Landslide monitoring and early warning method based on unmanned aerial vehicle photography and pixel analysis

Abstract

The invention provides a landslide monitoring and early warning method based on unmanned aerial vehicle photography and pixel analysis, and belongs to the technical field of geological disaster monitoring and early warning. Comprising the following steps: S1, geological survey; s2, arranging a marker; s3, collecting original point data; s4, collecting periodic data; s5, calculating plane displacement; s6, plane displacement is converted into space displacement; s7, analyzing displacement data; s8, performing early warning on a monitoring result to obtain an average displacement value a of the landslide mass marker, an average value b of components of the marker displacing along the main direction, an azimuth angle alpha and a pitch angle beta of the main direction, and an average angle theta deviating from the main direction; the destructive power of the landslide is strongest in the main direction and gradually weakened away from the main direction; the landslide risk is judged according to the analysis of the monitoring data, and early warning is given in time. According to the invention, the problems of ground disaster monitoring limitation and difficult quantification are effectively solved.

Description

TECHNICAL FIELD

[0001] The application provides a landslide monitoring and early warning method based on unmanned aerial vehicle photography and pixel analysis, and belongs to the technical field of geological disaster monitoring and early warning. BACKGROUND

[0002] There are many monitoring and early warning methods for landslide geological disasters, mainly including the following methods:

[0003] Ground deformation monitoring: observation piles, inclinometers and the like are arranged in the landslide body and the surrounding area, and displacement, settlement and other data are measured regularly, so that the movement trend of the landslide body can be directly understood. Leveling can accurately obtain elevation changes to determine whether there is subsidence; a total station can monitor horizontal displacement to analyze the sliding direction and speed of the landslide body.

[0004] Underground deformation monitoring: with the help of a borehole inclinometer, the deformation of soil or rock at different depths is monitored in the underground to understand the position and development of the potential sliding surface.

[0005] Underground water monitoring: water level observation wells are arranged in the landslide area to monitor changes in underground water level. Rising water level increases the weight of the soil and reduces the shear strength, which may trigger a landslide. Changes in water quality can also be analyzed to infer the erosion and softening effect of underground water activity on the rock-soil body.

[0006] Satellite remote sensing and aerial photogrammetry: satellite remote sensing can obtain information of a large area, and potential landslide signs such as surface deformation and vegetation changes can be analyzed by comparing multi-temporal images. Aerial photogrammetry can provide high-resolution images, stereoscopic observation and interpretation of topography, and identification of early micro-topographic features of landslides.

[0007] These methods generally have the problems of limited range and low accuracy. Deformation monitoring methods or underground water level monitoring methods can only quantitatively study local deformation and cannot be extended to the entire landslide body. Satellite remote sensing and aerial photogrammetry can only be analyzed in a plane and cannot be extended to three-dimensional space. In addition, they are difficult to quantitatively analyze the overall motion of the landslide body. SUMMARY

[0008] The application provides a landslide monitoring and early warning method based on unmanned aerial vehicle photography and pixel analysis. On the basis of traditional geological investigation, the displacement of each region of the entire landslide body is quantitatively analyzed at a millimeter level by the method of unmanned aerial vehicle photography and image analysis. Finally, the motion-related parameters of the entire landslide body are calculated by fitting into a three-dimensional coordinate system, and the warning value is given, effectively solving the problems of limited geological disaster monitoring and difficulty in quantitative analysis.

[0009] The specific technical solution is as follows:

[0010] A landslide monitoring and early warning method based on unmanned aerial vehicle photography and pixel analysis, comprising the following steps:

[0011] S1. Geological survey;

[0012] Geological survey includes two aspects: one is to identify landslide bodies, and the other is to determine the landslide perimeter, and then determine the range of landslide bodies in the study area.

[0013] (1) Landslide body identification

[0014] Landslide body identification is carried out through the following geological survey methods:

[0015] ① Topography

[0016] Mountain slope shape: If there are circle chair-shaped, horseshoe-shaped topography on the mountain slope, or there are sudden changes in slope, it may be a potential landslide.

[0017] Gully features: Pay attention to whether the gully has double-gully homology, gully wall has obvious scratches or terrace discontinuity phenomenon, if exists, it may have potential landslide.

[0018] Abnormal slope surface: Local pits, drum hills appear on the slope surface, or there are a large number of loose accumulations, which may indicate potential landslide risk.

[0019] ② Rock-soil mass characteristics

[0020] Rock fragmentation: If the rock joint fissure is developed, broken seriously, or the soil is loose, the particle size is mixed, the stability is poor, and the landslide is easy to occur.

[0021] Stratigraphic dislocation: Observe that the stratum has obvious dislocation, distortion, and new-old stratum inversion phenomenon, which may be a sign of potential landslide.

[0022] ③ Surface and vegetation

[0023] Surface cracks: Parallel or arc-shaped cracks appear on the surface of the mountain slope, especially when tensile cracks and shear cracks exist at the same time, which may be a potential landslide.

[0024] Abnormal vegetation: Trees are tilted and show "drunkard forest" and "sword tree" shapes, or the growth of vegetation is significantly different, which may be related to potential landslide.

[0025] (2) Landslide perimeter determination

[0026] The main methods for identifying landslide perimeter are as follows:

[0027] ① Geological and geomorphological identification method

[0028] Macroscopic geomorphic features: From a distance, if the slope has relatively low concave circle chair-shaped, horseshoe-shaped topography, or double-gully homology phenomenon, there may be a landslide, and the boundary of circle chair-shaped or horseshoe-shaped is often the landslide perimeter.

[0029] Micro-landform features: Close-up view, the back edge of the landslide often forms a concave or landslide lake, both sides have shear feather cracks, the front edge has bulging and longitudinal cracks, protruding landslide tongue, the boundaries of these special landforms can be determined as the perimeter of the landslide.

[0030] ②Stratigraphic lithology identification method

[0031] Sequence and occurrence: The strata of the landslide body are often disturbed during sliding, the sequence is chaotic, the structure is loose, and the occurrence may be discontinuous compared with the non-sliding slope section, and the old and new strata may be inverted abnormally, which can be used to divide the perimeter of the landslide.

[0032] Rock-soil mass characteristics: The rock-soil mass of the landslide body will be loose and broken due to sliding, which is obviously different from the surrounding complete rock-soil mass, and can be used as a basis for identifying the perimeter of the landslide.

[0033] ③Surface crack identification method

[0034] Crack morphology: The back edge of the landslide is generally a tensile crack, in the form of a circle chair; the side boundary is a shear feather crack; the front edge is a bulging and longitudinal crack. Tracking the extension range of these cracks can roughly determine the perimeter of the landslide.

[0035] Crack development: Observing whether the cracks have a lengthening and widening development trend, and the expansion direction of the newly appeared cracks or the original cracks can also help determine the expansion range and perimeter of the landslide.

[0036] S2. Marker arrangement

[0037] According to the range of the landslide body, 20-50 unequal grid points are arranged at equal intervals, uniformly distributed in each direction of the landslide body. A marker is placed at each of these points, and the markers are numbered according to the S-shaped rule from 1 to n. The horizontal direction of the grid point is consistent with the slope direction, and the vertical direction is perpendicular to the slope direction. The marker is composed of a small locator and a red ball on it, and the displacement of the red ball is consistent with the surface displacement of the location.

[0038] S3. Original point data collection

[0039] An unmanned aerial vehicle equipped with RTK and laser radar is used to take pictures of each marker, and the process is as follows:

[0040] ① On-site measurement, a rectangular area is framed according to the pixels with the marker as the center, and the long side is parallel to the horizontal plane. Mark the four corners of the rectangular area.

[0041] ② A vertical line is drawn through the center point of the rectangular area where the marker is located, perpendicular to the rectangular face.

[0042] ③ Raise the UAV to a certain point on the vertical line, the body azimuth angle is consistent with the strike of the inclined plane where the rectangle is located, and the inclination is opposite. Adjust the camera angle of view so that the camera focal point is directly opposite the center of the red ball. The camera angle of view β is complementary to the inclination angle of the inclined plane. Record the body azimuth angle α and the angle of view β.

[0043] ④ Adjust the position of the UAV along the vertical line direction, or adjust the camera magnification, so that the camera frame just frames the four corners of the rectangle. Record the geodetic coordinates and elevation of the UAV at this time, as well as the camera magnification a. Use the laser radar to measure the distance from the UAV to the marker.

[0044] ⑤ Take a photo. At this time, one pixel of the photo corresponds to a 1mm x 1mm planar space.

[0045] S4. Periodic data collection

[0046] The monitoring of the landslide body requires periodic data collection, with a period of one month.

[0047] After a period, according to the order of the marker numbers, take a second photo of each point. According to the original recorded UAV three-dimensional coordinates of each point, body azimuth and camera azimuth, camera magnification data, position the UAV to the same position as the original point and the camera view angle.

[0048] S5. Planar displacement calculation

[0049] The displacement of a certain marker is the displacement of the center of the red ball in the second photo relative to the center of the red ball in the first photo. Calculate the pixel value of the displacement of the center of the red ball in the rectangle inclined plane, and the displacement size can be calculated.

[0050] Get the pixel displacement value dx parallel to the strike and dy perpendicular to the strike of the red ball on the rectangle inclined plane. The planar displacement vector is (dx, dy), and the actual displacement size is

[0051] S6. Convert planar displacement to spatial displacement

[0052] Different markers on the landslide body in a local plane need to be converted into three-dimensional spatial displacement vectors.

[0053] Set the position of the UAV shooting point corresponding to a certain marker as the origin O, establish a three-dimensional coordinate system, with the north-south direction as the x-axis, the east-west direction as the y-axis, and the up-down direction as the z-axis. The marker is located at point P, and the rectangle inclined plane is A, with a strike of α and an inclination of 90°-β. The distance between P and O is r, the line connecting the two points is perpendicular to the plane A, and the azimuth of O pointing to P is α and the angle of view is β. P displaces x along the strike direction and y along the direction perpendicular to the strike direction. The conversion process is as follows:

[0054] ①Use the relationship between spatial trigonometric functions to solve the original coordinates of point P (x0, y0, z0). Where x0 = rsinαcosβ, y0 = rsinαsinβ, z0 = rcosβ.

[0055] ②The direction of the plane normal vector passes through point P and points to the origin O.

[0056] ③Unit vector of the slope direction

[0057] The direction of the slope is α, and the corresponding vector on the horizontal plane is

[0058] The inclination angle of the inclined plane is θ = 90° - β. Rotate the horizontal vector to the inclined plane, and we have:

[0059]

[0060] Unitization because So we have:

[0061]

[0062] ④Unit vector in vertical direction

[0063] Constructed on a horizontal plane Perpendicular vector

[0064] Rotate it onto the slope.

[0065] Unitization because So we have:

[0066]

[0067] ⑤Calculate three-dimensional spatial displacement:

[0068] It is known that point P is displaced by x along the strike direction and by y along the direction perpendicular to the strike direction within the inclined plane.

[0069] Then its displacement in three-dimensional space (△x, △y, △z) is:

[0070]

[0071] S7. Displacement Data Analysis

[0072] The displacement vector (a, b, c) of n markers in three-dimensional space is calculated, the average of the modulus of these displacement vectors is solved to represent the degree of landslide movement, and the main direction of landslide displacement is fitted using these displacement vectors for further analysis. The process is as follows:

[0073] ① Displacement vector of a certain marker Then the modulus of the vector is

[0074] The value is calculated for n vectors, and then the average is taken.

[0075] ② The covariance matrix of n vectors is calculated, and the eigenvalues and eigenvectors are solved. The direction of the eigenvector corresponding to the largest eigenvalue is the fitted main direction. Let the main direction vector be Then the azimuth angle α and the pitch angle β values can be calculated by the formula: α = arctan2(y, x),

[0076] ③ Displacement vector of a certain marker The modulus of the vector projected in the main direction is This value is calculated for all vectors, and then the average is taken.

[0077] ④ Displacement vector of a certain marker Given the main direction vector Then the angle between the two vectors is This value is calculated for all vectors, and then the average is taken.

[0078] S8. Monitoring results warning

[0079] The average displacement value a of the landslide markers, the average value b of the component of the marker displacement along the main direction, the azimuth angle α and the pitch angle β of the main direction, and the average angle θ deviating from the main direction are obtained. The destructive force of the landslide is strongest along the main direction and gradually weakens away from the main direction.

[0080] The average displacement value a of the landslide markers, the azimuth angle α and the pitch angle β of the main direction, the average value b of the component of the marker displacement along the main direction, and the average angle θ deviating from the main direction are obtained.

[0081] When a < 20 mm, the landslide body is relatively stable, and the surface rock-soil layer undergoes normal displacement changes under the influence of wind and water conditions; when 20 ≤ a < 80 mm, relative movement has occurred within the landslide body, accompanied by the appearance of cracks, and the landslide front is significantly deformed, indicating a large landslide hazard, and residents within the influence range of the landslide area need to be relocated; when a ≥ 80 mm, the landslide body has a significant landslide hazard and may slide at any time, and residents must be relocated immediately.

[0082] When 0 < b < 20 mm, the landslide body has a hidden danger; when b is greater than or equal to 20 mm, it has a major safety hazard and residents need to be immediately relocated.

[0083] When θ < 10°, the landslide body slides in one direction as a whole, the expansion angle of the two sides is not more than 20°, and the linear sliding occurs, at this time, the potential disaster area is the smallest, but the destructive power is the largest; when 10° < θ < 30°, the landslide body slides in a small fan shape, the expansion angle of the two sides is between 20° and 60°, at this time, the potential disaster area is medium, and the destructive power is medium; when θ > 30°, the landslide body slides in a large fan shape, the expansion angle of the two sides is greater than 60°, at this time, the potential disaster area is the largest, and the destructive power is the weakest.

[0084] According to the analysis of the above monitoring data, the landslide risk is judged, and early warning is made in time. BRIEF DESCRIPTION OF DRAWINGS

[0085] Figure 1 It is a schematic diagram of a landslide body;

[0086] Figure 2 It is a schematic diagram of a marker arrangement of the present application;

[0087] Figure 3 It is a schematic diagram of converting the plane displacement into the space displacement of the present application. DETAILED DESCRIPTION

[0088] The specific technical solutions of the present application are described in combination with the embodiments.

[0089] A landslide monitoring and early warning method based on unmanned aerial vehicle photography and pixel analysis, comprising the following steps:

[0090] S1. Geological survey

[0091] The purpose of the geological survey has two aspects, one is to identify the landslide body, and the other is to determine the landslide perimeter, and then to determine the range of the landslide body in the study area.

[0092] (1) Landslide body identification

[0093] The landslide body identification can be carried out by the following traditional geological survey methods:

[0094] ① Topography

[0095] Mountain slope shape: If there are circle chair-shaped, horseshoe-shaped topography on the mountain slope, or there are sudden changes in slope, the potential landslide body may exist.

[0096] Gully features: Pay attention to whether the gully has double-gully homology, gully wall has obvious scratches or terrace discontinuity, if there are, there may be a potential landslide.

[0097] Abnormal slope: Local pits, mounds, or large amounts of loose deposits on the slope surface may indicate potential landslide risk.

[0098] ②Rock and soil characteristics

[0099] Rock fragmentation: If the rock joints and fissures are well developed, severely fragmented, or the soil is loose and the particle size is mixed, the stability is poor, and the landslide is easy to occur.

[0100] Stratigraphic disturbance: Observe the obvious disturbance, distortion, and inversion of old and new strata, which may be a sign of potential landslide.

[0101] ③Surface and vegetation

[0102] Surface cracks: Parallel or arc-shaped cracks appear on the surface of the slope, especially when tensile and shear cracks exist simultaneously, which may be a potential landslide.

[0103] Abnormal vegetation: Trees are tilted and shaped like "drunkard forest" and "sword tree", or the growth of vegetation is significantly different, such as sparse or dead vegetation in some areas, which may be related to potential landslides.

[0104] (2) Landslide perimeter determination

[0105] As shown in Figure 1 , the identification methods of landslide perimeter mainly include the following:

[0106] ① Geological and geomorphic identification method

[0107] Macro-landform features: From a distance, if the slope has a relatively low concave circle chair shape, horseshoe-shaped landform, or double-gully coexistence phenomenon, there may be a landslide, and the boundary of the circle chair or horseshoe shape is often the landslide perimeter.

[0108] Micro-landform features: Close-up view, the back edge of the landslide often forms a concave land or landslide lake, with shear feather-shaped cracks on both sides, and bulging and longitudinal cracks, protruding landslide tongue on the front edge, etc. The boundary of these special landforms can be determined as the landslide perimeter.

[0109] ② Stratigraphic lithology identification method

[0110] Sequence and occurrence: The strata of the landslide body are often disturbed during sliding, with chaotic sequence, loose structure, and discontinuous occurrence compared to the non-sliding slope segment, with abnormal conditions such as inversion of old and new strata, which can be used to divide the landslide perimeter.

[0111] Rock and soil characteristics: The rock and soil of the landslide body will be loose and fragmented due to sliding, which is significantly different from the surrounding intact rock and soil, and can be used as a basis for identifying the landslide perimeter.

[0112] ③ Surface crack identification method

[0113] Crack morphology: the trailing edge of landslide is generally a tensile crack, which is in the shape of a circle chair; the side boundary is a shear feather crack; the front edge is a bulging and longitudinal crack. Tracking the extension of these cracks can roughly determine the perimeter of the landslide.

[0114] Crack development: observing whether the cracks have a trend of lengthening, widening, etc. The extension direction of newly appeared cracks or original cracks can also help determine the extension range and perimeter of the landslide.

[0115] S2. Marker arrangement

[0116] According to the range of landslide body, 20-50 unequal grid points are arranged at equal intervals and uniformly distributed in each direction of the landslide body. A marker is placed at each of these points. The markers are numbered from 1 to n according to the S-shaped rule. The horizontal direction of the grid points is consistent with the slope direction, and the vertical direction is perpendicular to the slope direction. The marker is composed of a small locator and a red ball on it. The displacement of the red ball is consistent with the surface displacement of the location, as shown in Figure 2 .

[0117] S3. Original point data collection

[0118] An unmanned aerial vehicle equipped with RTK and laser radar is used to take pictures of each marker. Assuming that the lens resolution is 8k, i.e. 7680x4320 pixels, the process is as follows:

[0119] ① On-site measurement: frame a 7680mmx4320mm rectangular area with the marker as the center, with the long side parallel to the horizontal plane. Mark the four corners of the rectangular area.

[0120] ② Draw a vertical line through the center point of the rectangular area where the marker is located.

[0121] ③ Raise the unmanned aerial vehicle and fly to a certain point on the vertical line. The body azimuth angle is consistent with the slope direction, and the inclination is opposite. Adjust the lens viewing angle so that the lens focal point is directly opposite the center of the red ball. The lens viewing angle β is complementary to the slope inclination. Record the body azimuth angle α and the viewing angle β.

[0122] ④ Adjust the position of the unmanned aerial vehicle along the vertical line direction, or adjust the lens magnification, so that the lens frame just frames the four corners of the rectangle. Record the geodetic coordinates and elevation of the unmanned aerial vehicle at this time using RTK, and the lens magnification a. Measure the distance from the unmanned aerial vehicle to the marker using laser radar.

[0123] ⑤ Take a picture. At this time, one pixel of the picture corresponds to a 1mmx1mm planar space.

[0124] S4. Periodic data collection

[0125] The monitoring of landslide body needs to be carried out periodically, and the period is usually one month.

[0126] After one cycle, take a second photo of each point in the order of the marker's serial number. Based on the original record of each point's UAV three-dimensional coordinates, body azimuth angle, lens azimuth angle, lens magnification, and other data, the UAV can be positioned one by one to the same position and lens view angle as the original point. The photo taken in this way is consistent with the first picture in terms of view angle and size.

[0127] S5. Plane displacement calculation

[0128] The displacement of a certain marker is the displacement of the red ball center in the second photo relative to the red ball center in the first photo. Only the displacement of the red ball center point in the rectangular slope needs to be calculated in pixel value to calculate the displacement size.

[0129] The OpenCV library of Python can be used to solve the red ball displacement pixel value. The following is the implementation idea and example code:

[0130] Implementation idea

[0131] Read image: use cv2.imread to read two photos.

[0132] Color space conversion: convert the image from BGR to HSV for more effective color recognition.

[0133] Color threshold segmentation: set threshold according to the color of the red ball in HSV space to segment the red ball area and get the mask.

[0134] Contour detection: find the contour in the mask to find the red ball contour.

[0135] Calculate the center coordinates: calculate the center coordinates of the red ball contour as the red ball position.

[0136] Calculate displacement: calculate the difference between the center coordinates of the red ball in the two pictures to get the displacement pixel value.

[0137] Example code

[0138] import cv2

[0139] import numpy as np

[0140] def find_ball_center(image):

[0141] hsv = cv2.cvtColor(image, cv2.COLOR_BGR2HSV)

[0142] lower_red = np.array([0, 100, 100])

[0143] upper_red = np.array([10, 255, 255])

[0144] mask1 = cv2.inRange(hsv, lower_red, upper_red)

[0145] lower_red = np.array([160, 100, 100])

[0146] upper_red = np.array([180, 255, 255])

[0147] mask2 = cv2.inRange(hsv, lower_red, upper_red)

[0148] mask = mask1 + mask2

[0149] contours, _ = cv2.findContours(mask, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)

[0150] if not contours:

[0151] return None

[0152] largest_contour = max(contours, key=cv2.contourArea)

[0153] M = cv2.moments(largest_contour)

[0154] if M["m00"] == 0:

[0155] return None

[0156] cX = int(M["m10"] / M["m00"])

[0157] cY = int(M["m01"] / M["m00"])

[0158]

[0159] The above process obtains the pixel displacement value dx parallel to the direction of the red ball on the inclined plane and the pixel displacement value dy perpendicular to the direction of the inclined plane. The plane displacement vector is (dx, dy), and the actual displacement size is

[0160] S6. Conversion of plane displacement into spatial displacement

[0161] Different markers on the landslide body are on a local plane. To coordinate the displacement vectors of the corresponding planes of each marker, they need to be converted into three-dimensional space displacement vectors.

[0162] ,like Figure 3 , set the position of a marker corresponding to the drone shooting point as the origin O, and establish three-dimensional spatial coordinates, with north and south as the x-axis, east and west as the y-axis, and up and down as the z-axis. The marker is located at point P, the rectangular slope is A, the slope direction is α, and the inclination angle is 90°-β. It is known that the distance from point P to point O is r, the line connecting the two points is perpendicular to plane A, the azimuth angle from point O to point P is α, and the downward angle is β. P is displaced by x along the strike direction in the plane and by y along the perpendicular strike direction. The conversion process is as follows:

[0163] ①Use the relationship between spatial trigonometric functions to solve the original coordinates of point P (x0, y0, z0). Where x0 = rsinαcosβ, y0 = rsinαsinβ, z0 = rcosβ.

[0164] ②The direction of the plane normal vector passes through point P and points to the origin O. ③Unit vector of the slope direction The direction of the slope is α, and the corresponding vector on the horizontal plane is The inclination angle of the inclined plane is θ = 90° - β. Rotate the horizontal vector onto the inclined plane, and we have:

[0165]

[0166] Unitization because So there is

[0167] ④Unit vector in vertical direction Constructed on a horizontal plane Perpendicular vector Rotate it onto the slope. Unitization because So there is

[0168]

[0169] ⑤Calculate three-dimensional spatial displacement:

[0170] It is known that point P is displaced by x along the strike direction and by y along the direction perpendicular to the strike direction within the inclined plane.

[0171] Then its displacement in three-dimensional space (△x, △y, △z) is:

[0172]

[0173] The above process can be implemented by Python.

[0174]

[0175]

[0176] S7. Displacement data analysis

[0177] The above calculation obtains the displacement vector (a, b, c) of n markers in three-dimensional space, the average value of the modulus of these displacement vectors can be solved to represent the degree of landslide sliding, and the main direction of landslide displacement is fitted with these displacement vectors for further analysis. The process is as follows:

[0178] ① The displacement vector of a certain marker Then the modulus of the vector is

[0179] Calculate this value for n vectors, and then take the average.

[0180] ② Calculate the covariance matrix of n vectors, and solve its eigenvalues and eigenvectors. The direction of the eigenvector corresponding to the largest eigenvalue is the fitted main direction. Let the main direction vector be Then the azimuth angle α and the pitch angle β value can be calculated by the formula: α = arctan2 (y, x),

[0181] ③ The displacement vector of a certain marker The modulus of the vector projected in the main direction is Calculate this value for all vectors, and then take the average.

[0182] ④ The displacement vector of a certain marker Given the main direction vector Then the angle between the two vectors is Calculate this value for all vectors, and then take the average.

[0183] The above steps can be implemented by Python language:

[0184]

[0185]

[0186] S8. Monitoring result early warning

[0187] The above process yields the average displacement a of the landslide marker, the average b of the marker displacement components along the main direction, the azimuth α and elevation β of the main direction, and the average angle θ of the deviation from the main direction. The destructive force of a landslide is strongest along the main direction and gradually weakens away from it.

[0188] In the above process, the average displacement value a of the landslide marker, the azimuth angle α and pitch angle β of the main direction, the average value b of the components of the marker displacement along the main direction, and the average angle θ of the deviation from the main direction are obtained.

[0189] When a is less than 20 mm, the landslide body is relatively stable, and the surface rock and soil layers undergo normal displacement changes under natural conditions such as feng shui. When 20 ≤ a < 80 mm, relative movement has occurred in the landslide body, accompanied by the appearance of cracks and obvious deformation of the landslide front edge. There is a greater landslide hazard, and residents within the affected area of ​​the landslide need to be relocated. When a ≥ 80 mm, the landslide body has a major landslide hazard and may collapse at any time, and residents must be relocated immediately.

[0190] When 0<b<20mm, the landslide body poses a safety hazard; when b≥20mm, it poses a major safety hazard and residents need to be evacuated immediately.

[0191] When θ is less than 10°, the landslide body slides in one direction as a whole, and the expansion angle of the boundaries on both sides does not exceed 20°, showing linear sliding. At this time, the potential disaster area is the smallest, but the destructive power is the greatest; when 10°<θ<30°, the landslide body slides in a small fan-shaped shape, and the expansion angle of the boundaries on both sides is between 20° and 60°. At this time, the potential disaster area is medium and the destructive power is medium; when θ is greater than 30°, the landslide body slides in a large fan-shaped shape, and the expansion angle of the boundaries on both sides is greater than 60°. At this time, the potential disaster area is the largest and the destructive power is the weakest.

[0192] Based on the analysis of the above monitoring data, the landslide risk is judged and timely warnings are issued.

Claims

1. A landslide monitoring and early warning method based on drone photography and pixel analysis, characterized in that: The following steps are involved: S1. Geological survey; The geological survey includes identification of the landslide body, secondly, determination of the landslide perimeter and thus the extent of the landslide body within the study area; S2. Marker placement; According to the range of the landslide body, grid points are set at equal intervals and evenly distributed in all directions of the landslide body; Subcontract a marker at these points; The marker consists of a small locator and a red ball on it; S3. Original point data collection; A drone equipped with RTK and LiDAR was used to take photos of each marker; S4. Periodic data collection; After one cycle, take a second photo of each point in the order of the marker numbers; and position the drone one by one to the same position and camera angle as the original point. S5. Plane displacement calculation; The displacement of a marker is the displacement of the center of the red ball in the second photo relative to the center of the red ball in the first photo. The displacement is calculated by calculating the pixel value of the displacement of the center of the red ball within the inclined plane of the rectangle. S6. Converting plane displacement into spatial displacement; Different markers on the landslide body are on a local plane. The displacement vectors of the corresponding planes of each marker should be integrated and converted into three-dimensional space displacement vectors. S7. Displacement data analysis; Calculate the displacement vectors (a, b, c) of n markers in three-dimensional space, calculate the average value of the modulus of these displacement vectors to represent the sliding degree of the landslide, and use these displacement vectors to fit the main direction of the landslide displacement; S8. Early warning of monitoring results; The average displacement value a of the landslide marker, the average value b of the marker displacement components along the main direction, the azimuth angle α and pitch angle β of the main direction, and the average angle θ of deviation from the main direction are obtained; the destructive force of the landslide is strongest along the main direction and gradually weakens away from the main direction; the landslide risk is judged based on the analysis of the monitoring data and timely warnings are issued.

2. The landslide monitoring and early warning method based on drone photography and pixel analysis according to claim 1 is characterized in that: In S1, the landslide perimeter determination methods include the following methods: ①Geological and geomorphological identification method Macro-geomorphological features: Observed from a distance, if the slope has relatively concave armchair-shaped or horseshoe-shaped landforms, or has double gullies with the same source, a landslide may occur. The armchair-shaped or horseshoe-shaped boundary is often the landslide perimeter. Micro-geomorphological features: Close inspection reveals that the landslide trailing edge often forms a depression or landslide lake, with shear feather cracks on both sides, and bulges and longitudinal cracks and a prominent landslide tongue on the leading edge. The boundaries of these special landforms are determined as the landslide perimeter. ② Stratigraphic lithology identification method Sequence and occurrence: The strata of a landslide are often disturbed during the sliding process, resulting in a disordered sequence and loose structure. Compared with the non-sliding slope section, the occurrence may also be discontinuous, with abnormal inversion of new and old strata. This is used to delineate the landslide perimeter. Rock and soil characteristics: The rock and soil of the landslide body will be loose and broken due to sliding, which is obviously different from the surrounding intact rock and soil. This serves as the basis for identifying the landslide perimeter; ③Surface crack identification method Crack morphology: The rear edge of a landslide is generally characterized by tensile cracks, shaped like a round chair; the lateral boundaries are characterized by shear feather cracks; and the leading edge is characterized by bulging and longitudinal cracks. By tracing the extent of these cracks, the landslide perimeter can be roughly determined. Crack development: Observe whether the cracks have a trend of lengthening and widening, and the expansion direction of new cracks or existing cracks, which can also help determine the expansion range and perimeter of the landslide.

3. The landslide monitoring and early warning method based on drone photography and pixel analysis according to claim 1 is characterized in that: The S3 specific method is: ① On-site measurement: With the marker as the center, draw a rectangular area based on the pixels, with the long side parallel to the horizontal plane; mark the four corners of the rectangular area; ② Draw a perpendicular line through the center point of the rectangle where the marker is located; ③ Raise the drone and fly it to a point on the vertical line. The azimuth angle of the aircraft is aligned with the slope of the rectangle, and their inclinations are opposite. Adjust the camera's downward angle so that the lens focus is directly on the center of the red ball. The downward angle β of the lens is complementary to the inclination angle of the slope. Record the aircraft's azimuth angle α and the downward angle β. ④ Adjust the drone's position along the vertical line, or adjust the lens magnification so that the lens image just frames the four corners of the rectangle; use RTK to record the drone's geodetic coordinates and elevation at this time, and the lens magnification a; use LiDAR to measure the distance from the drone to the landmark; ⑤Take a photo. At this time, 1 pixel in the photo corresponds to a 1mm×1mm plane space.

4. The landslide monitoring and early warning method based on drone photography and pixel analysis according to claim 1 is characterized in that: The specific method of converting S6 into a three-dimensional space displacement vector is: Set the position of a marker corresponding to the drone's shooting point as the origin O, and establish three-dimensional spatial coordinates, with north and south as the x-axis, east and west as the y-axis, and up and down as the z-axis; the marker is located at point P, the rectangular slope is A, the slope strike is α, and the inclination angle is 90°-β; the distance from point P to point O is r, the line connecting the two points is perpendicular to plane A, the azimuth angle from point O to point P is α, and the downward angle is β; P is displaced by x along the strike direction and y along the perpendicular strike direction within the plane; the conversion process is as follows: ①Use the relationship between spatial trigonometric functions to solve the original coordinates of point P (x0, y0, z0); where x0 = rsinαcosβ, y0 = rsinαsinβ, z0 = rcosβ; ②The direction of the plane normal vector passes through point P and points to the origin O. ③Unit vector of the slope direction The direction of the slope is α, and the corresponding vector on the horizontal plane is The inclination angle of the inclined plane is θ = 90° - β. Rotate the horizontal vector to the inclined plane, and we have: Unitization because have: ④Unit vector in vertical direction Constructed on a horizontal plane Perpendicular vector Rotate it onto the slope. Unitization because have: ⑤Calculate three-dimensional spatial displacement: It is known that point P is displaced by x along the strike direction and by y along the perpendicular strike direction within the inclined plane; Then its displacement in three-dimensional space (△x, △y, △z) is:

5. The landslide monitoring and early warning method based on drone photography and pixel analysis according to claim 1 is characterized in that: The specific process of S7 is as follows: ① Displacement vector of a marker Then the magnitude of the vector is Calculate this value for n vectors and then find the average; ② Calculate the covariance matrix of n vectors and solve their eigenvalues ​​and eigenvectors; the direction of the eigenvector corresponding to the maximum eigenvalue is the main direction of the fitting; let the main direction vector be After that, the azimuth angle α and pitch angle β are calculated by the formula: α=arctan2(y,x), ③ Displacement vector of a marker The magnitude of the projected vector in the principal direction is: Calculate this value for all vectors and then find the average; ④ Displacement vector of a marker The main direction vector is known to be The angle between the two vectors is This value is calculated for all vectors and then averaged.

6. The landslide monitoring and early warning method based on drone photography and pixel analysis according to claim 1 is characterized in that: In S8, the average displacement value a of the landslide marker, the azimuth angle α and the pitch angle β of the main direction, the average value b of the components of the marker displacement along the main direction, and the average angle θ of the deviation from the main direction are obtained; When a is less than 20 mm, the landslide body is relatively stable, and the surface rock and soil layers are subject to normal displacement changes under natural conditions of Feng Shui. When 20 ≤ a < 80 mm, relative movement has occurred within the landslide body, accompanied by the appearance of cracks and obvious deformation of the landslide front. There is a significant landslide hazard, and residents within the affected area need to be relocated. When a ≥ 80 mm, the landslide body has a major landslide hazard and may collapse at any time, and residents must be relocated immediately. When 0 < b < 20 mm, the landslide body poses a safety hazard; when b ≥ 20 mm, it poses a major safety hazard and residents must be evacuated immediately; When θ is less than 10°, the landslide body slides in one direction as a whole, and the expansion angle of the two side boundaries does not exceed 20°, showing linear sliding. At this time, the potential disaster area is the smallest, but the destructive force is the greatest. When 10°<θ<30°, the landslide body slides in a small fan-shaped manner, and the expansion angle of the two side boundaries is between 20° and 60°. At this time, the potential disaster area is medium and the destructive force is medium. When θ is greater than 30°, the landslide body slides in a large fan-shaped manner, and the expansion angle of the two side boundaries is greater than 60°. At this time, the potential disaster area is the largest and the destructive force is the weakest. Based on the analysis of the above monitoring data, the landslide risk is judged and timely warnings are issued.

Images