Photogrammetry Methods for High and Steep Slopes from Multiple Angles

Abstract

The present invention proposes a method for multi-angle close-up photogrammetry of a high-steep slope, comprising the following steps: S1: constructing an initial "low-resolution" slope basic model, and dividing the slope into a number of terrain units; S2: using methods such as upright photogrammetry and oblique photogrammetry to establish a basic model of the photographed slope, and extracting the terrain information of the slope through the basic model; S3: performing multi-angle close-up route planning and shooting according to the terrain information units determined in S2; the method can effectively reduce photography blind spots, and ensure that as many structural surfaces as possible face the shooting direction of the drone, so as to obtain the most comprehensive point cloud data, thereby improving the accuracy of the real-life model of the high-steep slope.

Description

Technical Field

[0001] The invention relates to the field of rock mass image acquisition, and in particular to a high and steep slope multi-angle close-up photogrammetry method. Background Art

[0002] High and steep rock slopes are important sources of disasters such as collapse and landslides. Such slopes often cause huge damage and are difficult to study. Therefore, the prevention and control of high and steep rock slope disasters has become a hot topic and focus of current geological disaster research.

[0003] Based on the "rock structure control theory", the controlling factor of the stability of high-steep rock slopes is the rock structure surface. Therefore, the first prerequisite for the study of high-steep rock slopes is to obtain key information such as the occurrence and trace length of the rock structure surface. The structural surface of high-steep slopes is often difficult to reach by manpower, or requires a lot of manpower and material resources and it is difficult to ensure the safety of personnel, so non-contact measurement methods are required to collect rock structure surface information. However, traditional non-contact measurement methods such as three-dimensional laser scanning, digital close-range photogrammetry, and conventional drone photogrammetry have problems such as unclear images and many blind spots, and often cannot identify structural surfaces with smaller widths. These structural surfaces are the key influencing factors affecting collapse and rockfall. Therefore, there is an urgent need for a photogrammetry method that can comprehensively obtain high-resolution structural surface images to provide support for the subsequent refined measurement and interpretation of slope rock structure surfaces. Summary of the invention

[0004] The present invention proposes a high and steep slope multi-angle close-up photogrammetry method, comprising the following steps:

[0005] S1: Construct an initial "low-resolution" slope basic model and divide the slope into several terrain units: obtain a small amount of low-resolution images of the target and its surroundings through drone photography; then use the acquired low-resolution images to obtain rough scene information of the target and its surroundings through image matching, aerial triangulation, and dense matching photogrammetry processing; finally, perform close track planning based on the rough scene information;

[0006] S2: using methods such as upright photogrammetry and oblique photogrammetry to establish a basic model of the photographed slope, and extracting the topographic information of the slope through the basic model;

[0007] S3: Based on the terrain information units determined in S2, multi-angle close-up flight route planning and shooting are carried out; the camera angle is perpendicular to each terrain unit for image acquisition to ensure that as much point cloud information as possible is obtained, and the three vertices of each terrain unit are respectively fitted to the corresponding spatial plane, and then a close-up flight plane parallel to the fitted plane is created within a specific distance;

[0008] S4: image acquisition of the structural surface, which is acquired by adjusting the shooting angle on the flight plane, and the shooting angle is set to be perpendicular to each group of structural surfaces;

[0009] S5: After the shooting angle, overlap rate, and speed parameters of the aerial survey are set, the three-dimensional route is imported into the UAV ground controller. The UAV reads the route and automatically performs the shooting task according to the planned parameters to collect images;

[0010] S6: Supplement targeted flight route shooting in key areas with blind spots;

[0011] S7: Construction of three-dimensional real-life model of slope surface.

[0012] Preferably, the terrain information of the slope in S2 is divided by the cusp catastrophe theory. When it is necessary to determine that it is within a certain plane, the point cloud data is preprocessed, and the point cloud is denoised by point cloud downsampling. Then, the local plane in the three-dimensional point cloud model is fitted by the KNN algorithm to calculate the normal vector perpendicular to the local plane of the point cloud, and use it as the normal vector of the central point cloud. If the angle between the normal vectors of adjacent point clouds is less than a preset threshold, it is considered that they can be placed in the same plane, and these point clouds are used to fit the plane; with the point cloud of the fitted plane as the center, continue to search for adjacent point clouds. If the angle between the normal vector of the adjacent point cloud and the plane is less than the preset threshold, the point cloud is also placed in the plane and the parameters of the plane are recalculated; repeat the steps until all point clouds are placed in a suitable plane to determine the plane.

[0013] Preferably, the threshold is 30°

[0014] Preferably, the standard potential energy function of the cusp catastrophe model is

[0015] U(x,a,b)=x 4 -ax 2 +bx (1)

[0016] Where U is the potential energy function; x is the state variable; a and b are the control variables.

[0017] The equilibrium equation is used to describe the degree of change of the potential energy function of the system, which can be obtained by differentiating the potential energy function

[0018]

[0019] By further derivation of the equilibrium equation and combining it with the equilibrium equation, we can derive an equation for comprehensively considering the change of system functions and the severity of their change. This paper defines it as the state function Δ, which can be expressed as

[0020] Δ=8a 3 -27b 2 (3)

[0021] When Δ<0, the potential energy function changes smoothly, and we believe that it is not in a mutation state under this condition; when Δ=0

[0022] 2, it is in a critical equilibrium state, that is, the critical state between mutation and non-mutation; when Δ>0, it indicates that the function has undergone a discontinuous change, and is in a mutation state at this time, which is determined to be a new plane.

[0023] Preferably, in S3, the method for determining the vertical direction of each terrain unit is to fit and describe it by a spatial inclined plane, and set the vertex coordinates of the lower base of the slope (ν1, ν2) and the vertex coordinates of the upper base (ν3, ν4) to describe the slope plane P0. According to the vertex coordinates of the spatial plane P0, its normal vector N can be calculated. For the slope, it is considered that its normal vector N should always point to the positive direction of the Z axis. During the flight, the drone lens should be oriented and perpendicular to the plane. According to the pinhole imaging principle, when the field of view angle of the camera is fov and the shooting distance is d, the corresponding ground imaging range (G) is:

[0024]

[0025] Then, multi-angle approach routes are planned in the flight plane. By simply setting the camera angle to be perpendicular to the fitting plane, high-definition images of each terrain unit can be obtained from directly in front of the terrain unit.

[0026] Preferably, the setting of the lens shooting angle refers to the definition of the inclination angle of the structural surface, and the shooting direction is the clockwise angle between the projection of the lens central axis direction on the horizontal plane and due north, and the shooting direction is opposite to the slope direction; the shooting pitch angle is the angle between the lens central axis and the horizontal direction, where looking up is positive and looking down is negative, and the range of the pitch angle is [-130°, 40°], and the pitch angles are all complementary angles of the inclination angle.

[0027] Preferably, the blind area can be known from the rough scene information obtained in S1, and manual photography is used to take a panoramic close-up shot to eliminate the blind area.

[0028] The multi-angle close-up photogrammetry method for high and steep slopes proposed in the present invention has the following beneficial effects: the method can effectively reduce photography blind spots and ensure that as many structural surfaces as possible face the direction of drone shooting, thereby obtaining the most comprehensive point cloud data, thereby improving the accuracy of the real-life model of the high and steep slope. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for describing the embodiments are briefly introduced below.

[0030] Figure 1 It is a framework diagram of the overall technical approach;

[0031] Figure 2 It is a cusp mutation model;

[0032] Figure 3 Schematic diagram of point cloud normal vector calculation method based on K nearest neighbor algorithm;

[0033] Figure 4 Schematic diagram of terrain unit division;

[0034] Figure 5 Create multi-angle photogrammetric diagrams for flight planes taking into account terrain;

[0035] Figure 6 This is a schematic diagram for calculating the shooting direction and pitch angle under normal circumstances;

[0036] Figure 7 Schematic diagram of multi-angle close-up photogrammetry considering the grouping characteristics of structural surface development;

[0037] Figure 8 It is a schematic diagram of the shooting direction and calculation of the pitch angle of the anti-tilt structural surface;

[0038] Fig. 9 This is a schematic diagram of shooting in the blind area of ​​the viewing angle; DETAILED DESCRIPTION

[0039] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0040] Example

[0041] The present invention proposes a high and steep slope multi-angle close-up photogrammetry method, comprising the following steps:

[0042] S1: Construct an initial “low-resolution” slope base model, such as Figure 4 As shown in the figure, the slope is divided into several terrain units: a small amount of low-resolution images of the target and its surroundings are obtained by drone photography; then the obtained low-resolution images are used to obtain rough scene information of the target and its surroundings through image matching, aerial triangulation, and dense matching photogrammetry processing; finally, close track planning is carried out based on the rough scene information;

[0043] When there is no known initial scene information for the object to be photographed, it is necessary to obtain a small amount of low-resolution images of these targets and their surroundings through regular UAV flights or manually operated UAV photography. Close-up photography of the target (especially emergency photography of landslides) often requires rapid acquisition of the target's initial scene information. Therefore, when shooting low-resolution images of the target, it is necessary to pay attention to the following: 1) When shooting, the UAV can be slightly farther away from the target to reduce the number of images required to cover the target and reduce the time required for photogrammetric processing; 2) While ensuring coverage of the target body, try to capture the environment around the target to avoid the safety of close flight due to lack of obstacle information around the target during subsequent track planning.

[0044] S2: Use methods such as upright photogrammetry and oblique photogrammetry to establish a basic model of the photographed slope. This method has a low resolution, often at the centimeter level, and the terrain information of the slope is extracted through the basic model;

[0045] According to the slope aspect and slope information, the slope is divided into multiple terrain units with similar slope aspects and slopes (see Appendix Figure 2 ). In this process, terrain division can be performed based on the cusp catastrophe theory. The essence of dividing terrain units is to find the boundaries of the terrain, that is, the locations where the physical geometric information between adjacent terrain units changes discontinuously. The cusp catastrophe theory is used to describe the phenomenon and process of discontinuous changes and mutations of variables in the system. When identifying the boundaries of terrain units, the physical geometric properties (normal vectors) of the point clouds at the junction of different terrain units are in a critical state of mutation.

[0046] The catastrophe theory mainly uses mathematical models to discuss the general law of the state change from one stable configuration to another in the dynamic system. The standard potential energy function of the cusp catastrophe model is:

[0047] U(x,a,b)=x 4 -ax 2 +bx (1)

[0048] Where U is the potential energy function; x is the state variable; a and b are the control variables.

[0049] The equilibrium equation is used to describe the degree of change of the potential energy function of the system, which can be obtained by differentiating the potential energy function

[0050]

[0051] By further derivation of the equilibrium equation and combining it with the equilibrium equation, we can derive an equation for comprehensively considering the change of system functions and the severity of their change. This paper defines it as the state function Δ, which can be expressed as

[0052] Δ=8a 3 -27b2 (3)

[0053] When Δ<0, the potential energy function changes smoothly, and we believe that it is not in a mutation state under this condition; when Δ=0, it is in a critical equilibrium state, that is, the critical state between mutation and non-mutation; when Δ>0, it indicates that the function has undergone discontinuous changes and is in a mutation state.

[0054] To determine a plane: Figure 3 As shown in the figure, the point cloud data is preprocessed, and the point cloud is denoised by downsampling. Then, the local plane in the three-dimensional point cloud model is fitted by the KNN (K-Nearest Neighbor) algorithm to calculate the normal vector perpendicular to the local plane of the point cloud and use it as the normal vector of the central point cloud. If the angle between the normal vectors of adjacent point clouds is less than a preset threshold, the empirical threshold is 30°, then it is considered that they can be placed in the same plane, and these point clouds are used to fit the plane; with the point cloud of the fitted plane as the center, continue to search for adjacent point clouds. If the angle between the normal vector of the adjacent point cloud and the plane is less than the preset threshold, then the point cloud is also placed in the plane and the parameters of the plane are recalculated; repeat the steps until all point clouds are placed in a suitable plane to determine the plane.

[0055] S3: Based on the terrain information units determined in S2, multi-angle close-up flight route planning and shooting are carried out; the camera angle is perpendicular to each terrain unit for image acquisition to ensure that as much point cloud information as possible is obtained, and the three vertices of each terrain unit are respectively fitted to the corresponding space plane, and then a close-up flight plane parallel to it is created within a specific distance of the fitting plane. The characteristic distance is set to 5-100m. Specifically:

[0056] The method to determine the vertical direction of each terrain unit is to fit it with a spatial slope. Assume that the coordinates of the vertex of the lower base of the slope (ν1, ν2) and the coordinates of the vertex of the upper base (ν3, ν4) are two sets of parameters to describe the slope plane P0. According to the vertex coordinates of the spatial plane P0, its normal vector N can be calculated. For the slope, it is believed that its normal vector N should always point to the positive direction of the Z axis. During the flight, the drone lens should be facing and perpendicular to the plane. According to the pinhole imaging principle, when the camera's field of view is fov and the shooting distance is d, the corresponding ground imaging range (G) is:

[0057]

[0058] Then, a multi-angle approach route is planned in the flight plane. By simply setting the camera angle to be perpendicular to the fitting plane, a high-definition image of each terrain unit can be obtained from the front of the terrain unit (see Figure 2). Figure 5 ).

[0059] like Figure 6 As shown in the figure, the setting of the lens shooting angle refers to the definition of the inclination angle of the structural surface. The shooting direction is the clockwise angle between the projection of the lens central axis on the horizontal plane and the true north. Generally, the shooting direction is opposite to the slope direction; the shooting pitch angle is the angle between the lens central axis and the horizontal direction, where looking up is positive and looking down is negative. The range of the pitch angle is [-130°, 40°], and the pitch angles are all complementary angles of the inclination angle.

[0060] like Figure 7 As shown in the figure, the structural surfaces of the slope are complex and changeable, but due to the multi-period nature of geological stress, the structural surfaces are often distributed in groups. The above-mentioned route perpendicular to the terrain unit can construct a high-definition real-life model of the slope surface, but many structural surfaces are not parallel to the slope surface, so it is necessary to specifically consider the output characteristics of the structural surface for targeted route planning, that is, set the shooting angle on the flight plane to be perpendicular to each group of structural surfaces.

[0061] like Figure 8 As shown, S4: image acquisition of the structural surface is obtained by adjusting the shooting angle on the flight plane, and the shooting angle is set to be perpendicular to each group of structural surfaces; the idea of ​​determining the shooting direction and the pitch angle is similar to the idea of ​​determining the spatial plane. Generally, the shooting direction is opposite to the inclination direction. For the anti-inclination structural surface, the shooting direction is the same as the inclination; the pitch angles are all the complementary angles of the inclination angle.

[0062] Then, S5: after the shooting angle, overlap rate, and speed parameters of the aerial survey are set, the three-dimensional route is imported into the UAV ground controller, the UAV reads the route, and automatically performs the shooting task according to the planned parameters to collect images;

[0063] like Fig. 9 As shown, S6: supplement the route with targeted photography in key areas with blind spots; the blind spots can be known from the rough scene information obtained in S1, and manual photography is used to take panoramic close-up photos to eliminate the photography blind spots.

[0064] S7: Construction of a three-dimensional real-life model of the slope surface. After the above multi-angle approach route planning is completed, the data processing can be performed by referring to the conventional model construction technology. The drone automatically executes the optimized multi-angle approach photography route and collects high-resolution images of the slope surface. These image data can then be imported into the mapping software for processing. The construction of a three-dimensional real-life model is an existing technology, and the DJI Zhitu software can be purchased for processing. The DJI Zhitu software can obtain a high-precision three-dimensional real-life model of the slope through steps such as image feature matching extraction - aerial triangulation - three-dimensional model generation.

[0065] The present invention constructs a basic site model of the study area, divides similar terrain into a terrain unit based on the basic site model, and fits a spatial plane according to each unit; creates a close flight plane parallel to the fitting plane at a specific distance (5 to 100 m) from the fitting plane; sets multi-angle close routes in the flight plane, and sets the camera angle to be perpendicular to the fitting plane; sets a targeted close flight for each group of structural surfaces in the flight plane according to the structural surface development grouping characteristics obtained from the field investigation and the basic model, and collects images at an angle perpendicular to the structural surfaces; supplements the route with targeted shooting in individual areas that are key research areas and prone to blind spots, and eliminates photographic blind spots; and constructs a three-dimensional real scene model by collecting images based on the above steps.

Claims

1. A high and steep slope multi-angle close-up photogrammetry method, characterized in that: The following steps are involved: S1: Construct an initial "low-resolution" slope basic model and divide the slope into several terrain units: obtain a small amount of low-resolution images of the target and its surroundings through drone photography; then use the acquired low-resolution images to obtain rough scene information of the target and its surroundings through image matching, aerial triangulation, and dense matching photogrammetry processing; finally, perform close track planning based on the rough scene information; S2: using the methods of upright photogrammetry and oblique photogrammetry to establish a basic model of the photographed slope, and extracting the topographic information of the slope through the basic model; The terrain information of the slope in S2 is divided by the cusp mutation model. When it is necessary to determine that it is within a certain plane, the point cloud data is preprocessed and the point cloud is denoised by downsampling. Then, the local plane in the three-dimensional point cloud model is fitted by the KNN algorithm to calculate the normal vector perpendicular to the local plane of the point cloud and use it as the normal vector of the center point cloud. If the angle between the normal vectors of adjacent point clouds is less than the preset threshold, they are considered to be placed in the same plane, and these point clouds are used to fit the plane; with the point cloud of the fitted plane as the center, continue to search for adjacent point clouds. If the angle between the normal vector of the adjacent point cloud and the plane is less than the preset threshold, the point cloud is also placed in the plane and the parameters of the plane are recalculated. Repeat the steps until all point clouds are placed in a suitable plane to determine the plane; The standard potential energy function of the cusp catastrophe model is: U(x,a,b)=x 4 -ax 2 +bx (1) Where U is the potential energy function; x is the state variable; a and b are the control variables; The equilibrium equation is used to describe the degree of change of the potential energy function of the system, which is obtained by derivation of the potential energy function. The equilibrium equation is further derived and combined with the equilibrium equation to derive an equation for comprehensively considering the change of the system potential energy function and its intensity of change. This paper defines it as the state function Δ, expressed as Δ=8a 3 -27b 2 (3) When Δ<0, the potential energy function changes steadily, and under this condition it is not in a mutation state; when Δ=0, it is in a critical equilibrium state, that is, the critical state between mutation and non-mutation; when Δ>0, it indicates that the function has undergone a discontinuous change, and is in a mutation state, which is determined to be a new plane; S3: Based on the terrain information units determined in S2, multi-angle close-up flight route planning and shooting are carried out; the camera angle is perpendicular to each terrain unit for image acquisition to ensure that as much point cloud information as possible is obtained, and the three vertices of each terrain unit are respectively fitted to the corresponding spatial plane, and then a close-up flight plane parallel to the fitted plane is created within a specific distance; S4: image acquisition of the structural surface, which is acquired by adjusting the shooting angle on the flight plane, and the shooting angle is set to be perpendicular to each group of structural surfaces; S5: After the shooting angle, overlap rate, and speed parameters of the aerial survey are set, the three-dimensional route is imported into the UAV ground controller. The UAV reads the route and automatically performs the shooting task according to the planned parameters to collect images; S6: Supplement targeted flight route shooting in key areas with blind spots; S7: Construction of three-dimensional real-life model of slope surface.

2. The high and steep slope multi-angle close-up photogrammetry method according to claim 1 is characterized in that: The threshold is 30°.

3. The high and steep slope multi-angle close-up photogrammetry method according to claim 1 is characterized in that: In S3, the method for determining the vertical direction of each terrain unit is to fit a spatial slope for description. The slope plane P0 is described by two sets of parameters: the vertex coordinates of the lower base of the slope (ν1, ν2) and the vertex coordinates of the upper base (ν3, ν4). The normal vector N is calculated according to the vertex coordinates of the spatial plane P0. For the slope, it is believed that its normal vector N should always point to the positive direction of the Z axis. During the flight, the drone lens should be oriented and perpendicular to the plane. According to the pinhole imaging principle, when the camera's field of view is fov and the shooting distance is d, the corresponding ground imaging range G is: Then, multi-angle approach routes are planned in the flight plane. By simply setting the camera angle to be perpendicular to the fitting plane, high-definition images of each terrain unit can be obtained from directly in front of the terrain unit.

4. The high and steep slope multi-angle close-up photogrammetry method according to claim 1 is characterized in that: The setting of the lens shooting angle refers to the definition of the inclination angle of the structural surface. The shooting direction is the clockwise angle between the projection of the lens central axis on the horizontal plane and the true north. The shooting direction is opposite to the slope direction; the shooting pitch angle is the angle between the lens central axis and the horizontal direction, where looking up is positive and looking down is negative. The range of the pitch angle is [-130°, 40°], and the pitch angle is the complementary angle of the inclination angle.

5. The high and steep slope multi-angle close-up photogrammetry method according to claim 1 is characterized in that: The blind spots are known from the rough scene information obtained from S1, and manual photography is used to take panoramic close-up photos to eliminate the blind spots.

Images