Slope collapse location extraction system, slope collapse location extraction method, and slope collapse location extraction program

Abstract

To provide an extraction system of a slope collapse spot which can extract a collapse spot of a slope at a low cost while enhancing extraction accuracy.SOLUTION: An extraction system of a slope collapse spot comprises: an acquisition unit which is configured to acquire a visible light image of a collapse extraction region on a ground surface and a vegetation index map indicating the distribution of a vegetation index in the collapse extraction region; and an extraction unit which is configured to execute detection processing of detecting a low index region in which the vegetation index is smaller than a threshold in the vegetation index map from the collapse extraction region, and extraction processing of extracting a collapse spot of a slope with pattern matching to a determination object region including the low index region in the visible light image.SELECTED DRAWING: Figure 2

Description

[Technical Field]

[0001] The present disclosure relates to a system for extracting slope collapse areas, a method for extracting slope collapse areas, and a program for extracting slope collapse areas. [Background technology]

[0002] As a preventive measure against the inflow of debris flow onto railway tracks, a technique for extracting slopes that are likely to collapse using a vegetation index is known (see Patent Document 1). [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Application Laid-Open No. 2013-221886 Summary of the Invention [Problem to be solved by the invention]

[0004] The above-mentioned technology does not target the extraction of slopes where collapses have occurred. Therefore, to extract the location of a collapsed slope, visible light images of the ground surface must be taken from the air and then an expert must visually inspect the area. This approach requires a great deal of effort and cost.

[0005] On the other hand, it is possible to extract collapsed areas using data captured by satellites of microwaves reflected from the Earth's surface, but it is difficult to extract small-scale collapses. It is also possible to extract collapsed areas using optical images that capture sunlight reflected from the Earth's surface at high resolution, but the costs of the optical images themselves and the image processing are both enormous.

[0006] One aspect of the present disclosure aims to provide a system for extracting slope collapse areas that can extract slope collapse areas at low cost while improving extraction accuracy. [Means for solving the problem]

[0007] One aspect of the present disclosure is a system for extracting slope collapse locations, comprising: an acquisition unit configured to acquire a visible light image of a collapse extraction area on the ground surface and a vegetation index map showing the distribution of vegetation indices in the collapse extraction area; and an extraction unit configured to perform a detection process to detect, from the collapse extraction area, low-index areas in the vegetation index map where the vegetation index is lower than a threshold value; and an extraction process to extract collapse locations on the slope by pattern matching against a determination target area including the low-index areas in the visible light image.

[0008] This configuration allows for the extraction of slope collapse areas without relying on the visual inspection of an expert. Furthermore, since image analysis is performed after narrowing down the target area, it is possible to reduce the amount of calculation required for image analysis while minimizing the risk of missing collapsed areas. Therefore, it is possible to extract slope collapse areas at low cost while improving extraction accuracy.

[0009] In one aspect of the present disclosure, the extraction unit may be configured to estimate the direction of slope failure in the determination region of the vegetation index map before pattern matching for the determination region of the visible light image in the extraction process. With this configuration, the amount of calculation required for rotation in pattern matching in the visible light image can be reduced.

[0010] In one aspect of the present disclosure, the extraction unit may be configured to set a threshold value based on the distribution of vegetation indices in a vegetation index map during the detection process. This configuration allows the determination target area to be appropriately set taking into account seasonal changes in vegetation indices, unusual topography, and other disturbances. This reduces the amount of calculation and improves extraction accuracy.

[0011] One aspect of the present disclosure may further include a recording unit configured to record the results of past extraction processes as history data. The extraction unit may be configured to, during the extraction process, exclude from the slope collapse extraction results any areas that were previously extracted as slope collapse locations and that are recorded in the history data as not being slope collapse locations. This configuration can prevent areas that were erroneously extracted in the past from being extracted again, thereby improving the accuracy of slope collapse location extraction.

[0012] In one aspect of the present disclosure, the extraction unit may be configured to exclude areas where the slope angle of the ground obtained from the digital elevation model is less than a certain value from the extraction results of slope collapse locations. This configuration can prevent the extraction of collapses based on erroneous determinations on gently sloping ground surfaces, thereby improving the accuracy of extraction of collapse locations.

[0013] Another aspect of the present disclosure is a method for extracting slope collapse locations, comprising the steps of: acquiring a visible light image of a collapse extraction area on the ground surface and a vegetation index map showing the distribution of vegetation indices in the collapse extraction area; detecting, from the collapse extraction area, low-index areas in the vegetation index map where the vegetation index is smaller than a threshold; and extracting collapsed areas on the slope by pattern matching against a determination target area including the low-index areas in the visible light image.

[0014] Another aspect of the present disclosure is a slope collapse location extraction program that causes a computer to perform the following steps: acquire a visible light image of a collapse extraction area on the ground surface and a vegetation index map showing the distribution of vegetation indices in the collapse extraction area; detect low-index areas in the vegetation index map from within the collapse extraction area where the vegetation index is smaller than a threshold; and extract collapsed areas on the slope by pattern matching against a determination target area that includes the low-index areas in the visible light image.

[0015] With these configurations, it is possible to extract collapsed areas on slopes at low cost while improving extraction accuracy. [Brief explanation of the drawings]

[0016] [Figure 1] FIG. 1 is a block diagram showing a schematic configuration of a system for extracting slope failure locations according to an embodiment. [Figure 2] FIG. 2A is an example of a visible light image of a collapsed area, and FIG. 2B is an example of a vegetation index map of the collapsed area. [Figure 3] FIG. 3A is an example of a vegetation index map in summer, and FIG. 3B is an example of a vegetation index map in winter. [Figure 4] FIG. 4A is a graph showing an example of the distribution of vegetation indices in summer, and FIG. 4B is a graph showing an example of the distribution of vegetation indices in winter. [Figure 5] FIG. 5 is a schematic diagram showing the procedure for estimating the collapse direction in a vegetation index map. [Figure 6] FIG. 6A is an example of a pattern template, FIG. 6B is an example of a visible light image of a collapsed extracted region, and FIG. 6C is an example of an extraction result of a collapsed portion of a slope. [Figure 7] FIG. 7A is a schematic diagram showing an example of a procedure for comparing extracted collapsed locations with specific locations, and FIG. 7B is an example of a tilt angle map. [Figure 8] FIG. 8 is a flow diagram of a method for extracting collapsed portions of a slope in the embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0017] Hereinafter, embodiments to which the present disclosure is applied will be described with reference to the drawings. [1. First embodiment] [1-1. Slope collapse location extraction system] 1, a system 1 for extracting slope collapse locations (hereinafter simply referred to as "extraction system 1") automatically extracts collapsed slope locations on the ground surface. The extraction system 1 includes an information processing device 10 and an imaging device 20.

[0018] <Information processing device> The information processing device 10 is configured by a computer having, for example, a processor such as a CPU (Central Processing Unit), a recording medium such as a memory, and input / output devices such as a keyboard and a display.

[0019] The computer constituting the information processing device 10 acquires visible light images and vegetation index maps, detects low index areas, and extracts slope collapse areas using a slope collapse extraction program recorded on a recording medium. The information processing device 10 has an acquisition unit 11, an extraction unit 12, an output unit 13, and a recording unit 14.

[0020] (Acquisition Department) The acquisition unit 11 is configured to acquire a visible light image of a collapsed area on the ground surface and a vegetation index map indicating the distribution of vegetation indices in the collapsed area.

[0021] The acquisition unit 11 generates a visible light image and a vegetation index map based on image data of the earth's surface captured by the imaging device 20. The image data captured by the imaging device 20 is transmitted to the acquisition unit 11 via a network such as the Internet.

[0022] The imaging device 20 is a camera or sensor mounted on an artificial satellite. By using such a camera, it is possible to survey the Earth's surface from directly above. However, the imaging device 20 may also be a camera mounted on a flying object such as an aircraft or drone.

[0023] The acquisition unit 11 generates a visible light image as shown in Fig. 2A by combining the red wavelength data, green wavelength data, and blue wavelength data contained in the image data sent from the imaging device 20. The acquisition unit 11 also generates a vegetation index map as shown in Fig. 2B by combining the red wavelength data and near-infrared wavelength data contained in the image data sent from the imaging device 20.

[0024] The red wavelength data used in the visible light image and the red wavelength data used in the vegetation index map are the same data. In other words, the acquisition unit 11 generates a combination of a visible light image and a vegetation index map that includes the same ground surface area.

[0025] A vegetation index map is a map that shows the distribution of vegetation indices, which represent the state of vegetation on the earth's surface (for example, the amount of plants, activity, etc.). In the vegetation index map of Figure 2B, areas with a higher vegetation index are brighter and closer to white. Areas with a lower vegetation index are darker and closer to black. The magnitude of the vegetation index at each point (i.e., each pixel) is determined by the brightness on the vegetation index map.

[0026] In the example of Figure 2B, the vegetation index of the black area in the center is lower than that of the other areas, so the white area in the center of the visible light image in Figure 2A, which corresponds to the vegetation index map in Figure 2B, is determined to be an area with few vegetation.

[0027] In this embodiment, the vegetation index that constitutes the vegetation index map is the NDVI (Normalized Difference Vegetation Index). However, the vegetation index may also be the GNDVI (Green Normalized Difference Vegetation Index). When the GNDVI is used, green wavelength data is used instead of red wavelength data when creating the vegetation index map.

[0028] (Extraction part) The extraction unit 12 is configured to extract collapsed portions of a slope using the visible light image and the vegetation index map generated by the acquisition unit 11. The extraction unit 12 executes a detection process and an extraction process.

[0029] In the detection process, the extraction unit 12 detects low-index areas in the vegetation index map where the vegetation index is lower than a threshold from among the collapsed extracted areas included in the visible light image and the vegetation index map. In the example of Figure 2B, the black area in the center is detected as a low-index area. A low-index area is an area where there is little vegetation and it is predicted that the slope may be collapsing.

[0030] As shown in Figures 3A and 3B, the vegetation index changes with the seasons. Figure 3A is a vegetation index map of a certain area in summer, and Figure 3B is a vegetation index map of the same area in winter. As is clear from the figures, the map is darker overall in winter than in summer. In other words, the vegetation index is generally lower in winter than in summer.

[0031] Furthermore, the brightness of the vegetation index map may change due to the influence of the position of the sun and the presence of clouds depending on the timing of image capture by the image capture device 20. Furthermore, the vegetation index may contain abnormal values ​​due to special topography such as pond P shown in Figure 3A or the presence of buildings.

[0032] Therefore, if the threshold for detecting low-index regions is set to a fixed value, the low-index regions may not be detected properly due to the influence of the above-mentioned disturbances, which may result in an increase in the amount of calculation, incorrect extraction of collapsed areas, or failure to extract collapsed areas.

[0033] Therefore, in the detection process, the extraction unit 12 sets a threshold value based on the distribution of vegetation indices in the vegetation index map. Hereinafter, a method for setting a threshold value by the extraction unit 12 will be specifically described.

[0034] First, the extraction unit 12 creates a distribution map of vegetation indices as shown in Fig. 4A. The horizontal axis of Fig. 4A represents the vegetation index, and the vertical axis represents the number of pixels (i.e., frequency). Fig. 4A shows the distribution of vegetation indices in a certain area in summer.

[0035] The extraction unit 12 obtains a truncated normal distribution TND from this distribution map by cutting out the top 3% of the entire distribution and cutting out the lower region that is a distance D or more away from the mode M. The distance D is the horizontal distance from the mode M to the top 3%. In other words, the width of the truncated normal distribution TND is twice the distance from the mode M to the top 3%. The top 3% of the entire distribution corresponds to outliers.

[0036] Next, the extraction unit 12 calculates the mean value A1 of the truncated normal distribution TND and an adjustment value C1=α×σ by multiplying the standard deviation σ of the truncated normal distribution TND by a constant α. The extraction unit 12 calculates the threshold value T1 (=A1−C1) by subtracting the adjustment value C1 from the mean value A1. The constant α is a positive number, for example, 1.5, that is set recursively based on experience.

[0037] 4B is a distribution map of the vegetation index in winter, when the vegetation index is generally lower than in summer. In the distribution map of FIG. 4B, the extraction unit 12 calculates the threshold value T2 (=A2-C2) from the average value A2 and the adjustment value C2 of the truncated normal distribution TND, as in FIG. 4A.

[0038] The threshold value T2 for winter in Fig. 4B is smaller than the threshold value T1 for summer in Fig. 4A. Therefore, for example, a location that is not determined to be a low index area in summer (i.e., an area where plants exist) can be prevented from being determined to be a low index area (i.e., an area where no plants exist) when the amount or activity of plants decreases in winter.

[0039] Subsequently, in the extraction process, the extraction unit 12 estimates the direction of collapse of the slope in the determination target region of the vegetation index map before pattern matching with the determination target region of the visible light image, which will be described later.

[0040] The determination target area is an area including a low index area and its surroundings within the collapsed area included in the visible light image and the vegetation index map. The determination target area is set at a plurality of locations within the collapsed area.

[0041] The extraction unit 12 detects a low index area LA from within the determination target area JA of the vegetation index map, as shown in Fig. 5. In the detected low index area LA, the extraction unit 12 counts the number of low index pixels whose vegetation index is smaller than the above-mentioned threshold in both the vertical and horizontal directions.

[0042] In Figure 5, the number of low-index pixels in the left-right direction (i.e., in each row) is shown to the right of the determination target area JA, and the number of low-index pixels in the up-down direction (i.e., in each column) is shown below the determination target area JA.

[0043] In the example of Figure 5, the number of low-index pixels in each row increases rapidly near the center. Also, the number of low-index pixels in each column increases gradually toward the right. From this result, the extraction unit 12 infers that the direction of the collapse is the direction indicated by the arrow in Figure 5.

[0044] Specifically, the extraction unit 12 calculates the gradient magnitude (i.e., the rate of change in pixel value of the low-index pixel) and the gradient direction (i.e., the direction of change in pixel value of the low-index pixel value) for each pixel in the low-index region LA, and creates a gradient histogram. The extraction unit 12 determines the peak of the created gradient histogram as the direction of collapse (i.e., the representative direction) of the low-index region LA.

[0045] After estimating the direction of the collapse, the extraction unit 12 extracts the collapsed part of the slope by pattern matching with the determination target area in the visible light image. Specifically, the extraction unit 12 extracts the collapsed part by superimposing pre-recorded template patterns of multiple collapsed parts on the determination target area while rotating and scaling them.

[0046] 6A shows an example of a template pattern of a collapsed portion. The pattern of the collapsed portion includes information on the direction of collapse in addition to the shape of the collapsed portion. The extraction unit 12 applies the template pattern to each of a plurality of determination target regions included in the visible light image shown in FIG. 6B.

[0047] When applying the template pattern, the extraction unit 12 rotates the template pattern in each determination target area so that it matches the direction of the collapse estimated on the vegetation index map. As shown in Figure 6C, the extraction unit 12 extracts the area where the template pattern matches as the slope collapse location (the area surrounded by a rectangle in Figure 6C).

[0048] In the extraction process, the extraction unit 12 excludes from the extraction result of slope collapse locations, among areas that have been extracted in the past as slope collapse locations, specific locations that are recorded in the history data of the recording unit 14 as not being slope collapse locations. The history data of the recording unit 14 includes records of whether or not the collapse locations extracted by the extraction unit 12 were actually collapse locations as a result of visual judgment by a person (i.e., an operator). In other words, the extraction unit 12 excludes from the extraction result, among the areas to be determined, areas that the extraction unit 12 has mistakenly detected as collapse locations in the past.

[0049] The determination of whether the collapsed portion extracted by the extraction unit 12 is the same as the specific portion recorded in the history data is performed, for example, by the following procedure: As shown in Fig. 7A, the extraction unit 12 reads out, from the history data, a rectangle that circumscribes the range in which the template pattern was rotated and enlarged when the specific portion was extracted as the collapsed portion, as a first frame F1, and overlays it on the visible light image.

[0050] Furthermore, the extraction unit 12 overlays a rectangle circumscribing the range in which the template pattern was rotated and enlarged when the currently extracted collapsed portion as a second frame F2 on the visible light image. If the center of gravity G1 of the first frame F1 is inside the second frame F2 and the center of gravity G2 of the second frame F2 is inside the first frame F1, the extraction unit 12 determines that the collapsed portion extracted by the extraction unit 12 is the same as the specific portion recorded in the history data.

[0051] Furthermore, in the extraction process, the extraction unit 12 excludes areas where the inclination angle of the ground surface obtained from the digital elevation model is less than a certain value from the extraction results of slope collapse locations. The digital elevation model is composed of elevation data for each cell obtained by dividing the ground surface into a certain mesh (for example, 5 m).

[0052] The extraction unit 12 creates the tilt angle map shown in Fig. 7B from the difference in elevation between adjacent cells of the digital elevation model. In the tilt angle map of Fig. 7B, the darker the color of the cell, the larger the tilt angle.

[0053] If an extracted collapsed portion includes an exclusion area with a tilt angle less than a certain value, the extraction unit 12 excludes the collapsed portion from the extraction result. The threshold value of the tilt angle for determining the exclusion area is, for example, 30°. In other words, the extraction unit 12 excludes areas with a tilt angle less than 30°.

[0054] (output section) The output unit 13 is configured to output the extraction result of the extraction unit 12. The output unit 13 may be a display that displays the extraction result. Alternatively, the output unit 13 may be a transmitter that transmits the extraction result to a recording medium, a control device, a computer, a terminal, or the like external to the information processing device 10.

[0055] The operator visually checks the collapsed areas of the slope included in the extraction results output from the output unit 13 to see if there is any collapse.

[0056] (Recording Department) The recording unit 14 is configured to record the results of past extraction processes performed by the extraction unit 12 and the results of visual inspections by workers as history data.

[0057] [1-2. How to extract slope collapse locations] The method for extracting slope collapse locations shown in Fig. 8 includes an acquisition step S10, a detection step S20, an extraction step S30, and an output step S40. The method for extracting slope collapse locations of this embodiment can be performed using, for example, the slope collapse location extraction system 1 shown in Fig. 1.

[0058] <Acquisition process> In this step, a visible light image of the collapsed area on the ground surface and a vegetation index map showing the distribution of vegetation indices in the collapsed area are obtained.

[0059] <Detection process> In this step, a low index area in which the vegetation index is smaller than a threshold value is detected from the collapsed area in the vegetation index map.

[0060] <Extraction process> In this step, collapsed areas of the slope are extracted by pattern matching on the determination target area including the low index area in the visible light image.

[0061] <Output process> In this process, the extracted collapsed areas of the slope are output, and the output results are provided to the worker who will visually inspect the area.

[0062] [1-3.Effects] According to the embodiment described above in detail, the following effects can be obtained. (1a) Slope collapse areas can be extracted without relying on the visual inspection of an expert. In addition, image analysis is performed after narrowing down the target area, which reduces the amount of calculation required for image analysis while preventing failure to extract collapsed areas. As a result, slope collapse areas can be extracted at low cost while improving extraction accuracy.

[0063] (1b) By estimating the direction of slope collapse in the vegetation index map, the amount of calculation required for rotation can be reduced in pattern matching in visible light images.

[0064] (1c) By setting the threshold for detecting low-index areas based on the distribution of vegetation indices, it is possible to appropriately set the judgment target area taking into account disturbances such as seasonal changes in vegetation indices and special topography. This reduces the amount of calculation and improves extraction accuracy.

[0065] (1d) By excluding areas that are recorded in the historical data as not being slope collapse areas from the results of slope collapse area extraction, it is possible to prevent areas that were mistakenly extracted in the past from being extracted again, thereby improving the accuracy of slope collapse area extraction.

[0066] (1e) By excluding areas where the surface slope angle obtained from the digital elevation model is less than a certain value from the results of detecting slope collapses, it is possible to prevent erroneous detection of collapses on gently sloping surfaces, thereby improving the accuracy of detecting collapsed areas.

[0067] 2. Other Embodiments Although the embodiments of the present disclosure have been described above, it goes without saying that the present disclosure is not limited to the above-described embodiments and can take on various forms.

[0068] (2a) In the system and method for extracting slope collapse locations according to the above embodiment, it is not essential to estimate the direction of slope collapse in the vegetation index map. Therefore, pattern matching in the visible light image may be performed without estimating the direction of slope collapse.

[0069] (2b) In the system and method for extracting slope failure locations according to the above embodiment, the threshold for detecting low index areas does not necessarily have to be set based on the distribution of vegetation indexes. This threshold may be set based on information other than the distribution of vegetation indexes.

[0070] (2c) The function of one component in the above embodiments may be distributed among multiple components, or the functions of multiple components may be integrated into one component. Also, part of the configuration of the above embodiments may be omitted. Furthermore, at least part of the configuration of the above embodiments may be added to or substituted for the configuration of another of the above embodiments. All aspects included in the technical idea identified by the wording of the claims are embodiments of the present disclosure. [Explanation of symbols]

[0071] 1...slope collapse location extraction system, 10...information processing device, 11...acquisition unit, 12...extraction unit, 13...output unit, 14...recording unit, 20...imaging device.

Claims

1. an acquisition unit configured to acquire a visible light image of a collapsed area on the ground surface and a vegetation index map showing a distribution of vegetation indices in the collapsed area; an extraction unit configured to execute a detection process of detecting an area in the vegetation index map where the vegetation index is smaller than a threshold value as a low-index area in the visible light image, and an extraction process of extracting a collapsed portion of a slope by pattern matching with a determination target area including the low-index area in the visible light image; A slope collapse location extraction system comprising:

2. The slope collapse location extraction system according to claim 1, A system for extracting slope collapse locations, wherein the extraction unit is configured to estimate the direction of slope collapse in the determination target area of ​​the vegetation index map before pattern matching the determination target area of ​​the visible light image in the extraction process.

3. The system for extracting a slope collapse location according to claim 1 or 2, A system for extracting slope collapse locations, wherein the extraction unit is configured to set the threshold value based on the distribution of vegetation indices in the vegetation index map during the detection process.

4. The system for extracting a slope failure location according to any one of claims 1 to 3, a recording unit configured to record past results of the extraction process as history data; The extraction unit is configured to exclude from the extraction results of slope collapse areas any areas that have been previously extracted as slope collapse areas but that are recorded in the historical data as not being slope collapse areas.

5. The system for extracting a slope failure location according to any one of claims 1 to 4, The extraction unit is configured to exclude areas where the slope angle of the earth's surface obtained from the digital elevation model is less than a certain value from the extraction results of slope collapse areas.

6. A method for extracting slope failure locations executed by a computer, comprising: acquiring a visible light image of a collapsed area on the ground surface and a vegetation index map showing a distribution of vegetation indices in the collapsed area; detecting an area in the vegetation index map where the vegetation index is smaller than a threshold as a low-index area in the visible light image; extracting a collapsed portion of the slope by pattern matching with a determination target region including the low index region in the visible light image; A method for extracting slope collapse locations, comprising:

7. Obtaining a visible light image of a collapsed area on the ground surface and a vegetation index map showing a distribution of vegetation indices in the collapsed area; detecting an area in the vegetation index map where the vegetation index is smaller than a threshold as a low-index area in the visible light image; extracting a collapsed portion of the slope by pattern matching with a determination target region including the low index region in the visible light image; A slope collapse extraction program that runs the following on a computer.

Images