Cavern image expansion graph generation method based on adjacent front image complementation

Abstract

The invention discloses a cavern image expansion graph generation method based on adjacent front image complementation, and provides a set of cavern image expansion graph generation method based on adjacent front image complementation, which comprises the full technical process of image acquisition, data processing, pixel complementation extraction and pixel matrix picture output. Through front high-definition image acquisition and pixel extraction, the problem that many geological elements are invisible from the side face due to overexcavation or underexcavation of rock excavation cavern wall construction is well solved, the method can be well suitable for normal rock excavation caverns, and precision loss caused by image distortion is reduced to the maximum extent; and meanwhile, a complex operation flow of image distortion correction is also avoided.

Description

Technical Field

[0001] The invention relates to the field of underground engineering survey, and is particularly suitable for a method for generating a cavern image expansion map based on complementary adjacent frontal images. Background Art

[0002] Traditional geological cataloging combines on-site operations such as measuring with a tape measure, drawing lines with meter paper, and mass production with a compass with internal operations such as CAD vectorization. The final result is a CAD cataloging result marked with two-dimensional line segments and points. The result displays the geological information and relative position relationship of the engineering site in the form of points and lines. The display of engineering geological information is relatively limited, and it is impossible to achieve intuitive display of all geological elements and delivery of results.

[0003] With the rapid development of image processing technology, control measurement technology, and real-life image stitching technology, the application of real-life modeling and technology has provided more convenient data collection methods for surveying, mapping, and exploration. Visual analysis and output technology have also greatly improved the operating efficiency of various professional business applications and reduced business costs. At present, the construction geology of municipal, water conservancy, transportation and other industries has realized the real-life geological cataloging of excavated slopes and caverns. By constructing a three-dimensional real-life model and model-based geological information identification and calculation, the collection of geological elements in the excavation project and the three-dimensional visualization results display have been realized. However, it is still facing the problem of lack of two-dimensional real-life results that meet the requirements of construction cataloging.

[0004] Most of the existing methods for generating two-dimensional real-life results are done manually by stitching together multiple images of engineering parts in CAD, relying on manual stretching or scaling of adjacent photos to match each other. The operation is time-consuming and laborious, and the technical threshold is high. A few methods use photography along the axis of the cave toward the inside of the cave, and then extract the route of stitching the images of engineering parts by extracting the distant side walls and arch parts. However, this shooting method will cause image loss due to local occlusion caused by over-excavation and under-excavation during construction. At the same time, there are also problems such as poor image quality and large distortion caused by extracting distant side images, which cannot be applied to the excavation of caverns in normal projects. Summary of the invention

[0005] The present invention aims to provide a method for generating a cavern image expansion diagram based on complementary adjacent frontal images, so as to solve the problem of two-dimensional real scene results required by construction cataloging.

[0006] To achieve the above object, the present invention adopts the following technical solutions: The method for generating a cavern image expansion map based on complementary adjacent frontal images of the present invention comprises the following steps: S1, arrange shooting points according to fixed shooting intervals, complete the front image collection of the horizontal ring and top arch direction of the excavated cavern at each shooting point, and complete the aerial triangulation calculation of the excavated cavern through real scene modeling software to obtain the image position coordinates and image rotation matrix; S2, obtain the cross-sectional shape and size of the target output section cavern according to the engineering design drawing, and output the empty pixel matrix IMG of the expansion diagram of the excavated cavern according to the target pixel; S3, inversely calculate the engineering part and three-dimensional space coordinates corresponding to each pixel point P in the matrix IMG; S4, using k-means clustering algorithm and elbow turning point to divide the image into several sub-image groups according to spatial position; calculating the coordinates of the projection point P' of point P on the hole axis and the plane distance from point P' to the mass points of all sub-image groups; taking the sub-image group with the minimum plane distance as the target image group a; S5, indexing the image information of the image group a, wherein the image information includes the image name, file path, image plane position, image rotation matrix R, image principal point coordinates, and lens pixel focal length f; and calculating the shooting angle and shooting range of the image group a according to the image information of the image group a; S6, calculate the front image requirements of point P according to the direction θ of the hole axis, including the plane position and the shooting angle; find the image photos whose shooting range matches the plane position and has the same shooting angle from the image group a; S7, inversely calculating the pixel position of the image photo corresponding to the point P according to the three-dimensional space coordinates of the point P, and obtaining the RGB value of the point P; S8, summarizing the RGB values ​​of each point in the matrix IMG, outputting them in PNG format, and obtaining an expanded image of the real scene of the excavated cavern.

[0007] Furthermore, the cross-sectional shape and dimensions described in step S2 include the height of the straight wall section, the width of the cavern, the angle of the arc section, the radius of the arc section, the direction and slope of the cave axis.

[0008] Furthermore, the size of the empty pixel matrix IMG in step S2 is N*L*3, and the order from top to bottom is left wall bottom-left wall top-top arch-right wall top-right wall bottom, N is the number of rows of the matrix IMG, L is the number of columns of the matrix IMG, and 3 is the three color values ​​of R, G, and B.

[0009] Furthermore, the engineering parts in step S3 include a left wall, a top arch, and a right wall; and the three-dimensional space coordinates are calculated respectively according to the engineering parts to which they belong.

[0010] Furthermore, the shooting angle includes the rotation angles along the X-axis, Y-axis and Z-axis when the image is shot, following the principle that clockwise is a negative value and counterclockwise is a positive value.

[0011] Furthermore, the image photos corresponding to the point P include a left wall image photo A, a right wall image photo B, and a top arch image photo C according to the engineering part corresponding to the point P.

[0012] Furthermore, in step S7, the RGB value of point P is first obtained in the image photograph of the engineering part corresponding to point P; and secondly, the RGB value of point P is obtained in the image photograph adjacent to the engineering part corresponding to point P.

[0013] The advantage of the present invention is that it provides a method for generating cavern image expansion maps based on complementary adjacent front images, including a full technical process from image acquisition, data processing, pixel complementary extraction, and pixel matrix output images. The front high-definition image acquisition and pixel extraction effectively solve the problem that many geological elements are not visible from the side due to over-excavation or under-excavation of the rock excavation cavern wall construction. The present invention can be well applied to normal rock excavation caverns, minimizes the accuracy loss caused by image distortion, and also avoids the complex operation process of image distortion correction.

[0014] The present invention realizes intelligent grouping of image groups through the k-means elbow algorithm, maps the engineering coordinates of the excavated caverns in combination with the empty pixel matrix points and the projection coordinates of the coordinates on the cave axis, extracts the position information of the aerial triangulated image of the nearest image group and the rotation matrix to inversely calculate the shooting posture information to determine the front image that meets the requirements, calculates the corresponding image point positions in the front image and extracts the RGB values ​​to fill the pixel matrix, thereby realizing the generation of high-definition and distortion-free high-definition image expansion maps of excavated caverns, solving the long-standing problems in the construction geology fields of water conservancy, hydropower, municipal engineering and transportation engineering. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 The present invention is a flowchart of the method for generating a cavern image expansion map based on complementary adjacent frontal images.

[0016] Figure 2 It is a schematic diagram of the on-site frontal image acquisition and shooting method described in the present invention.

[0017] Figure 3 It is a schematic diagram of the correspondence between the two-dimensional unfolded diagram and the three-dimensional space coordinates of the present invention.

[0018] Figure 4 It is a schematic diagram of extracting the RGB values ​​of the front image pixels to the corresponding positions of the empty pixel matrix in the method of the present invention.

[0019] Figure 5 It is an expanded image diagram of the excavated cavern completed by the method of the present invention. DETAILED DESCRIPTION

[0020] The technical solutions in the embodiments of the present invention are described clearly and completely below. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0021] Example 1 Figure 1 As shown, the cavern image expansion diagram generation method based on adjacent front image complementation of the present invention is described in detail by taking a city gate cavern (round arch straight wall type), and comprises the following steps: S1, arrange shooting points for the excavated cavern according to fixed shooting intervals, complete the horizontal ring and top arch direction front image collection of the excavated cavern at each shooting point, complete the aerial triangulation calculation of the excavated cavern through real scene modeling software, and obtain the position coordinates of the image photos and the image rotation matrix and other information. The shooting method diagram is shown in the figure. Figure 2 shown.

[0022] S2, according to the engineering design drawings, obtain the cross-sectional shape and size information of the target output section cavern, and according to the image point size D Pixel Output the empty pixel matrix IMG of the excavation cavern expansion diagram. The cross-sectional shape and size information of the cavern include the height of the straight wall section, the width of the cavern, the angle of the arc section, the radius of the arc section, the direction of the cavern axis and the slope. The empty pixel matrix IMG is output according to the standard scale designed in the engineering design drawing. The size of IMG is N*L*3, and from top to bottom, it follows the order of left wall bottom-left wall top-top arch-right wall top-right wall bottom. Each position contains 3 color values ​​(R, G, B). The calculation is shown in formula (1).

[0023] (1) In the present invention, the calculation result is rounded to the nearest integer.

[0024] In formula (1), N is the total number of rows of the empty pixel matrix IMG, L is the total number of columns, and N is b H is the number of rows of the left and right side walls in IMG, and HD is the number of rows of the top arch section. b R is the side wall height, in meters. h is the radius of the arch segment, in m; Angle is the included angle of the arch segment, in °; Pixel is the side length of the square of the engineering part corresponding to the pixel point, in m; ZH Q is the starting pile number of the target tunnel section, in m; ZH Z The ending pile number of the target tunnel section, in meters.

[0025] S3, inversely calculate the engineering part corresponding to each pixel point P in the pixel matrix IMG and the three-dimensional space coordinates of the point P. The three-dimensional space coordinates of the point P include the stake number ZH n(That is, the X corresponding to point P n and Y n ) and elevation H n The engineering parts corresponding to the pixel point P include the left wall, the top arch, and the right wall. Therefore, the three-dimensional spatial coordinates of the point P are calculated according to the engineering parts to which it belongs.

[0026] In addition, the pile number calculation of point P in the entire cavern can refer to formula (2) for pile number positioning and identification when outputting results.

[0027] (2) For any pixel point P(n, l) in the pixel matrix IMG, when n <N b When , the engineering part corresponding to point P is the left wall of the excavated cavern. The three-dimensional space coordinate calculation formula of point P is: (3) When N b ≤n <N b When +HD, the engineering part corresponding to point P is the arc segment of the excavated cavern top arch. Considering the three-dimensional projection transformation of some pixel points in the arc segment in the direction of the cavern and the spatial vector perpendicular to the direction of the cavern, the arc space coordinates of point P are obtained.

[0028] (4) When n ≥ N b When +HD, the engineering part corresponding to point P is the right wall of the excavated cavern, and the three-dimensional space coordinate calculation formula of point P is: (5) Where l is the number of columns in the pixel matrix where point P is located, and n is the number of rows in the pixel matrix where point P is located. α is the starting angle of the arc segment, in degrees, which is converted to radians in the calculation, rad. θ is the direction of the tunnel axis, in degrees, which is converted to radians in the calculation, rad. Dip is the slope of the excavated cavern, in %, and W is the width of the cavern, in meters. X e , Y e , Z e The coordinates of the bottom plate of the cave axis corresponding to the starting pile number of the cave.

[0029] Then, the image photo and spatial position relationship (i.e., frontal image) containing the three-dimensional space coordinate point are calculated based on the image group information. Specifically: like Figure 3 As shown, it is point P in the excavated cave. Figure 3 The three-dimensional space coordinates of B and the midpoint of the pixel matrix IMG Figure 3 The position correspondence of A.

[0030] S4, using the k-means clustering algorithm and the elbow method, divide the images obtained in step S1 into several sub-image groups according to their spatial positions. The specific formula is as follows: (6) In the formula, i is the number of point cloud groups, c i To allocate to the first i Group point cloud group number of points, x i , y i To calculate the i The center point coordinates of the point cloud group, SSE is the single point coordinates of each point cloud group ( x , y ) to the center point of the point cloud group ( x i , y i ) is the sum of the distances.

[0031] Then get point P ( n，l ) After the three-dimensional engineering coordinates P (Xn, Yn, Hn) corresponding to the point P are calculated, the coordinates (Xn', Yn', Hn') of the projection point P' of the hole axis are calculated, and the plane distances of the point P' to the positions of the particles of all sub-image groups are calculated based on the pixel matrix IMG. The sub-image group with the minimum plane distance is taken as the target image group a. The formula is as follows: (7) S5, index the image information of image group a, including the image name, file path, image plane position, image rotation matrix R, image principal point coordinates, lens pixel focal length f, etc. corresponding to image group a, and complete the calculation of the shooting angle information and shooting range information of image group a based on the above image information.

[0032] The shooting angle information (ω, ϕ, k) represents the rotation angles along the X-axis, Y-axis, and Z-axis when the image is shot, and is calculated according to the principle that clockwise is a negative value and counterclockwise is a positive value.

[0033] (8) S6, calculate the front image requirements of point P according to the direction θ of the hole axis, including the plane position and shooting angle. According to the engineering part corresponding to point P, the front image of point P is divided into left wall image photo A, right wall image photo B, and top arch image photo C.

[0034] Find the image photos whose shooting range matches the plane position and has the same shooting angle from the image group a, that is, the image photos that meet the requirements of the front image of point P.

[0035] The shooting direction of the left wall image A should be perpendicular to the direction θ of the hole axis, and point P should be within the image FOV range, which should meet the following requirements: (9) Where FOV is the field of view of the camera and lens, in degrees; f is the focal length of the lens, in millimeters; Dx is the size of the short side of the sensor, in millimeters; and Dy is the size of the long side of the sensor, in millimeters.

[0036] The right wall image photo B and the top arch image photo C refer to the left wall image requirements.

[0037] S7, based on the three-dimensional spatial coordinates of point P, reversely calculate the pixel position PT (x n ,y n ), get the RGB value of point P.

[0038] For the image point P on the left wall ( n , l ), that is, n <N b , load the left wall front image photo A, calculate the image point position of image photo A corresponding to each position in turn, and extract the RGB value of the image point position.

[0039] like Figure 4 As shown, it shows the point P on the left wall of the cave ( n，l ) Figure 4 B is the front view photo on the left wall Figure 4 The pixel position PT on A ( x n 、y n )’s correspondence.

[0040] For the pixel position PT of the left wall point P ( x n 、y n ), when x n The n value is greater than the number of rows of image photo A, and can be obtained by searching the top arch image C.

[0041] For the image point P of the cave ceiling, that is, N b <n<N b +HD, load the top arch front image photo C, calculate the image point position of the image photo C corresponding to each position in turn, and extract the RGB value of the image point position.

[0042] For the case where the photo pixel does not exist, that is, the pixel position PT (x n ,y n ) n The value of n is greater than the number of rows in the image C. b <n<Nb +HD / 2 can be obtained by searching the left wall front image A; b +HD / 2≤n <N b In the case of +HD, it can be obtained by searching the right wall frontal image B.

[0043] For the image point P on the right wall, that is, n≥N b +HD, load the right wall front image photo B, calculate the image point position of image photo B corresponding to each position in turn, and extract the RGB value of the image point position.

[0044] If a photo pixel does not exist, that is, the pixel position PT (x n ,y n ) n The n value is less than 0 and can be obtained by searching the top arch part image C.

[0045] S8, summarizing the RGB values ​​of each point in the matrix IMG, outputting them in PNG format, and obtaining an expanded image of the real scene of the excavated cavern.

[0046] Example 2 In a pumped storage project, the cavern is a gate cavern with a width of W=7.5m and a vertical wall section height of H. b =4.75m, top arch arc radius Rh =3.8m, top arch arc angle Angle = 164°, tunnel axis strike θ = 246°, slope Dip = 6.5%, starting pile number ZH Q =135, end pile number ZH Z =155, starting point coordinate X e =64580.56,Y e =3923494.70, Z e =448.5. The excavated cavern is composed of a number of major joints and fissures, with a small opening, and the filling is mainly calcium, iron and rock debris, and is partially closed. The resolution size parameter Pixel is no more than 1cm, and the resolution of the cavern image expansion map meets the visibility of the joints and fissures.

[0047] After completing the image acquisition and aerial triangulation calculation on site, the empty pixel matrix IMG of the excavated cavern image expansion diagram is constructed according to the cavern design parameters. The IMG size is N*L*3, and from top to bottom, it follows the order of left wall bottom-left wall top-top arch-right wall top-right wall bottom. Refer to the formula and follow D Pixel The calculated values ​​are rounded to the nearest integer and get N=1746, L=2000, where H b =475, HD=796.

[0048] Starting from the matrix pixel point P (150, 250), first calculate the engineering part pile number ZH corresponding to the pixel point according to formula (2): n . After comparison, we can see that n <H b (150<475), that is, the point should be on the left wall of the cavern. Calculate the three-dimensional spatial coordinates of point P (X n , Y n and H n ).

[0049] The images are automatically grouped through the K-means clustering algorithm and the elbow method. The coordinates of the projection point P' of the hole axis corresponding to point P are calculated based on the three-dimensional spatial coordinates of point P. The plane distance between point P' and the mass points of each sub-image group is calculated, and the sub-image group with the minimum plane distance is taken as the target image group a.

[0050] The image name, plane position, principal point coordinates, rotation matrix, focal length and other information of the image group a and the lens CCD size (Dx and Dy) of the mobile phone VIVO S15 used for shooting are read in sequence, and the shooting posture ω, ϕ, k . The front image requirement of point P is calculated according to the direction θ of the hole axis. Then, combined with the field of view FOV, the left wall front image A, the top arch front image C and the right wall front image B corresponding to point P are determined.

[0051] Calculate the pixel position PT (x n and n ), after comparing whether the number of image rows exceeds the image height, select the xth image from the left wall front image A. n row, y n The RGB value of the pixel in the column is extracted and placed in the (150, 250) position of the empty pixel matrix IMG.

[0052] Complete all the empty values ​​of the empty pixel matrix IMG in the above order, and output the IMG matrix in PNG image format to obtain the image expansion diagram of the excavated cavern in this section, such as Figure 5 shown.

Claims

1. A method for generating a cavern image expansion map based on complementary adjacent frontal images, characterized in that: The following steps are involved: S1, arrange shooting points according to fixed shooting intervals, complete the front image collection of the horizontal ring and top arch direction of the excavated cavern at each shooting point, and complete the aerial triangulation calculation of the excavated cavern through real scene modeling software to obtain the image position coordinates and image rotation matrix; S2, obtain the cross-sectional shape and size of the target output section cavern according to the engineering design drawing, and output the empty pixel matrix IMG of the expansion diagram of the excavated cavern according to the target pixel; S3, inversely calculate the engineering part and three-dimensional space coordinates corresponding to each pixel point P in the matrix IMG; S4, using the k-means clustering algorithm and elbow inflection point to divide the image into several sub-image groups according to spatial location; Calculate the coordinates of the projection point P' of point P on the hole axis and the plane distance from point P' to the mass points of all sub-image groups; take the sub-image group with the minimum plane distance as the target image group a; S5, indexing the image information of the image group a, wherein the image information includes the image name, file path, image plane position, image rotation matrix R, image principal point coordinates, and lens pixel focal length f; and calculating the shooting angle and shooting range of the image group a according to the image information of the image group a; S6, calculate the front image requirements of point P according to the direction θ of the hole axis, including the plane position and the shooting angle; find the image photos whose shooting range matches the plane position and has the same shooting angle from the image group a; S7, inversely calculating the pixel position of the image photo corresponding to the point P according to the three-dimensional space coordinates of the point P, and obtaining the RGB value of the point P; S8, summarizing the RGB values ​​of each point in the matrix IMG, outputting them in PNG format, and obtaining an expanded image of the real scene of the excavated cavern.

2. The method for generating a cavern image expansion map based on complementary adjacent frontal images according to claim 1, characterized in that: The cross-sectional shape and dimensions described in step S2 include the height of the straight wall section, the width of the cavern, the angle of the arc section, the radius of the arc section, the direction and slope of the cave axis.

3. The method for generating a cavern image expansion map based on complementary adjacent frontal images according to claim 1, characterized in that: The size of the empty pixel matrix IMG in step S2 is N*L*3, and the order from top to bottom is left wall bottom - left wall top - top arch - right wall top - right wall bottom, N is the number of rows of the matrix IMG, L is the number of columns of the matrix IMG, and 3 is the three color values ​​of R, G, and B.

4. The method for generating a cavern image expansion map based on complementary adjacent frontal images according to claim 1, characterized in that: The engineering parts in step S3 include the left wall, the top arch, and the right wall; the three-dimensional space coordinates are calculated respectively according to the engineering parts to which they belong.

5. The method for generating a cavern image expansion map based on complementary adjacent frontal images according to claim 1, characterized in that: The shooting angle includes the rotation angle along the X-axis, Y-axis and Z-axis when the image is shot, following the principle that clockwise is a negative value and counterclockwise is a positive value.

6. The method for generating a cavern image expansion map based on complementary adjacent frontal images according to claim 1, characterized in that: The image photos corresponding to the point P include a left wall image photo A, a right wall image photo B, and a top arch image photo C according to the engineering part corresponding to the point P.

7. The method for generating a cavern image expansion map based on complementary adjacent frontal images according to claim 6, characterized in that: In step S7, the RGB value of point P is first obtained in the image photo of the engineering part corresponding to point P; secondly, the RGB value of point P is obtained in the image photo adjacent to the engineering part corresponding to point P.

Images