System for monitoring glacier dynamics by unmanned aerial vehicle low-altitude flight photogrammetry method

Abstract

The invention belongs to the technical field of glacier identification, and provides a system for monitoring glacier dynamics by an unmanned aerial vehicle low-altitude flight photogrammetry method, which comprises the following steps that: a glacier area unmanned aerial vehicle low-altitude photogrammetry module selects an unmanned aerial vehicle as required, and plans a route and parameters; the image preprocessing module is used for extracting key feature points from each image by using Pi * 4Dcap; fine three-dimensional information, an ortho-image and a glacier digital surface model DSM are obtained through point cloud generation and densification, and glacier dynamic information can be identified by utilizing the glacier digital surface model DSM after correction and interpolation processing are carried out on the three-dimensional information, the ortho-image and the glacier digital surface model DSM; calculating the elevation, volume, speed and mass change of the glacier based on a digital surface model (DSM) of the glacier; the glacier fissure is identified by utilizing the orthoimage and the glacier digital surface model DSM, and the system can comprehensively and accurately monitor the dynamic information of the glacier.

Description

Technical Field

[0001] This invention belongs to the field of glacier identification technology, specifically a system for monitoring glacier dynamics using a low-altitude drone photogrammetry method. Background Technology

[0002] Glacier distribution areas have harsh natural environments, with low temperatures, complex terrain, and changeable weather. For a long time, monitoring glacier changes has mainly relied on traditional manual ground-based observations. This method is time-consuming, labor-intensive, and carries certain risks. In maritime glacier areas, manual monitoring is even more difficult.

[0003] The development of satellite remote sensing technology has made large-scale glacier monitoring possible. However, glacier areas are affected by factors such as spatiotemporal resolution, clouds, and snow cover. In particular, maritime glacier areas are rich in water vapor, and cloudless remote sensing data is scarce. Although airborne radar can acquire high spatiotemporal resolution data and is not limited by weather, it is expensive and difficult to operate, making it difficult to promote and apply it on a large scale and at a high frequency in mountainous areas. Ground observation and satellite remote sensing are difficult to combine effectively, which cannot meet the needs of accurate monitoring of dynamic changes in glaciers.

[0004] Therefore, this invention provides a system for monitoring glacier dynamics using a low-altitude unmanned aerial vehicle (UAV) photogrammetry method. Summary of the Invention

[0005] In order to overcome the shortcomings of the prior art, at least one technical problem raised in the background art is solved.

[0006] The technical solution adopted by this invention to solve its technical problem is:

[0007] In a first aspect, the present invention provides a system for monitoring glacier dynamics using a low-altitude unmanned aerial vehicle (UAV) photogrammetry method, comprising:

[0008] Low-altitude photogrammetry module for glacier areas: Select appropriate UAV cameras based on glacier characteristics and set flight parameters appropriately to carry out photogrammetry;

[0009] Image preprocessing module: Imports comprehensive image information captured by drone camera, performs feature point extraction and matching, and projects the three-dimensional object onto the two-dimensional image plane through projection transformation;

[0010] Digital Surface Model (DSM) Generation Module: Based on multiple images that have undergone projection transformation, the module uses SfM technology to reconstruct the three-dimensional structure of the glacier and generate a digital surface model (DSM) and orthophotos.

[0011] Glacier dynamic information identification module: Calculates glacier dynamic change information based on the glacier digital surface model (DSM) and orthophotos;

[0012] Glacier fissure identification module: Identifies glacier fissures using orthophotos;

[0013] Secondly, the present invention provides a method for monitoring glacier dynamics using low-altitude unmanned aerial vehicle (UAV) photogrammetry, including:

[0014] S1: Select a suitable UAV camera based on the characteristics of the glacier and set the flight parameters appropriately to carry out photogrammetry;

[0015] S2: Import the comprehensive information of images captured by the drone camera, extract and match feature points, and project the three-dimensional object onto the two-dimensional image plane through projection transformation;

[0016] S3: Based on multiple images that have undergone projection transformation, the three-dimensional structure of the glacier is restored using SfM technology to generate a digital surface model (DSM) and orthophotos.

[0017] S4: Calculate dynamic change information of glaciers based on the Digital Surface Model (DSM) and orthophotos;

[0018] S5: Identify glacial fissures using orthophotos.

[0019] The beneficial effects of this invention are as follows: it solves the problems of traditional manual monitoring, reduces labor costs and risk factors, and makes up for the shortcomings of satellite remote sensing technology, which may be unable to accurately capture subtle changes in local glaciers; UAVs can conduct detailed photography and monitoring of specific areas of glaciers, and can discover details such as tiny cracks on the glacier surface and local elevation changes, providing more comprehensive data support for the accurate monitoring of dynamic changes in glaciers. Attached Figure Description

[0020] The invention will now be further described with reference to the accompanying drawings.

[0021] Figure 1 This is a system module diagram of the system for monitoring glacier dynamics using the UAV low-altitude flight photogrammetry method of the present invention;

[0022] Figure 2 This is a flowchart illustrating the steps of the UAV low-altitude flight photogrammetry method for monitoring glacier dynamics according to the present invention. Detailed Implementation

[0023] To make the technical means, creative features, objectives and effects of this invention easier to understand, the invention will be further described below in conjunction with specific embodiments.

[0024] Example 1

[0025] like Figure 1 As shown in the embodiment of the present invention, the system for monitoring glacier dynamics using a UAV low-altitude flight photogrammetry method includes:

[0026] Low-altitude photogrammetry module for glacier areas using drones: Select a suitable drone camera based on the characteristics of the glacier area, such as an area of ​​approximately 1 km². 2 Choose a rotary-wing drone camera for left or right sides, and a fixed-wing or hybrid drone camera for large or long areas. The drone's single flight time should be no less than 30 minutes. Based on the daily changes in temperature and wind speed, the best flight time is from 7:00 AM to 12:00 PM. Plan the payload reasonably and arrange at least two flights. The first flight is used for route debugging, and the subsequent flights are for normal operation.

[0027] During the mission planning phase, the drone flight path planning tool Pix4Dcapture needs to be used to ensure that the image coverage area is sufficient.

[0028] Phase control points were deployed at the glacier terminus and on both sides of the glacier tongue to assess image accuracy. When there were no phase control points, cross-validation was used for evaluation. During aerial survey design, flight path planning tools were used to ensure image coverage, guaranteeing a forward overlap of more than 75% and a lateral overlap of more than 60%, with a flight altitude below 120 meters.

[0029] Specifically, during the mission planning phase, the drone flight path planning tool Pix4Dcapture needs to be used to ensure that the image coverage area is sufficient.

[0030] The requirements for overlap parameters are: forward overlap of more than 75% and lateral overlap of more than 60% to fully guarantee image stitching in subsequent stages.

[0031] Regarding flight altitude parameters: Flight altitude directly affects ground resolution (GSD). While excessively high flight altitudes reduce the number of images that need to be captured, they also reduce the detail of the images. This invention sets the flight altitude of the drone camera below 120 meters to ensure that the captured images have appropriate resolution.

[0032] It is important to note that during the first flight, the flight path should be carefully adjusted, parameters such as aircraft attitude and high-altitude wind speed should be recorded, and the image acquisition effect should be checked. If it does not match the preset parameters, it should be optimized. The second flight should be conducted as normal aerial photography according to the optimized flight path.

[0033] Image preprocessing module: Imports comprehensive image information captured by the drone camera; the comprehensive image information includes: EXIF ​​data, GPS location, and camera intrinsic and extrinsic parameter matrix data; to facilitate subsequent processing;

[0034] In the first step of image preprocessing, Pix4Dcapture uses a feature detection algorithm to extract key feature points from each image; it also uses the feature detection algorithm to detect local features in the image; and finally, it projects the three-dimensional object onto the two-dimensional image plane through projection transformation.

[0035] First, image alignment is performed, followed by feature point extraction and matching;

[0036] Based on scale-invariant features, the formula L(x,y,σ)=G(x,y,σ)×I(x,y) is used to detect identical feature parts in an image; where L(x,y,σ) is the image after Gaussian blurring at different scales, G(x,y,σ) is the Gaussian kernel function, and I(x,y) is the input image.

[0037] Feature points between different images are matched based on the nearest neighbor matching method; the nearest neighbor matching method is used to find corresponding feature points between two images;

[0038] Projection transformation is performed using a pinhole camera model; the specific formula for the pinhole camera model is as follows: Where X, Y, and Z are the homogeneous coordinates of a point in three-dimensional space; s is a preset scaling factor representing the depth of the object point; u and v are image coordinates; 1 is an extra component value, facilitating matrix operations and transformations; K represents the camera intrinsic parameter matrix data, which can be expressed as: Where fx and fy are the focal lengths of the camera in the x-axis and y-axis directions, respectively, representing the reciprocal of the actual physical length corresponding to each pixel on the image plane; cx and cy are the optical centers of the camera, representing the intersection points of the camera's optical axis and the image plane; [R|t] is the camera extrinsic parameter matrix, where R is a 3×3 rotation matrix used to describe the rotation of the UAV camera relative to the world coordinate system, consisting of three rotation angles; and t is a 3×1 translation vector used to describe the translation of the UAV camera relative to the world coordinate system.

[0039] By using a pinhole camera model, three-dimensional objects can be projected onto a two-dimensional image plane;

[0040] Digital Surface Model (DSM) Generation Module: Based on multiple images that have undergone projection transformation, the three-dimensional structure of the glacier is reconstructed using SfM technology;

[0041] The essential matrix describes the geometric relationship of the same point in space under the coordinate systems of two drone cameras from two different perspectives. It is only related to the relative position of the drone cameras. For example, the positional relationship of two different drone cameras when they are shooting the same target. The positional relationship can be vertical or horizontal in space.

[0042] Specifically, the essential matrix F satisfies the following formula: in, X1 and X2 are the normalized coordinates of a point in space in two different drone camera coordinate systems, respectively. For example, two drone cameras in different locations are shooting the same coordinate point, which is the coordinate point in the coordinate systems of the two drone cameras. It should be noted that the normalized coordinates refer to the coordinates of the image coordinates transformed to the camera coordinate system through the camera intrinsic parameters and Z=1.

[0043] In SfM, the essential matrix is ​​used to recover the relative motion between cameras, helping to determine the pose relationship between different images, thus providing a basis for subsequent 3D point reconstruction;

[0044] Then, using the triangulation rules, the three-dimensional coordinates are recovered using point pairs from multiple images; the triangulation formula is as follows: C1 + λ1d1 = C2 + λ2d2; where C1 and C2 are the optical centers of the two cameras; λ1 and λ2 are preset scaling factors; d1 and d2 are the line-of-sight directions of the two cameras;

[0045] After feature extraction, feature matching, and triangulation, preliminary three-dimensional information of the scene can be obtained. This information is presented in the form of a sparse point cloud. The sparse point cloud consists of a series of discrete three-dimensional points, each of which represents a feature location in the scene. These points are obtained by triangulation calculations on the feature points matched in images from different viewpoints.

[0046] Based on the stereo matching method, the disparity value between corresponding pixels is calculated by finding corresponding pixels in images from different viewpoints, and then the disparity value is converted into a depth map according to the camera's intrinsic parameters.

[0047] For example, in two images taken by different drones, two images taken by drones from different locations can be matched, and a depth map can be obtained by calculating the disparity value.

[0048] The disparity is calculated for every two or more sets of images using a formula, and a depth map is generated using the disparity calculation. The specific formula is as follows: Where f is the camera focal length, b is the baseline distance, and d is the parallax value;

[0049] Once the depth map is obtained, it can be converted into a 3D point cloud. For each pixel in the image, the coordinates of the point in 3D space can be calculated based on its corresponding depth value and the camera's intrinsic parameters, thereby generating a new 3D point. These newly generated points are then merged with the previous sparse point cloud, thus achieving point cloud densification.

[0050] In actual shooting, drone cameras are not perfectly perpendicular to the ground and have a certain tilt angle, which will cause geometric distortion in the image; at the same time, the undulation of the terrain will also make the position and shape of ground objects in the image inconsistent with the actual situation; in order to obtain images that can accurately reflect the real situation of the ground, geometric correction is required to eliminate the influence of these factors.

[0051] By using preset control points and their corresponding points on the image, the homography matrix H is calculated; by using the principle of perspective transformation, each pixel in the image can be transformed and mapped to the corrected position to obtain an orthophoto.

[0052] The specific process involves using the formula: Where (X, Y) is a point on the image, and (X1, Y1) is the new coordinate position;

[0053] The generated orthophotos facilitate subsequent identification of glacier fissures;

[0054] The obtained point cloud data is further processed. Point cloud data consists of a series of discrete 3D points that record the positional information of the object's surface. This point cloud data is the basis for generating the DSM. For example, after performing a 3D scan of a glacier in a region, point cloud data containing information about the surfaces of various objects such as buildings, trees, and the ground will be obtained. Since the point cloud data is discrete, while the DSM needs to be a continuous surface model, the nearest neighbor interpolation algorithm is needed to fill the blank areas between the points.

[0055] The nearest neighbor interpolation method is used to fill the blank areas between points. This method is simple and direct. For the location to be interpolated, the nearest point cloud data point is found, and then the height value of that point is assigned to the interpolation point. The blank areas between points are filled in this way to generate the surface information of the glacier.

[0056] Because there are certain errors in camera pose estimation and 3D point reconstruction during the entire image alignment and 3D reconstruction process, these errors will cause the reconstruction results to deviate from the actual situation; it is necessary to use bundle adjustment to eliminate the deviation.

[0057] The goal of bundled adjustments is to simultaneously optimize the position and pose of all cameras and the position of all 3D points, minimizing reprojection error. Reprojection error refers to the error between the point on the image plane projected by the camera's intrinsic and extrinsic parameters and the corresponding point in the actual image.

[0058] Specifically, for n 3D points Xi (i = 1, 2, 3, ..., n) and m cameras, for the j-th camera (j = 1, 2, 3, ..., m), its intrinsic parameter matrix is ​​Kj, and its extrinsic parameter matrix is... The 3D point Xi is projected onto the image plane using the intrinsic and extrinsic parameters of the j-th camera to obtain the point. The corresponding point in the actual image is x. ij The objective function for bundled adjustment is typically expressed as: in, Represents the square of the Euclidean distance;

[0059] The above steps of processing and adjustment generate a digital surface model (DSM) of the glacier, which facilitates the identification of subsequent dynamic information of the glacier.

[0060] The Digital Surface Model of Glaciers (DSM) is a digital representation of glacier height information. It is a dataset that expresses the height values ​​of glaciers at discrete points such as regular grid points or irregular triangular grids within a certain range. The Digital Surface Model of Glaciers (DSM) not only records the undulations of the terrain, but also covers the height of all objects on the ground, presenting a complete surface morphology.

[0061] Glacier dynamic information identification module: Calculates glacier dynamic change information based on the glacier digital surface model (DSM) and orthophotos; glacier dynamic change information includes: glacier elevation change, glacier volume change, and glacier flow velocity;

[0062] The core principle is to determine the elevation changes of a glacier by comparing the height changes of a specific point corresponding to the digital surface model (DSM) of a glacier at different times. This process is usually used in dynamic monitoring of glaciers, where dynamic information of the glacier is estimated by periodically generating digital surface models (DSM) of the glacier area.

[0063] Specifically, the change in glacier elevation at a given point (x, y);

[0064] The following formula can be used to calculate: Δh(x,y)=h2(x,y)-h1(x,y); where Δh(x,y) is the change in glacier elevation, h1(x,y) is the glacier digital surface model (DSM) data for the first period, and h2(x,y) is the glacier digital surface model (DSM) data for the second period.

[0065] In this invention, the first period and the second period are used only to indicate the order of time, with the first period being earlier than the second period. Their purpose is only to distinguish the glacier digital surface model (DSM) data from different periods.

[0066] Based on the calculation results, the following three cases can be identified:

[0067] If Δh(x, y) is greater than zero, it means that the glacier elevation at that point has increased.

[0068] If Δh(x, y) is less than zero, then the glacier elevation at that point has decreased.

[0069] If Δh(x, y) equals zero, then the glacier elevation at the point does not change significantly;

[0070] Furthermore, the change in glacier volume can be calculated by calculating the change in glacier elevation;

[0071] Specifically, to estimate the volume of a glacier with a known area S, the average elevation change Δh(x,y) of the glacier in this area is calculated. The volume of the glacier can then be calculated using the formula: V=S·Δh(x,y). Similarly, the mass change of the glacier can be calculated using the formula based on its density: M=ρ·V; where ρ is the density of the glacier.

[0072] By utilizing digital surface model (DSM) data of glaciers from different periods, information on the volume changes of glaciers can be inferred.

[0073] Furthermore, based on orthophotos, glacier flow velocities were calculated using COSI-Corr.

[0074] Specifically, the displacement of each point on the glacier surface is extracted from two orthophotos taken at different times, and the flow velocity is calculated based on the displacement and the time difference Δt between the drone images.

[0075] For the correlation point region between two orthophotos I1(x1, y1) and I2(x2, y2) from different periods, the displacement is calculated using the correlation point region; the velocity change of the glacier is then inferred.

[0076] The displacement in the horizontal component is Δx = x2 - x1; the displacement in the vertical direction is Δy = y2 - y1; then the total displacement can be expressed by the formula: Its total flow velocity can be expressed as:

[0077] Glacier Crevice Recognition Module: Based on high-resolution orthophotos from UAVs, the key to extracting glacier crevices lies in recognizing the edges, texture variations, and morphological features in the images;

[0078] Canny edge detection is one of the key steps in extracting glacial crevasses, because glacial crevasses are usually represented by prominent edges in images. Based on this, Canny edge detection is used to extract glacial crevasses.

[0079] The Canny algorithm first uses gradients to calculate the edges of an image. In an image, edges typically correspond to rapid changes in pixel values. Specifically, the color contrast between glacier surface fissures and the main body of the glacier is very large. Therefore, the pixel values ​​of the main body of the glacier and the fissures in the image will change rapidly, such as rapidly changing from white to black, i.e., from glacier to glacier fissures; or rapidly changing from black to white, i.e., from glacier fissures to glacier.

[0080] The core formula for gradient calculation is: in, and It is the gradient in the x and y directions;

[0081] After calculating the gradient magnitude G, non-maximum suppression is needed to obtain finer edges. This step checks the neighboring pixels of each pixel in its gradient direction. If the gradient magnitude of the pixel is not a local maximum, it is set to zero. This can eliminate the blurring of the edges and retain only the true edge points.

[0082] After nonmaximum suppression, edge points need to be further filtered using dual threshold detection; two thresholds are set: a low threshold T... L and high threshold T H Gradient magnitude greater than high threshold T H Pixels with gradient magnitudes less than a low threshold T are identified as strong edge points. L Pixels are excluded, while gradient magnitudes are between the low threshold T. L and high threshold T H Pixels between points are retained as edge points if they are connected to strong edge points, otherwise they are excluded; this can effectively remove false edges caused by noise while preserving true edge information.

[0083] Finally, morphological processing methods are used, through multiple steps such as dilation, erosion, and binarization, to effectively identify glacial fissure information from the images.

[0084] The technical solution of this embodiment is as follows: Based on the Digital Surface Model (DSM) of glaciers, data from different periods are compared to calculate the dynamic information of the glacier; the elevation and volume changes of a certain point of the glacier are calculated using a simple formula, which can intuitively determine whether the elevation of the glacier at that point is rising, falling, or remaining unchanged; based on orthophotos, the displacement of each point on the glacier surface is extracted from two orthophotos from different periods using the COSI-Corr method, and the flow velocity is calculated based on the displacement and the time difference between the images; the Canny edge detection algorithm is used to extract glacier fissures. This algorithm, through gradient calculation, non-maximum suppression, and dual threshold detection, can accurately identify prominent edges in the image and effectively extract glacier fissure information.

[0085] Example 2

[0086] like Figure 2 As shown in Example 1, the present invention also provides a method for monitoring glacier dynamics using a low-altitude unmanned aerial vehicle (UAV) photogrammetry method, comprising:

[0087] S1: Select a suitable drone camera based on the characteristics of the glacier area, such as a glacier area of ​​approximately 1 km². 2 Rotary-wing UAV cameras are selected for left and right sides, while fixed-wing or hybrid UAV cameras are selected for large or long areas. Phase control points are set up at the end of the glacier and on both sides of the glacier tongue to evaluate image accuracy. When there are no phase control points, cross-validation evaluation is used. When designing aerial surveys, flight path planning tools are used to ensure image coverage, guaranteeing a forward overlap of more than 75% and a lateral overlap of more than 60%, and a flight altitude of less than 120 meters.

[0088] S2: Import comprehensive image information captured by the drone camera; the comprehensive image information includes: EXIF ​​data, GPS location, and camera intrinsic and extrinsic parameter matrix data; to facilitate subsequent operation and processing;

[0089] Pix4Dcapture uses feature detection algorithms to extract key feature points from each image; it also uses feature detection algorithms to detect local features in the image; finally, it projects the 3D object onto the 2D image plane through projection transformation.

[0090] Based on scale-invariant features, the formula L(x,y,σ)=G(x,y,σ)×I(x,y) is used to detect identical feature parts in an image; where L(x,y,σ) is the image after Gaussian blurring at different scales, G(x,y,σ) is the Gaussian kernel function, and I(x,y) is the input image.

[0091] Feature points between different images are matched using the nearest neighbor matching method; corresponding feature points between two images are found using the nearest neighbor matching method; projection transformation is performed using a pinhole camera model; the specific formula for the pinhole camera model is as follows: Where X, Y, and Z are the homogeneous coordinates of a point in three-dimensional space; s is a preset scaling factor representing the depth of the object point; u and v are image coordinates; 1 is an extra component value, facilitating matrix operations and transformations; K represents the camera intrinsic parameter matrix data, which can be expressed as: Where fx and fy are the focal lengths of the camera along the x and y axes, respectively, representing the reciprocal of the actual physical length corresponding to each pixel on the image plane; cx and cy are the optical centers of the camera, representing the intersection of the camera's optical axis and the image plane; [R|t] is the camera extrinsic parameter matrix, where R is a 3×3 rotation matrix used to describe the rotation of the UAV camera relative to the world coordinate system, consisting of three rotation angles; t is a 3×1 translation vector used to describe the translation of the UAV camera relative to the world coordinate system; by using the pinhole camera model, three-dimensional objects can be projected onto a two-dimensional image plane.

[0092] S3: Based on multiple projected images, the 3D structure of the glacier is reconstructed using SfM technology; the essential matrix F is used to describe the geometric relationship of the same point in space under two different viewpoints in the coordinate systems of two UAV cameras, and it depends only on the relative position of the UAV cameras; specifically, the essential matrix F satisfies the following formula: in, X1 and X2 are the normalized coordinates of a point in space in two different UAV camera coordinate systems, respectively. It should be noted that normalized coordinates refer to the coordinates of the image coordinates transformed to the camera coordinate system through the camera intrinsic parameters and Z=1.

[0093] Using triangulation, the 3D coordinates are recovered. The triangulation formula is as follows: C1 + λ1d1 = C2 + λ2d2; where C1 and C2 are the optical centers of the two cameras; λ1 and λ2 are preset scaling factors; and d1 and d2 are the line-of-sight directions of the two cameras. Based on stereo matching, corresponding pixels are found between images from different viewpoints, and the disparity value between them is calculated. Then, based on the camera's intrinsic parameters, the disparity value is converted into a depth map. A depth map is generated using disparity calculation. The specific formula is as follows: Where f is the camera focal length, b is the baseline distance, and d is the parallax value;

[0094] By using preset control points and their corresponding points on the image, the homography matrix H is calculated; by using the principle of perspective transformation, each pixel in the image can be transformed and mapped to the corrected position to obtain an orthophoto.

[0095] The specific process involves using the formula: Where (X, Y) is a point on the image, and (X1, Y1) is the new coordinate position;

[0096] The generated orthophotos facilitate subsequent identification of glacier fissures;

[0097] The obtained point cloud data is further processed. Nearest neighbor interpolation is used to fill the blank areas between points to generate glacier surface information. The bundled adjustment method is used to optimize the position and pose of all cameras and the position of all 3D points to minimize the reprojection error. The reprojection error refers to the error between the point on the image plane projected by the camera's intrinsic and extrinsic parameters and the corresponding point in the actual image.

[0098] Specifically, for n 3D points Xi (i = 1, 2, 3, ..., n) and m cameras, for the j-th camera (j = 1, 2, 3, ..., m), its intrinsic parameter matrix is ​​Kj, and its extrinsic parameter matrix is... The 3D point Xi is projected onto the image plane using the intrinsic and extrinsic parameters of the j-th camera to obtain the point. The corresponding point in the actual image is x. ij The objective function for bundled adjustment is typically expressed as: in, Represents the square of the Euclidean distance;

[0099] The above steps of processing and adjustment generate a digital surface model (DSM) of the glacier, which facilitates the identification of subsequent dynamic information of the glacier.

[0100] S4: Calculate glacier dynamic change information based on glacier digital surface model (DSM) and orthophotos; glacier dynamic change information includes: glacier elevation change, glacier volume change, and glacier flow velocity; the core principle is to determine the glacier elevation change by comparing the elevation change data of a specific point corresponding to the glacier digital surface model (DSM) at different times; this process is usually applied in glacier dynamic monitoring, by regularly generating glacier digital surface models (DSM) of glacier areas to infer the dynamic information of glaciers;

[0101] S5: Based on high-resolution orthophotos from UAVs, the key to extracting glacial crevasses lies in identifying edges, texture variations, and morphological features in the images;

[0102] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A system for monitoring glacier dynamics using low-altitude unmanned aerial vehicle (UAV) photogrammetry, characterized by: include: Low-altitude photogrammetry module for glacier areas: Select appropriate UAV cameras based on glacier characteristics and set flight parameters appropriately to carry out photogrammetry; Image preprocessing module: Imports comprehensive image information captured by drone camera, performs feature point extraction and matching, and projects the three-dimensional object onto the two-dimensional image plane through projection transformation; Digital Surface Model (DSM) Generation Module: Based on multiple images that have undergone projection transformation, the module uses SfM technology to reconstruct the three-dimensional structure of the glacier and generate a digital surface model (DSM) and orthophotos. Glacier dynamic information identification module: Calculates glacier dynamic change information based on the glacier digital surface model (DSM) and orthophotos; Glacier fissure identification module: Identifies glacier fissures using orthophotos.

2. The system for monitoring glacier dynamics using the UAV low-altitude flight photogrammetry method according to claim 1, characterized in that: The specific steps for feature point extraction and matching are as follows: The formula L(x,y,σ) = G(x,y,σ) × I(x,y) is used to detect identical feature parts in the image; where L(x,y,σ) is the image after Gaussian blurring at different scales, G(x,y,σ) is the Gaussian kernel function, and I(x,y) is the input image; feature points between different images are matched based on the nearest neighbor matching method.

3. The system for monitoring glacier dynamics using the UAV low-altitude flight photogrammetry method according to claim 1, characterized in that: The specific process of projecting a three-dimensional object onto a two-dimensional image plane through projection transformation is as follows: projection transformation is performed using a pinhole camera model; the specific formula is: Where X, Y, and Z are the homogeneous coordinates of a point in three-dimensional space; s is a preset scaling factor representing the depth of the object point; u and v are image coordinates; 1 is an extra component value, facilitating matrix operations and transformations; K represents the camera intrinsic parameter matrix data, which can be expressed as: Where fx and fy are the focal lengths of the camera in the x-axis and y-axis directions, respectively, representing the reciprocal of the actual physical length corresponding to each pixel on the image plane; cx and cy are the optical centers of the camera, representing the intersection of the camera's optical axis and the image plane; [R|t] is the camera extrinsic parameter matrix, where R is a 3×3 rotation matrix used to describe the rotation of the UAV camera relative to the world coordinate system, which consists of three rotation angles; and t is a 3×1 translation vector used to describe the translation of the UAV camera relative to the world coordinate system.

4. The system for monitoring glacier dynamics using the UAV low-altitude flight photogrammetry method according to claim 1, characterized in that: The specific process for generating an orthophoto is as follows: The homography matrix H is calculated using preset control points and their corresponding points on the image; each pixel in the image is transformed using the perspective transformation principle, mapping it to its corrected position to obtain the orthophoto. The specific process involves using the formula: Here, (X, Y) is a point on the image, and (X1, Y1) is the new coordinate position.

5. The system for monitoring glacier dynamics using the UAV low-altitude flight photogrammetry method according to claim 1, characterized in that: The specific information on the dynamic changes of the glacier includes: changes in glacier elevation, changes in glacier volume, and glacier flow velocity.

6. The system for monitoring glacier dynamics using the UAV low-altitude flight photogrammetry method according to claim 5, characterized in that: The specific process for calculating the glacier elevation change is as follows: Calculate using the formula: Δh(x,y) = h2(x,y) - h1(x,y); Δh(x,y) represents the glacier elevation change, h1(x,y) is the glacier digital surface model (DSM) data from the first period, and h2(x,y) is the glacier digital surface model (DSM) data from the second period; if Δh(x,y) is greater than zero, it indicates that the glacier elevation at that point has increased; if Δh(x,y) is less than zero, it indicates that the glacier elevation at that point has decreased; if Δh(x,y) is equal to zero, it indicates that the glacier elevation at that point has not changed significantly.

7. The system for monitoring glacier dynamics using the UAV low-altitude flight photogrammetry method according to claim 5, characterized in that: The specific process for calculating the volume change of a glacier is as follows: estimate the volume of a glacier with a known area S, calculate the average elevation change Δh(x,y) of the glacier in this area, and then calculate the volume of the glacier using the formula: V=S·Δh(x,y).

8. The system for monitoring glacier dynamics using the UAV low-altitude flight photogrammetry method according to claim 5, characterized in that: The specific process for calculating glacier flow velocity is as follows: The displacement of each point on the glacier surface is extracted from two orthophotos taken at different times, and the flow velocity is calculated based on the displacement and the time difference Δt between the drone image and the time of capture. The horizontal displacement Δx = x2 - x1; the vertical displacement Δy = y2 - y1; the total displacement can be expressed by the formula: The total flow velocity can be expressed as:

9. The system for monitoring glacier dynamics using the UAV low-altitude flight photogrammetry method according to claim 1, characterized in that: The process of identifying glacial fissures using orthophotos is as follows: glacial fissures are extracted using Canny edge detection based on high-resolution orthophotos from UAVs; finally, morphological processing methods are used, including dilation, erosion, and binarization, through multiple steps to identify glacial fissure information from the images.

10. A method for monitoring glacier dynamics using low-altitude unmanned aerial vehicle (UAV) photogrammetry, characterized by: include: S1: Select a suitable UAV camera based on the characteristics of the glacier and set the flight parameters appropriately to carry out photogrammetry; S2: Import the comprehensive information of images captured by the drone camera, extract and match feature points, and project the three-dimensional object onto the two-dimensional image plane through projection transformation; S3: Based on multiple images that have undergone projection transformation, the three-dimensional structure of the glacier is restored using SfM technology to generate a digital surface model (DSM) and orthophotos. S4: Calculate dynamic change information of glaciers based on the Digital Surface Model (DSM) and orthophotos; S5: Identify glacial fissures using orthophotos.

Images