Geological disaster detection system based on aerial photography

Abstract

The invention belongs to the field of geological disaster detection, and discloses an aerial photography-based geological disaster detection system, which comprises an SAR image acquisition device, an SAR image processing device, an unmanned aerial vehicle device and a geological disaster detection device, the SAR image acquisition device is used for acquiring a first SAR image and a second SAR image of a geological disaster detection area, and the shooting time of the first SAR image is earlier than the shooting time of the second SAR image; the SAR image processing device is used for acquiring a shooting area according to the first SAR image and the second SAR image; the unmanned aerial vehicle device is used for shooting the shooting area to obtain a visible light image; the geological disaster detection device is used for identifying the visible light image and judging whether a preset type of geological disaster exists in the visible light image or not. The method can effectively shorten the average value of the time duration between the occurrence moments of all geological disaster events and the moments when the department responsible for detection obtains the detection results, and facilitates the more timely discovery of the areas where the geological disaster events occur.

Description

Technical Field

[0001] The present invention relates to the field of geological disaster detection, and in particular to a geological disaster detection system based on aerial photography. Background Art

[0002] Using drone aerial photography technology to detect geological disasters is an efficient detection method that can promptly identify areas where geological disasters such as landslides exist. However, when using drone technology to detect geological disasters, the existing technology generally divides the entire detection area into multiple small areas, and then uses drones to shoot all the small areas, and obtains the detection results based on the images obtained. However, after a geological disaster occurs, there is a long time between the start of the geological disaster event and the time when the department responsible for detection obtains the detection results in some small areas that are photographed later, resulting in the inability to timely discover these areas where geological disasters occur. Summary of the invention

[0003] The purpose of the present invention is to disclose a geological disaster detection system based on aerial photography to solve the technical problems raised in the background technology.

[0004] In order to achieve the above object, the present invention adopts the following technical solution:

[0005] The present invention provides a geological disaster detection system based on aerial photography, including a SAR image acquisition device, a SAR image processing device, an unmanned aerial vehicle device and a geological disaster detection device;

[0006] The SAR image acquisition device is used to acquire a first SAR image and a second SAR image of a geological disaster detection area, wherein the shooting time of the first SAR image is earlier than the shooting time of the second SAR image;

[0007] The SAR image processing device is used to obtain a shooting area according to the first SAR image and the second SAR image;

[0008] The drone device is used to photograph the shooting area and obtain visible light images;

[0009] The geological disaster detection device is used to identify the visible light image and determine whether there is a preset type of geological disaster in the visible light image;

[0010] The step of photographing the photographing area to obtain a visible light image includes:

[0011] Calculate the clustering value of each shooting area separately;

[0012] Based on the clustering values, a clustering algorithm is used to divide all clustering values ​​into two clusters;

[0013] Get the average value of the clustering values ​​in the two clusters respectively;

[0014] Use CA to represent the cluster with the larger average cluster value among the two clusters, and use CB to represent the other cluster;

[0015] First, the shooting area corresponding to the cluster value in CA is photographed to obtain a visible light image, and then the shooting area corresponding to the cluster value in CB is photographed to obtain a visible light image.

[0016] Preferably, the SAR image acquisition device comprises an earth observation remote sensing satellite carrying a synthetic aperture radar;

[0017] The earth observation remote sensing satellite equipped with synthetic aperture radar is used to image the geological disaster detection area through synthetic aperture radar to obtain the first SAR image and the second SAR image of the geological disaster detection area.

[0018] Preferably, acquiring the shooting area according to the first SAR image and the second SAR image includes:

[0019] Partitioning the second SAR image to obtain a plurality of areas to be determined;

[0020] Based on the first SAR image, it is judged whether each area to be judged meets the preset judgment condition, and the area to be judged that meets the preset judgment condition is used as the shooting area.

[0021] Preferably, partitioning the second SAR image to obtain a plurality of areas to be determined includes:

[0022] Calculate the zoning control coefficient;

[0023] The second SAR image is partitioned based on the partition control coefficient to obtain a plurality of areas to be determined.

[0024] Preferably, the calculation formula of the partition control coefficient is:

[0025]

[0026] ctridx represents the partition control coefficient of the second SAR image, scp represents the set of pixels in the second SAR image, nscp represents the total number of pixels in scp, and pixl i It represents the pixel value of pixel i in scp, ctrbs represents the preset control value, and baseval represents the preset reference value.

[0027] Preferably, partitioning the second SAR image based on the partition control coefficient to obtain a plurality of areas to be determined includes:

[0028] Use the following formula to calculate the length of the area to be judged:

[0029]

[0030] Use the following formula to calculate the width of the area to be judged:

[0031]

[0032] Wherein, Lth and Wth represent the length and width of the area to be determined, respectively, and Lbs and Wbs represent the preset length and width, respectively;

[0033] Dividing the second SAR image into a plurality of to-be-processed regions with a length of Lth and a width of Wth;

[0034] The to-be-judged region generation process is performed on the to-be-processed region to obtain the to-be-judged region.

[0035] Preferably, performing a process of generating a region to be determined on the region to be processed to obtain the region to be determined includes:

[0036] The first step is to take all the areas to be processed as elements of the set Udeal;

[0037] The second step is to randomly select a processing area from Udeal, use the extracted processing area as the comparison area, use the extracted processing area as an element in the set NU, and delete the extracted processing area from Udeal;

[0038] The third step is to calculate the adjacency value of each area to be processed that is adjacent to the comparison area and belongs to Udeal;

[0039] The fourth step is to determine whether the minimum adjacency value meets the set adjacency value threshold. If so, the area to be processed with the smallest adjacency value is deleted from Udeal, and the area to be processed with the smallest adjacency value is stored in NU. The area to be processed with the smallest adjacency value is used as the new comparison area and the process goes to the third step; if not, the process goes to the fifth step.

[0040] Step 5: Take the area corresponding to all elements in NU as an area to be judged;

[0041] Step 6: Clear the elements in NU and proceed to step 2.

[0042] Preferably, the adjacency value is obtained in the following manner:

[0043] cmp represents the comparison area, and Ucmp represents the set of all to-be-processed areas adjacent to cmp and belonging to Udeal;

[0044] Let deal represent the area to be processed, then the adjacency value corresponding to deal is:

[0045] neivaldeal =|avep cmp -avep deal |

[0046] neival deal Indicates the adjacency value corresponding to deal, avep cmp and avep deal They are the mean pixel values ​​of the pixels in cmp and deal respectively.

[0047] Preferably, judging whether each to-be-judged area meets a preset judgment condition based on the first SAR image includes:

[0048] The area to be judged is represented by jud2, and the area where all the pixels in the first SAR image with the same coordinates as the pixels in jud are located is represented by jud1;

[0049] Calculate the judgment value:

[0050]

[0051] jdevl jud2 Indicates the judgment value corresponding to jud2, NF jud2 Indicates the total number of pixels in jud2 that meet the following inequality:

[0052]

[0053] pixl j,2 Represents the pixel value of pixel j in jud2, pixl j,1 represents the pixel value of the pixel with the same coordinates as pixel j in the first SAR image, pixlm represents the maximum value of the pixel value of the pixel in the second SAR image, NL jud2 Represents the total number of pixels in jud2;

[0054] If jdevl jud2 If it is greater than the preset judgment value threshold, it means that jud2 meets the preset judgment condition.

[0055] Preferably, identifying the visible light image to determine whether there is a preset type of geological disaster in the visible light image includes:

[0056] The visible light image is input into a pre-trained recognition model for recognition to determine whether there is a preset type of geological disaster in the visible light image.

[0057] Beneficial effects:

[0058] In the process of using unmanned aerial photography technology to detect geological disasters, the present invention obtains the shooting area based on two SAR images with different shooting times, and then divides the clustering values ​​of the shooting area into two clusters, and preferentially shoots the shooting area corresponding to the cluster value in the cluster with a larger average value of the cluster value, thereby effectively reducing the number of shooting areas that need to be shot, while giving priority to shooting areas with a higher probability of geological disaster events occurring in all required shooting areas, which can effectively shorten the average value of the length of time between the occurrence time of all geological disaster events and the time when the department responsible for detection obtains the detection results, which is conducive to more timely discovery of areas where geological disaster events occur. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0060] Figure 1 The present invention is a schematic diagram of a geological disaster detection system based on aerial photography.

[0061] Figure 2 The diagram is a schematic diagram of the process of photographing the photographing area corresponding to the cluster value in CA to obtain a visible light image. DETAILED DESCRIPTION

[0062] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. The components of the embodiments of the present invention generally described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the claimed invention, but merely represents the selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making creative work belong to the scope of protection of the present invention.

[0063] like Figure 1 In one embodiment shown, the present invention provides a geological disaster detection system based on aerial photography, including a SAR image acquisition device, a SAR image processing device, a UAV device and a geological disaster detection device;

[0064] The SAR image acquisition device is used to acquire a first SAR image and a second SAR image of a geological disaster detection area, wherein the shooting time of the first SAR image is earlier than the shooting time of the second SAR image;

[0065] The SAR image processing device is used to obtain a shooting area according to the first SAR image and the second SAR image;

[0066] The drone device is used to photograph the shooting area and obtain visible light images;

[0067] The geological disaster detection device is used to identify the visible light image and determine whether there is a preset type of geological disaster in the visible light image;

[0068] The step of photographing the photographing area to obtain a visible light image includes:

[0069] Calculate the clustering value of each shooting area separately;

[0070] Based on the clustering values, a clustering algorithm is used to divide all clustering values ​​into two clusters;

[0071] Get the average value of the clustering values ​​in the two clusters respectively;

[0072] Use CA to represent the cluster with the larger average cluster value among the two clusters, and use CB to represent the other cluster;

[0073] First, the shooting area corresponding to the cluster value in CA is photographed to obtain a visible light image, and then the shooting area corresponding to the cluster value in CB is photographed to obtain a visible light image.

[0074] In the process of using drone aerial photography technology to detect geological disasters, the shooting area is acquired based on two SAR images with different shooting times, and then the clustering values ​​of the shooting area are divided into two clusters, and the shooting areas corresponding to the cluster values ​​in the cluster with a larger average cluster value are preferentially photographed. In this way, the number of shooting areas that need to be photographed can be effectively reduced, and the shooting areas with a higher probability of geological disaster events occurring in all the required shooting areas can be photographed preferentially. The average value of the length of time between the occurrence time of all geological disaster events and the time when the department responsible for detection obtains the detection results can be effectively shortened, which is conducive to more timely discovery of areas where geological disaster events occur.

[0075] In addition, the reason why the present invention does not directly detect geological disasters based on SAR images is that the pixel values ​​(including amplitude, phase, etc.) of the pixels in the SAR images are easily affected by various factors, which may cause changes in image quality and thus affect the accuracy of applications such as geological disaster monitoring and change detection. For example:

[0076] (1) Atmospheric factors:

[0077] Atmospheric humidity: The water vapor content in the atmosphere affects the propagation of SAR signals, causing phase delay and amplitude attenuation, which in turn affects the accuracy of pixel values. Especially in cloudy or humid areas, the propagation of radar signals may be significantly affected.

[0078] Air pressure and temperature changes: These atmospheric factors affect the propagation speed and attenuation of signals, which may lead to systematic errors, especially in image data processing over long time scales.

[0079] (2) Weather factors:

[0080] Precipitation (rain, snow, fog): Precipitation is an important factor affecting the quality of SAR images. Heavy precipitation can cause scattering and absorption of radar signals, thus affecting the intensity of pixels. Snow, fog and other weather conditions can also attenuate radar signals, resulting in reduced image quality.

[0081] Clouds: Although SAR has the ability to penetrate clouds, under certain extreme weather conditions, the thickness and type of clouds may have a certain impact on the signal, especially affecting the reflection intensity of ground objects.

[0082] Therefore, after acquiring the SAR image, the present invention uses a drone to perform secondary confirmation, so that the probability of misidentification can be greatly reduced.

[0083] Preferably, the SAR image acquisition device comprises an earth observation remote sensing satellite carrying a synthetic aperture radar;

[0084] The earth observation remote sensing satellite equipped with synthetic aperture radar is used to image the geological disaster detection area through synthetic aperture radar to obtain the first SAR image and the second SAR image of the geological disaster detection area.

[0085] The SAR image acquisition device of the present invention can image the geological disaster detection area in a fixed cycle, for example, once every week. Except for the shooting time, the other shooting parameters of each imaging are the same. Therefore, the second SAR image of the present invention can be the latest SAR image, and the first SAR image is another SAR image whose shooting time is closest to the shooting time of the second SAR image. At this time, the length of the shooting time between the two SAR images is equal to the fixed cycle.

[0086] Preferably, acquiring the shooting area according to the first SAR image and the second SAR image includes:

[0087] Partitioning the second SAR image to obtain a plurality of areas to be determined;

[0088] Based on the first SAR image, it is judged whether each area to be judged meets the preset judgment condition, and the area to be judged that meets the preset judgment condition is used as the shooting area.

[0089] By partitioning, it is helpful to screen out areas with a high probability of geological disaster events as shooting areas, thereby reducing the probability of shooting in areas where no geological disaster events exist, which is conducive to more timely discovery of areas where geological disasters exist.

[0090] Preferably, partitioning the second SAR image to obtain a plurality of areas to be determined includes:

[0091] Calculate the zoning control coefficient;

[0092] The second SAR image is partitioned based on the partition control coefficient to obtain a plurality of areas to be determined.

[0093] The partitioning process of the present invention is not based on fixed length and width, because such a partitioning method cannot be applied to SAR images of different states, which may easily lead to the area to be judged obtained after partitioning being too large or too small. If the area to be judged is too large, the result obtained in the subsequent judgment process of the shooting area will be inaccurate; if the area to be judged is too small, the number of shooting areas obtained will be too large, which will affect the efficiency of geological disaster identification in the address disaster detection area and will not be conducive to timely discovery of existing geological disasters.

[0094] Preferably, the calculation formula of the partition control coefficient is:

[0095]

[0096] ctridx represents the partition control coefficient of the second SAR image, scp represents the set of pixels in the second SAR image, nscp represents the total number of pixels in scp, and pixl i It represents the pixel value of pixel i in scp, ctrbs represents the preset control value, and baseval represents the preset reference value.

[0097] The partition control coefficient of the present invention is calculated based on the difference degree of pixel values ​​of all pixel points in the second SAR image. The greater the difference degree, the greater the probability of the existence of a geological disaster event in the second SAR image, and the larger the partition control coefficient, so that the length and width of the area to be judged calculated subsequently can be smaller, which is conducive to obtaining more accurate partition results and more comprehensively identifying areas where geological disaster events exist; the smaller the difference degree, the smaller the partition control coefficient, so that the length and width of the area to be judged calculated subsequently can be larger, which is conducive to reducing the number of areas to be judged, thereby reducing the number of shooting areas and improving the efficiency of obtaining visible light images.

[0098] In some embodiments, the preset control value may be 128.

[0099] In some embodiments, the preset reference value may be 10.

[0100] In some embodiments, the pixel value may be a grayscale value or a phase value.

[0101] Grayscale value:

[0102] Type: Grayscale images usually represent the grayscale level of amplitude information. SAR images can be displayed as visually understandable grayscale images after conversion. The grayscale value reflects the signal strength and is usually processed by logarithmic transformation of the amplitude (for example, taking the logarithm and expressing it in dB).

[0103] Value range: If logarithmically transformed (for example, in dB), the grayscale value range is generally 0 to 255 (8-bit image).

[0104] Phase value:

[0105] Type: The pixel value of the phase image represents the phase information of the echo signal, usually the phase part of the complex number. The phase value is used in interferometry (such as differential interferometry SAR, InSAR) and other applications to obtain surface deformation information.

[0106] Value range: The phase value range is -π to π (i.e. -180° to 180°). Because the phase is periodic, values ​​outside this range will be wrapped around.

[0107] Preferably, partitioning the second SAR image based on the partition control coefficient to obtain a plurality of areas to be determined includes:

[0108] Use the following formula to calculate the length of the area to be judged:

[0109]

[0110] Use the following formula to calculate the width of the area to be judged:

[0111]

[0112] Wherein, Lth and Wth represent the length and width of the area to be determined, respectively, and Lbs and Wbs represent the preset length and width, respectively;

[0113] Dividing the second SAR image into a plurality of to-be-processed regions with a length of Lth and a width of Wth;

[0114] The to-be-judged region generation process is performed on the to-be-processed region to obtain the to-be-judged region.

[0115] In the above formula, The floor symbol.

[0116] The length and width of the judgment area of ​​the present invention can be adaptively changed based on the change of the partition control coefficient, so that the present invention can obtain a shooting area whose length and width are more in line with actual shooting requirements.

[0117] In some embodiments, the preset length is one-fifth of the length of the image obtained when the drone takes pictures at the maximum flight altitude; the preset width is one-fifth of the width of the image obtained when the drone takes pictures at the maximum flight altitude.

[0118] Preferably, performing a process of generating a region to be determined on the region to be processed to obtain the region to be determined includes:

[0119] The first step is to take all the areas to be processed as elements of the set Udeal;

[0120] The second step is to randomly select a processing area from Udeal, use the extracted processing area as the comparison area, use the extracted processing area as an element in the set NU, and delete the extracted processing area from Udeal;

[0121] The third step is to calculate the adjacency value of each area to be processed that is adjacent to the comparison area and belongs to Udeal;

[0122] The fourth step is to determine whether the minimum adjacency value meets the set adjacency value threshold. If so, the area to be processed with the smallest adjacency value is deleted from Udeal, and the area to be processed with the smallest adjacency value is stored in NU. The area to be processed with the smallest adjacency value is used as the new comparison area and the process goes to the third step; if not, the process goes to the fifth step.

[0123] Step 5: Take the area corresponding to all elements in NU as an area to be judged;

[0124] Step 6: Clear the elements in NU and proceed to step 2.

[0125] The above-mentioned calculation process of the present invention mainly splices the areas to be processed with similar pixel value characteristics. In this way, the pixel value characteristic differences between the areas to be judged can be amplified, and the integrity of the areas where geological disasters occur in the areas to be judged can be improved, so that the areas with address disasters can be identified more accurately, which is beneficial to the accuracy of the recognition results of the subsequent recognition process of whether there is a preset type of geological disaster using an image recognition algorithm.

[0126] Preferably, the adjacency value is obtained in the following manner:

[0127] cmp represents the comparison area, and Ucmp represents the set of all to-be-processed areas adjacent to cmp and belonging to Udeal;

[0128] Let deal represent the area to be processed, then the adjacency value corresponding to deal is:

[0129] neival deal =|avep cmp -avep deal |

[0130] neival deal Indicates the adjacency value corresponding to deal, avep cmp and avep deal They are the mean pixel values ​​of the pixels in cmp and deal respectively.

[0131] In some embodiments, the adjacency value threshold is set to one tenth of the average value of the pixel values ​​of the pixel points in the second SAR image.

[0132] Preferably, judging whether each to-be-judged area meets a preset judgment condition based on the first SAR image includes:

[0133] The area to be judged is represented by jud2, and the area where all the pixels in the first SAR image with the same coordinates as the pixels in jud are located is represented by jud1;

[0134] Calculate the judgment value:

[0135]

[0136] jdevl jud2 Indicates the judgment value corresponding to jud2, NF jud2 Indicates the total number of pixels in jud2 that meet the following inequality:

[0137]

[0138] pixl j,2Represents the pixel value of pixel j in jud2, pixl j,1 represents the pixel value of the pixel with the same coordinates as pixel j in the first SAR image, pixlm represents the maximum value of the pixel value of the pixel in the second SAR image, NL jud2 Represents the total number of pixels in jud2;

[0139] If jdevl jud2 If it is greater than the preset judgment value threshold, it means that jud2 meets the preset judgment condition.

[0140] In the process of judging whether the area to be judged meets the preset judgment conditions, the present invention screens the pixel points in jud2 by setting an inequality, so that the larger the range of pixel points whose pixel values ​​in the second SAR image change more than those in the first SAR image, the larger the judgment value, thereby identifying the area with a high probability of geological disaster events as the shooting area.

[0141] In some embodiments, the preset judgment value threshold may be 0.05.

[0142] Preferably, identifying the visible light image to determine whether there is a preset type of geological disaster in the visible light image includes:

[0143] The visible light image is input into a pre-trained recognition model for recognition to determine whether there is a preset type of geological disaster in the visible light image.

[0144] In some embodiments, the pre-trained recognition model may be a CNN or a FCN.

[0145] CNN is good at automatically extracting features from images, and is particularly suitable for complex pattern recognition in SAR images. By training a deep convolutional network, CNN can learn high-level features of disaster areas (such as texture, morphological changes, etc.) and identify and segment disaster areas.

[0146] FCN is a network designed for pixel-level segmentation tasks, which is suitable for pixel-level segmentation of disaster areas in SAR images. It can accurately identify the boundaries of geological disaster areas and is widely used in the detection of geological disasters such as floods and landslides.

[0147] In some embodiments, the preset types of geological disasters include landslides, cracks, etc.

[0148] Preferably, the calculation formula of the clustering value is:

[0149]

[0150] clust sis the cluster value of the shooting area s, jdevl s is the judgment value corresponding to s, jdevl max is the maximum value of the judgment value corresponding to all shooting areas, area s represents the area of ​​s, area max Indicates the maximum value of the area of ​​all shooting areas, numoth s represents the total number of other shooting areas adjacent to s, numl represents the total number of all shooting areas, λ1, λ2 and λ3 are the judgment value influence weight, area influence weight and quantity influence weight respectively.

[0151] Since the judgment of whether the area to be judged meets the preset judgment conditions is based on the judgment value, the present invention uses the area to be judged whose judgment value is greater than the preset judgment value threshold as the shooting area. Therefore, each shooting area has a corresponding area to be judged, and the judgment value of the area to be judged is the judgment value of the corresponding shooting area.

[0152] The clustering value of the present invention is comprehensively considered from the aspects of judgment value, area and the state of adjacent areas. Therefore, if the judgment value of s is larger, the area is smaller, and the number of adjacent shooting areas is greater, the corresponding clustering value is larger. In this way, the probability of giving priority to shooting areas with a high probability of geological disaster events can be increased, while avoiding the shooting areas corresponding to the clustering values ​​in the cluster being too dispersed during subsequent clustering, thereby shortening the total flight distance and further improving the efficiency of the shooting process.

[0153] The present invention further classifies the shooting areas by clustering values. Therefore, it is possible to give priority to shooting areas with a high probability of geological disasters while improving the shooting efficiency as much as possible.

[0154] In some embodiments, the judgment value influence weight, the area influence weight, and the quantity influence weight may be 0.5, 0.3, and 0.2, respectively.

[0155] Preferably, if Figure 2 , photograph the photographing area corresponding to the cluster value in CA to obtain a visible light image, including:

[0156] The shooting areas corresponding to the cluster values ​​in CA are stored in the set SA;

[0157] Obtain the shortest path PA that traverses all shooting areas in PA;

[0158] All shooting areas in SA are photographed based on PA to obtain visible light images corresponding to all shooting areas in SA.

[0159] The shortest path can be found using Dijkstra's algorithm.

[0160] In some embodiments, when shooting, the drone device flies to the sky above the center of the area corresponding to the shooting area in the actual space, and shoots vertically downward to obtain a visible light image.

[0161] In addition, the method of photographing the photographing area corresponding to the clustering value in CB to obtain the visible light image is the same as the method of photographing the photographing area corresponding to the clustering value in CA to obtain the visible light image.

[0162] In some embodiments, photographing all photographing areas in SA based on PA to obtain visible light images corresponding to all photographing areas in SA includes:

[0163] According to PA, the system flies to each shooting area in SA in turn to shoot, and obtains the visible light image corresponding to each shooting area in SA respectively.

[0164] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. A geological disaster detection system based on aerial photography, characterized in that: It includes a SAR image acquisition device, a SAR image processing device, an unmanned aerial vehicle device and a geological disaster detection device; The SAR image acquisition device is used to acquire a first SAR image and a second SAR image of a geological disaster detection area, wherein the shooting time of the first SAR image is earlier than the shooting time of the second SAR image; The SAR image processing device is used to obtain a shooting area according to the first SAR image and the second SAR image; The drone device is used to photograph the shooting area and obtain visible light images; The geological disaster detection device is used to identify the visible light image and determine whether there is a preset type of geological disaster in the visible light image; The step of photographing the photographing area to obtain a visible light image includes: Calculate the clustering value of each shooting area separately; Based on the clustering values, a clustering algorithm is used to divide all clustering values ​​into two clusters; Get the average value of the clustering values ​​in the two clusters respectively; Use CA to represent the cluster with the larger average cluster value among the two clusters, and use CB to represent the other cluster; First, the shooting area corresponding to the cluster value in CA is photographed to obtain a visible light image, and then the shooting area corresponding to the cluster value in CB is photographed to obtain a visible light image.

2. A geological disaster detection system based on aerial photography according to claim 1, characterized in that: The SAR image acquisition device includes an earth observation remote sensing satellite carrying a synthetic aperture radar; The earth observation remote sensing satellite equipped with synthetic aperture radar is used to image the geological disaster detection area through synthetic aperture radar to obtain the first SAR image and the second SAR image of the geological disaster detection area.

3. A geological disaster detection system based on aerial photography according to claim 1, characterized in that: Acquiring a shooting area according to the first SAR image and the second SAR image includes: Partitioning the second SAR image to obtain a plurality of areas to be determined; Based on the first SAR image, it is judged whether each area to be judged meets the preset judgment condition, and the area to be judged that meets the preset judgment condition is used as the shooting area.

4. A geological disaster detection system based on aerial photography according to claim 3, characterized in that: The second SAR image is partitioned to obtain multiple areas to be judged, including: Calculate the zoning control coefficient; The second SAR image is partitioned based on the partition control coefficient to obtain a plurality of areas to be determined.

5. A geological disaster detection system based on aerial photography according to claim 4, characterized in that: The calculation formula of the partition control coefficient is: ctridx represents the partition control coefficient of the second SAR image, scp represents the set of pixels in the second SAR image, nscp represents the total number of pixels in scp, and pixl i It represents the pixel value of pixel i in scp, ctrbs represents the preset control value, and baseval represents the preset reference value.

6. A geological disaster detection system based on aerial photography according to claim 5, characterized in that: The second SAR image is partitioned based on the partition control coefficient to obtain multiple areas to be judged, including: Use the following formula to calculate the length of the area to be judged: Use the following formula to calculate the width of the area to be judged: Wherein, Lth and Wth represent the length and width of the area to be determined, respectively, and Lbs and Wbs represent the preset length and width, respectively; Dividing the second SAR image into a plurality of to-be-processed regions with a length of Lth and a width of Wth; The to-be-judged region generation process is performed on the to-be-processed region to obtain the to-be-judged region.

7. A geological disaster detection system based on aerial photography according to claim 6, characterized in that: The processing of generating the area to be judged is performed on the area to be processed to obtain the area to be judged, including: The first step is to take all the areas to be processed as elements of the set Udeal; The second step is to randomly select a processing area from Udeal, use the extracted processing area as the comparison area, use the extracted processing area as an element in the set NU, and delete the extracted processing area from Udeal; The third step is to calculate the adjacency value of each area to be processed that is adjacent to the comparison area and belongs to Udeal; The fourth step is to determine whether the minimum adjacency value meets the set adjacency value threshold. If so, the area to be processed with the smallest adjacency value is deleted from Udeal, and the area to be processed with the smallest adjacency value is stored in NU. The area to be processed with the smallest adjacency value is used as the new comparison area and the process goes to the third step; if not, the process goes to the fifth step. Step 5: Take the area corresponding to all elements in NU as an area to be judged; Step 6: Clear the elements in NU and proceed to step 2.

8. A geological disaster detection system based on aerial photography according to claim 7, characterized in that: The adjacency value is obtained as follows: cmp represents the comparison area, and Ucmp represents the set of all to-be-processed areas adjacent to cmp and belonging to Udeal; Let deal represent the area to be processed, then the adjacency value corresponding to deal is: neival deal =|avep cmp -avep deal | neival deal Indicates the adjacency value corresponding to deal, avep cmp and avep deal They are the mean pixel values ​​of the pixels in cmp and deal respectively.

9. The geological disaster detection system based on aerial photography according to claim 3 is characterized in that: Determining whether each area to be determined meets a preset determination condition based on the first SAR image includes: The area to be judged is represented by jud2, and the area where all the pixels in the first SAR image with the same coordinates as the pixels in jud are located is represented by jud1; Calculate the judgment value: jdevl jud2 Indicates the judgment value corresponding to jud2, NF jud2 Indicates the total number of pixels in jud2 that meet the following inequality: pixl j,2 Represents the pixel value of pixel j in jud2, pixl j,1 represents the pixel value of the pixel with the same coordinates as pixel j in the first SAR image, pixlm represents the maximum value of the pixel value of the pixel in the second SAR image, NL jud2 Represents the total number of pixels in jud2; If jdevl jud2 If it is greater than the preset judgment value threshold, it means that jud2 meets the preset judgment condition.

10. The geological disaster detection system based on aerial photography according to claim 1, characterized in that: Identify visible light images to determine whether there are preset types of geological disasters in the visible light images, including: The visible light image is input into a pre-trained recognition model for recognition to determine whether there is a preset type of geological disaster in the visible light image.

Images