiment 1 includes a processor PR, a memory ME, and a storage medium KI, as shown in FIG. 2, and may further include an input unit NY and an output unit SY, if necessary.
[0026] The processor PR is the well-known core of a computer that operates hardware according to software. The memory ME is composed of, for example, a dynamic random access memory (DRAM) and a static random access memory (SRAM). The storage medium KI is composed of, for example, a hard disk drive (HDD: Hard Disk Drive), a solid state drive (SSD: Solid State Drive), and a read only memory (ROM). The storage medium KI stores a program PRG. The program PRG is a group of instructions that specifies the content of the processing to be executed by the processor PR.
[0027] The input unit NY and the output unit SY are composed of, for example, an input interface and an output interface for transmitting input signals NS and output signals SS related to the operation of the processor PR to and from the outside of the landslide disaster area detection device CK.
[0028] Regarding the relationship between the functions and the hardware configuration in the landslide disaster area detection device CK, on the hardware, the processor PR executes the program PRG stored in the storage medium KI using the memory ME, and, as necessary, controls the operation of the input unit NY and the output unit SY to realize the functions of each unit from the acquisition unit SH to the generation unit SE (shown in Figure 1).
[0029] Operation of the First Embodiment FIG. 3 is a flowchart (part 1) showing the operation of the landslide disaster area detection device CK of the first embodiment.
[0030] FIG. 4 is a flowchart (part 2) showing the operation of the landslide disaster area detection device CK of the first embodiment.
[0031] The operation of the landslide disaster area detection device CK of the first embodiment will be described with reference to the flowcharts of FIGS.
[0032] Detection of landslide disaster areas in mountainous areas Step ST11: The acquisition unit SH (shown in FIG. 1) acquires a remote sensing image RG1 taken before the occurrence of the landslide disaster, and a remote sensing image RG2 taken after the occurrence of the landslide disaster.
[0033] Step ST12: The detection unit KE (shown in FIG. 1) judges whether or not each pixel in the remote sensing images RG1 and RG2 indicates a vegetation area, and detects the coordinates of the pixel indicating the vegetation area SR (not shown).
[0034] Land cover classification is the process of classifying the attributes of each pixel in remote sensing images RG1 and RG2 taken from the sky. Each pixel in remote sensing images RG1 and RG2 registered on a map contains information about where on the Earth it is. It is possible to investigate from remote sensing images RG1 and RG2 what objects (e.g., vegetation, soil, asphalt) are covering the surface of that point.
[0035] To detect the above-mentioned vegetation region SR, for example, in the case of a multispectral image, the NDVI (Normalized Difference Vegetation Index) shown in equation (1) can be used, using the reflectance characteristics obtained from the multispectral image. TIFF0007595804000001.tif15166 NIR: Spectral reflectance in the near infrared region Red: Spectral reflectance in the red range
[0036] A multispectral image is an image in which the reflectance spectrum characteristics, which have a continuous distribution with respect to wavelength, are decomposed into several wavelength bands, and image data is acquired for each wavelength band. With a multispectral image, it is possible to acquire the planar reflectance at multiple wavelengths for the same subject.
[0037] For the detection of the vegetation region SR described above, for example, in the case of a polarimetric SAR image, the Radar Thin-Vegetation Index (RVI) shown in equation (2) can be used. TIFF0007595804000002.tif15166 σ: Backscattering coefficient (antilogarithm)
[0038] The subscripts indicate the polarization direction of the transmitted and received radio waves, with H representing the horizontal direction and V representing the vertical direction. For example, σHV is the backscattering coefficient when a vertically polarized radio wave is received in response to a horizontally polarized transmitted radio wave.
[0039] A multi-polarized SAR image is a SAR image obtained when several polarized signals are acquired simultaneously.
[0040] Step ST13: The acquisition unit SH acquires a digital elevation model SHM of the area where the landslide disaster occurred.
[0041] Step ST14: The calculation unit SA (shown in FIG. 1) calculates the slope angle KK of the target point TC (not shown) by using the digital elevation model SHM.
[0042] Landslides occur when soil and sand flow from higher ground to lower ground, and therefore occur where the terrain is inclined. For the reasons mentioned above, the magnitude of the slope of the terrain is used as one of the criteria for determining whether or not a landslide has occurred at a point.
[0043] Step ST15: The judgment unit HA (shown in Figure 1) judges, based on the result of the judgment of the vegetation area SR in step ST12, whether or not there is a possibility that the target point TC captured in the remote sensing images RG1 and RG2 was a point that was in the vegetation area SR before the landslide disaster occurred and that is no longer in the vegetation area SR after the landslide disaster occurred, i.e., whether or not there is a possibility that the target point TC is a point that is no longer in the vegetation area SR due to the occurrence of the landslide disaster.
[0044] Step ST16: The determination unit HA determines, from the tilt angle KK calculated in step ST14, whether or not the tilt angle KK of the target point TC captured in the remote sensing images RG1 and RG2 exceeds a predetermined tilt angle threshold value KKth.
[0045] Step ST17: When the judgment unit HA judges in step ST15 that the target point TC may be a point that is no longer a vegetation area SR due to the occurrence of a landslide disaster and when the judgment unit HA judges in step ST16 that the slope angle KK of the target point TC exceeds the slope angle threshold KKth, the judgment unit HA judges that the target point TC is a landslide disaster area in a mountainous region.
[0046] Step ST18: In contrast to step ST17, the determining unit HA determines that the target point TC is not a landslide disaster area in a mountainous region.
[0047] Detection of landslide disaster areas in urban areas Step ST19: The acquisition unit SH acquires a SAR image SG1 taken before the landslide disaster occurs, and a SAR image SG2 taken after the landslide disaster occurs.
[0048] The locations and times photographed in the SAR images SG1 and SG2 are the same as the locations and times photographed in the remote sensing images RG1 and RG2 (obtained in step ST11).
[0049] Step ST20: The calculation unit SA calculates the variation BA (not shown) of the backscattering coefficient (antilogarithm) in the local region KR (not shown) based on the SAR images SG1 and SG2.
[0050] The variance in a local region in an image means the non-uniformity of pixel values (quantitatively expressed by various indices for evaluating statistical variance, such as variance, standard deviation, and standard deviation error) calculated from all pixel values when an area of M pixels vertical by N pixels horizontal is extracted. The calculation unit SA calculates the variance BA over the entire SAR images SG1 and SG2. The value of the variance BA in the local region KR at coordinates (i, j) in the image is calculated from the pixel values of M pixels vertical by N pixels horizontal including the coordinates (i, j).
[0051] The backscattering coefficient (antilogarithm) is the backscattering coefficient expressed as a real number greater than or equal to 0. The backscattering coefficient (antilogarithm) is generally expressed in decibels (dB) or as a real number greater than or equal to 0.
[0052] Step ST21: The calculation unit SA calculates the difference between the evaluation values of the backscattering coefficient variation BA before and after the occurrence of a disaster at the same location, i.e., the difference obtained by subtracting the evaluation value of the variation BA before the landslide disaster from the evaluation value of the variation BA after the landslide disaster.
[0053] In urban areas affected by landslides, the contrast of the SAR images SG1 and SG2, i.e., the spatial distribution of the backscattering coefficient, increases. In areas with high contrast in the SAR images SG1 and SG2, the pixel values fluctuate sharply due to differences in pixel positions. Therefore, by evaluating the variation BA within the local region KR, it is possible to determine whether or not the region is high contrast.
[0054] By evaluating the spatial variability BA of the backscattering coefficient in SAR images SG1 and SG2 acquired before and after the occurrence of a landslide disaster and comparing the magnitude of the evaluated values of the variability BA, it is possible to estimate the presence or absence of damage from landslide disasters in urban areas.
[0055] Step ST22: The judgment unit HA judges whether a difference evaluation value HCs (not shown), which is the difference between the evaluation values of the spatial variation BA of the backscattering coefficient before and after the occurrence of a landslide disaster, exceeds a predetermined threshold evaluation value HCth (not shown).
[0056] Step ST23: The judgment unit HA excludes the vegetation area SR covering both before and after the occurrence of the landslide disaster detected based on the remote sensing images RG1 and RG2 in step ST12 from the targets for which the presence or absence of a landslide disaster in urban areas should be detected using the SAR images SG1 and SG2 acquired in step ST19.
[0057] The reason for the above exclusion is that the landslide disaster area, which was vegetation area SR before the landslide disaster, was detected based on remote sensing images RG1 and RG2, and on the other hand, the vegetation area SR is susceptible to changes over time, which may lead to erroneous detection.
[0058] FIG. 5 shows the vicinity of a landslide disaster area in a mountainous region according to the first embodiment.
[0059] Step ST24: The judgment unit HA limits the area to the surrounding area SYR of the mountainous landslide disaster area SAJ detected based on the remote sensing images RG1 and RG2 in order to judge only changes in the evaluation values of the variance BA of the SAR images SG1 and SG2 outside the area where the landslide disaster occurred (changes that may occur due to various changes in the earth's surface).
[0060] The SAR images SG1 and SG2 are captured over a very wide area, and even a slight change, such as a change in the position of the vehicle, can cause the local variation BA to change. Therefore, as shown in Figure 5, the target area for detecting the landslide disaster area SIJ in the urban area is limited to the surrounding area SYR of the landslide disaster area SAJ in the mountain area.
[0061] As a specific method for limiting the area to the surrounding area SYR, for example, a smoothing process is performed on the image showing the landslide disaster area SAJ in the mountainous area to expand the area, and pixels whose pixel values exceed a certain value are defined as the surrounding area SYR (the closer a point is to the landslide disaster area SAJ in the mountainous area, the higher the pixel value after the smoothing process described above, so points far from the mountainous disaster area are excluded based on a predetermined threshold value.) In addition, for example, an area within a certain distance from the image of the landslide disaster area SAJ in the mountainous area is defined as the surrounding area SYR.
[0062] Step ST25: The determination unit HA determines whether or not the point was a vegetated ground point SC before the occurrence of the landslide disaster, and whether or not the point was a vegetated ground point SC after the occurrence of the landslide disaster.
[0063] Step ST26: When the judgment unit HA judges in step ST25 that the vegetation area point SC was not a vegetation area point SC before the landslide disaster occurred and that the vegetation area point SC was not a vegetation area point SC after the landslide disaster occurred, and when the judgment unit HA judges in step ST22 that the difference evaluation value HCs exceeds the threshold evaluation value HCth, it judges that the landslide disaster area SIJ is an urban landslide disaster area.
[0064] Step ST27: In contrast to step ST26, the determining unit HA determines that the area is not a landslide disaster area SIJ in an urban area.
[0065] FIG. 6 shows the relationship among the vegetation points SC, the difference evaluation value HCs, the threshold evaluation value HCth, and the landslide disaster area SIJ in the urban area in the first embodiment.
[0066] Figures 6A, 6B, 6C, and 6D respectively show vegetated areas SC before the landslide disaster occurred, vegetated areas SC after the landslide disaster occurred, points where the difference evaluation value HCs of the local variation BA of the radar backscattering coefficient increased to exceed the above-mentioned threshold evaluation value HCth, and points determined to be landslide disaster areas SIJ in urban areas.
[0067] FIG. 6A, for example, shows that pixel (1, 6) was a vegetation site SC before the landslide occurred.
[0068] FIG. 6B, for example, shows that pixel (4, 1) is a vegetation spot SC after the occurrence of a landslide disaster.
[0069] FIG. 6C shows, for example, that pixel (2, 1) is the point where the difference evaluation value HCs of the local variation BA of the radar backscattering coefficient has increased so much as to exceed the threshold evaluation value HCth.
[0070] FIG. 6D shows, for example, that pixel (6, 2) is determined to be a landslide disaster area SIJ in the urban area because (1) it was not a vegetation area SC before the landslide disaster occurred (as shown in FIG. 6A), (2) it is not a vegetation area SC after the landslide disaster occurred (as shown in FIG. 6B), and (3) it is a point where the difference evaluation value HCs of the local variation BA of the radar backscattering coefficient has increased to exceed the threshold evaluation value HCth (as shown in FIG. 6C).
[0071] Step ST28 (initial step ST23): The generation unit SE (shown in FIG. 1) generates both the landslide disaster area SAJ in mountainous areas and the landslide disaster area SIJ in urban areas as a landslide disaster area distribution JB.
[0072] Effects of the First Embodiment As described above, the landslide disaster area detection device CK of embodiment 1 makes it possible to detect areas affected by landslide disaster areas SIJ in urban areas with the same accuracy as that for detecting landslide disaster areas SAJ in mountainous areas.
[0073] Embodiment 2. Second Embodiment The landslide disaster area detection device CK of the second embodiment will be described.
[0074] The landslide disaster area detection device CK of the second embodiment utilizes the characteristic that in a landslide disaster, soil and sand flows down along the inclination direction of the slope from a higher place to a lower place. More specifically, the landslide disaster area detection device CK of the second embodiment reduces erroneous detection of a landslide disaster area SAJ in mountainous areas by performing morphological transformation using a kernel created from the inclination angle KK obtained from the digital elevation model SHM.
[0075] Functions of the Second Embodiment 7 is a functional block diagram of the landslide disaster area detection device CK of the embodiment 2. The functions of the landslide disaster area detection device CK of the embodiment 2 will be described with reference to FIG.
[0076] As shown in FIG. 7, the landslide disaster area detection device CK of the second embodiment includes an acquisition unit SH to a generation unit SE, similar to the landslide disaster area detection device CK of the first embodiment (shown in FIG. 1).
[0077] On the other hand, as shown in FIG. 7, the landslide disaster area detection device CK of the second embodiment differs from the landslide disaster area detection device CK of the first embodiment in that it further includes a conversion unit HE and a creation unit SS.
[0078] The conversion unit HE corresponds to the "first conversion unit" and the "second conversion unit."
[0079] The creation section SS corresponds to the "creation section."
[0080] The conversion unit HE has the following functions: The first transformation step is to convert the tilt angle at the target point TC into a tangent. The second transformation step is to perform morphological transformation of the landslide disaster area SAJ in mountainous areas.
[0081] The creation unit SS has the following functions. Create a kernel for morphological transformation according to the magnitude of the tangent and the direction of the slope at the target point TC.
[0082] Hardware Configuration of the Second Embodiment The hardware configuration of the landslide disaster area detection device CK of the second embodiment is similar to the hardware configuration of the landslide disaster area detection device CK of the first embodiment (shown in FIG. 2).
[0083] Operation of the Second Embodiment FIG. 8 is a flowchart (part 1) showing the operation of the landslide disaster area detection device CK of the second embodiment.
[0084] FIG. 9 is a flowchart (part 2) showing the operation of the landslide disaster area detection device CK of the second embodiment.
[0085] The operation of the landslide disaster area detection device CK of the second embodiment will be described with reference to the flowcharts of FIGS.
[0086] As is clear from a comparison of Figures 8 and 9, which show the operation of the landslide disaster area detection device CK of embodiment 2, with Figures 3 and 4, which show the operation of the landslide disaster area detection device CK of embodiment 1, the operation of the landslide disaster area detection device CK of embodiment 2 differs from the operation of the landslide disaster area detection device CK of embodiment 1 and further includes steps ST51 to ST53.
[0087] The following mainly describes the operations of steps ST51 to ST53 of the landslide disaster area detection device CK of the second embodiment.
[0088] FIG. 10 shows morphological transformation (part 1) using a kernel KN created from a slope angle KK obtained from the digital elevation model SHM of the second embodiment.
[0089] FIG. 11 shows morphological transformation (part 2) using a kernel KN created from a slope angle KK obtained from the digital elevation model SHM of the second embodiment.
[0090] Step ST51: The conversion unit HE (shown in FIG. 7) converts the inclination angle KK calculated from the digital elevation model (described in the first embodiment) into a tangent.
[0091] The tilt angle KK is calculated for the surrounding pixels. Here, the surrounding pixels refer to, for example, the central pixel of a 3×3 pixel (9 pixels) array, and the surrounding pixels refer to the 8 pixels excluding the central pixel.
[0092] Step ST52: The creation unit SS (shown in FIG. 7) creates a kernel KN for performing shrinking and expansion processing (morphological transformation) of the image shown in FIG. 11 in the downward slope direction shown in FIG. 10 with a width proportional to the tangent of the inclination angle KK.
[0093] The above creation by the creation unit SS utilizes the property that when comparing two points, the potential energy of the soil at the higher position increases in proportion to the tangent of the slope angle KK, and therefore the runoff distance RK when soil flows out increases in accordance with the tangent of the slope angle KK.
[0094] The direction of the pixel value "1" in Fig. 11 indicates the direction of the morphological transformation on the plane. The width of the kernel KN, which indicates how many pixels away from the center the pixel value "1" is located, indicates the length of the transformation, which indicates how many pixels are reduced or enlarged.
[0095] With respect to pixel (3, 3) shown in Fig. 10 as a reference, the altitude decreases downward, to the lower right, and to the right. As a result, as shown in Fig. 11, the kernel KN extends in the above three directions.
[0096] In FIG. 10, the degree of decrease in altitude from pixel (3, 3) in the lower right direction is large, that is, the absolute value of the tangent is large, so in FIG. 11, the kernel KN has a large width in the lower right direction.
[0097] Step ST53: The transformation unit HE performs morphological transformation on the landslide disaster area SAJ in the mountainous area (determined in step ST37) by using the morphological transformation kernel KN that reflects the above-mentioned tilt angle KK.
[0098] In addition to being used to deform shapes, morphological transformations are also used to remove noise. Specifically, erosion removes isolated pixels that have no neighboring pixels by shrinking them to less than one pixel. On the other hand, dilation removes tiny missed detections by filling them in by expanding the surrounding image.
[0099] Effects of the Second Embodiment As described above, in the landslide disaster area detection device CK of embodiment 2, by shrinking the landslide disaster area, it is possible to obtain the effect of removing detection results other than those along the slope, and on the other hand, by subsequently performing an expansion process using the exact same kernel KN, it is possible to restore the landslide disaster area that remains unremoved, i.e., the landslide disaster area along the slope, to its state before shrinkage.
[0100] The above-described embodiments may be combined with each other without departing from the gist of the present disclosure, and components in each embodiment may be appropriately deleted or modified, or other components may be added. [Industrial Applicability]
[0101] INDUSTRIAL APPLICABILITY The landslide disaster area detection device according to the present disclosure can be used to detect landslide disaster areas in urban areas with the same accuracy as that for detecting landslide disaster areas in mountainous regions. [Explanation of symbols]
[0102] 1 CK landslide disaster area detection device, HA judgment unit, HE conversion unit, JB landslide disaster area distribution, KE detection unit, RG remote sensing image, SA calculation unit, SAJ landslide disaster area, SE generation unit, SG SAR image, SH acquisition unit, SHM digital elevation model, SIJ landslide disaster area, SS creation unit.
Claims
1. A first acquisition unit that acquires remote sensing images before and after the occurrence of a landslide disaster; a first detection unit that detects a vegetation area from the remote sensing image; A second acquisition unit that acquires a digital elevation model of the area where the landslide disaster occurred; a first calculation unit that calculates a slope angle of a target point from the digital elevation model; a first determination unit that determines that the target point in the remote sensing image is a landslide disaster area in a mountainous region when the target point is a vegetated area before the occurrence of the landslide disaster, is a vegetated area that is no longer a vegetated area after the occurrence of the landslide disaster, and is a point whose slope angle exceeds a predetermined slope angle threshold; A third acquisition unit that acquires SAR images before and after the occurrence of the landslide disaster; a second calculation unit that calculates a variation in a backscattering coefficient in a local region from the SAR image; A third calculation unit that calculates a difference in the variation before and after the occurrence of the landslide disaster; A second determination unit that limits the area to the periphery of the landslide disaster area in the mountainous area; a third determination unit that determines that the area is a landslide disaster area in an urban area when the difference exceeds a predetermined threshold value and the area is not a vegetation area both before and after the occurrence of the landslide disaster; a generation unit for generating a landslide disaster area distribution based on the landslide disaster area in the mountainous area and the landslide disaster area in the urban area; A landslide disaster area detection device comprising:
2. A first conversion unit that converts the tilt angle at the target point into a tangent; a generation unit for generating a morphological transformation kernel according to the magnitude of the tangent and the direction of the gradient at the target point; A second transformation unit that performs morphological transformation on the landslide disaster area in the mountainous area; The landslide disaster area detection device according to claim 1 , further comprising:
3. The first acquisition unit acquires remote sensing images before and after the occurrence of a landslide disaster; The first detection unit detects a vegetation area from the remote sensing image; The second acquisition unit acquires a digital elevation model of the area where the landslide disaster occurred, The first calculation unit calculates a slope angle of the target point from the digital elevation model, a first determination unit determines that the target point in the remote sensing image is a landslide disaster area in a mountainous region when the target point is a vegetated area before the occurrence of the landslide disaster, is a vegetated area that is no longer a vegetated area after the occurrence of the landslide disaster, and is a point whose slope angle exceeds a predetermined slope angle threshold; a third acquisition unit that acquires SAR images before and after the occurrence of the landslide disaster; The second calculation unit calculates a variation in a backscattering coefficient in a local region from the SAR image, The third calculation unit calculates a difference in the variation before and after the occurrence of the landslide disaster, The second determination unit limits the area to the vicinity of the landslide disaster area in the mountainous area, a third determination unit that, when the difference exceeds a predetermined threshold and the area is not a vegetated area both before and after the occurrence of the landslide disaster, determines that the area is a landslide disaster area in an urban area; The generation unit generates a landslide disaster area distribution based on the landslide disaster area in the mountain area and the landslide disaster area in the urban area. A method for detecting landslide disaster areas.
4. The first conversion unit converts the tilt angle at the target point into a tangent; a generation unit generates a morphological transformation kernel according to the magnitude of the tangent and the direction of the gradient at the target point; The second transformation unit performs morphological transformation on the landslide disaster area in the mountainous area, The landslide disaster area detection method according to claim 3 .
5. When the difference in the variation of the backscattering coefficient in a local area from SAR images before and after the occurrence of a landslide disaster exceeds a predetermined threshold, it is determined that the area is a landslide disaster area in an urban area. Disaster area detection device.
Citation Information
Patent Citations
Multi-temporal multi-polarization SAR landslide extraction method
CN113408547A
Method for accurately identifying new landslide area based on multi-source time sequence remote sensing image
CN114241333A
Method for extracting unstable part of slope
JP2013221886A
Topography image creation method and topography image creation device
JP2013231843A
Landslide area detection device and program
JP2021056008A
