Remote sensing orthoimage seamline generation method

Abstract

The present invention provides a remote sensing orthoimage seamline generation method, comprising: constructing constrained triangulation networks in effective intersection areas of adjacent orthoimages one by one; transforming an image topology search problem into a polygon selection problem of a triangle to generate a primary seamline network; and optimizing primary seamlines on the basis of building vector data and road vector data to generate an advanced seamline network. The advanced seamline network generated in the present invention can effectively avoid areas such as buildings and roads, and can achieve large-range image mosaicking, a mosaicking result is independent of the processing sequence, the flexibility and high efficiency are achieved during mosaicking processing, and the accumulation of errors is avoided.

Description

A method for generating mosaic lines in remote sensing orthophotos Technical Field

[0001] This invention belongs to the field of remote sensing image processing, and specifically relates to a method for generating mosaic lines in remote sensing orthophotos. Background Technology

[0002] Remote sensing image mosaicking involves stitching together multiple orthophotos with overlapping areas to create a larger-scale orthophoto. Automatic generation of seam lines is a crucial step in large-scale remote sensing image mosaicking. Existing seam line generation methods often focus only on adjacent images, neglecting to consider how to connect individual seam lines to form a seam line network, thereby creating the effective mosaic polygons for each orthophoto—that is, the range of pixels in each orthophoto that contribute to the mosaicking. For small-scale image mosaicking, this approach has little impact on applications. However, for large-scale image mosaicking, such as hundreds or thousands of images of a single region, the impact of processing efficiency and error accumulation on applications must be considered. Therefore, connecting individual seam lines to form a seam line network, and thus creating the effective mosaic polygons for each orthophoto, is essential for large-scale orthophoto mosaicking applications.

[0003] Existing techniques use conventional Voronoi diagrams to generate seam lines, but this cannot guarantee that the generated seam lines will be within the overlapping area of ​​adjacent orthophotos. This can lead to areas in the mosaicking result that are not covered by the images. Other researchers have proposed a method to connect individual seam lines into effective mosaic polygons, but this method requires knowledge of flight strip information within the region, limiting its applicability. Mainstream commercial photogrammetry software includes orthophoto mosaicking capabilities and has proposed its own solutions for seam line generation. Some systems do not consider how to connect the individual seam lines of each image to form a seam line network, while others, although capable of generating seam line networks, do not disclose their algorithms, and the seam lines may still be located in areas with obstacles such as buildings and trees.

[0004] Therefore, how to automatically generate the mosaicking network required for multi-orthophoto mosaicking, and how to effectively refine the mosaicking network so that the mosaicking lines can avoid passing through obstacles such as buildings and trees, has become one of the research hotspots in the fields of computer vision, photogrammetry and remote sensing. Summary of the Invention

[0005] This invention proposes a method for generating mosaic lines in remote sensing orthophotos. This method constructs constrained triangular networks in the effective intersection regions of adjacent orthophotos, transforming the image topology search problem into a polygon selection problem to generate a primary mosaic line network. The primary mosaic lines are then optimized using building and road vector data to generate a higher-level mosaic line network. The higher-level mosaic line network generated by this invention can effectively avoid areas such as buildings and roads, while also enabling large-scale orthophoto mosaicking. The mosaicking result is independent of the processing order, providing flexibility and efficiency in mosaicking and avoiding error accumulation. The method specifically includes the following steps:

[0006] S1 uses a tracking algorithm to sequentially obtain the effective regions of each orthophoto in the image set, and simplifies the effective regions to obtain the effective convex polygons of each orthophoto; the image set includes multiple orthophotos, multiple original images and corresponding DEM data; wherein multiple orthophotos constitute multiple pairs of adjacent orthophotos;

[0007] S2 identifies a first intersection region between effective convex polygons of a pair of adjacent orthophotos, the first intersection region including different types of boundary point sets and different types of boundary line sets; and constructs a constrained triangular mesh based on the different types of boundary point sets and boundary line sets in the first intersection region; the constrained triangular mesh includes multiple triangular patches; wherein different types of boundary point sets constitute different types of boundary line sets;

[0008] S3 extracts the target triangular facet set by determining the boundary point set to which each vertex in each triangular facet of the constrained triangular mesh belongs; the target triangular facet set includes multiple target triangular facets; each target triangular facet consists of one boundary line and two non-boundary lines;

[0009] S4 takes the midpoint of two non-boundary lines in each target triangular facet to obtain a midpoint set; the midpoint set includes multiple adjacent midpoints and two end midpoints; and sequentially connects the adjacent midpoints in the midpoint set to each other, and connects the two end midpoints to the nearest intersection point in the boundary point set in the first intersection region to obtain a pair of primary mosaic lines of adjacent orthophotos.

[0010] S5 Repeat steps S2 to S4 to connect all the initial mosaic lines of adjacent orthophotos to form a primary mosaic line network.

[0011] S6 optimizes the primary tessellation network based on building vector data to obtain an intermediate tessellation network; the intermediate tessellation network includes multiple intermediate tessellation lines;

[0012] S7 optimizes the intermediate mosaic line network based on road vector data to obtain the advanced mosaic line network.

[0013] Specifically, the different types of boundary point sets mentioned in step S2 include upper boundary point sets, lower boundary line point sets, and intersection point sets; the different types of boundary lines include upper boundary line sets, lower boundary line sets, and special boundary line sets; two adjacent upper boundary points in the upper boundary point set constitute an upper boundary line; two adjacent lower boundary points in the lower boundary point set constitute a lower boundary line; multiple upper boundary lines constitute an upper boundary line set; multiple lower boundary lines constitute a lower boundary line set; and boundary lines containing intersection points are considered special boundary lines.

[0014] Specifically, the method for extracting the target triangular facet set in step S3 is as follows:

[0015] Determine the boundary point set category to which each vertex in each triangular facet belongs in the constrained triangular network, remove triangular faces where any vertex belongs to the intersection point set and where all vertices belong to the same boundary point set category, and extract the target triangular facet set.

[0016] Specifically, step S6 further includes:

[0017] S61 obtains the original building positioning model based on the original building vector data corresponding to each original building in multiple orthophotos, and uses the model to calculate the first region of each original building in the corresponding orthophoto in turn; the first region includes multiple edge points, denoted as the first edge point set;

[0018] S62 will select the original buildings that exist in the first region in multiple orthophotos to form a primary building set. Then, the common intersection of each target building in the primary building set with the multiple first regions in multiple orthophotos will be obtained to obtain the target intersection region of each target building in multiple orthophotos. The multiple edge points in the target intersection region will be denoted as the second edge point set.

[0019] S63 sequentially filters all original buildings within a preset threshold range, centered on each second edge point, to form a set of intermediate buildings {R1, R2, ..., Rn}. For each intermediate building in the set, its first region on different orthophotos is calculated, forming a set of first regions for each intermediate building. The common intersection of all the first regions in the set of first regions of each intermediate building is taken to obtain the second region of each intermediate building in multiple orthophotos. The second regions of all intermediate buildings in multiple orthophotos constitute a set of second regions {S1, S2, ..., Sn}. The correspondence between Ri and Si is recorded, where i = 1 to n.

[0020] S64 performs an intersection test on the second region set {S1,S2,...,Sn} for all intermediate buildings' second regions, merges the second intersecting regions that intersect each other to form a third region V1, and takes the non-intersecting second regions as other third regions to form the third region set {V1,V2,...,Vm (m≤n)}.

[0021] S65 extracts all second edge points concentrated in the third region set {V1,V2,...,Vq,...,Vm} to form the third edge point set {P1,P2,...,Pj,...,Pw}; sequentially, taking each third edge point Pj in the third edge point set {P1,P2,...,Pj,...,Pw} as the center, it searches for high-level buildings {B1,B2,...,Bk,...,Bt(t≤n)} within a preset threshold range, and at the same time finds the third region Vq corresponding to each high-level building Bk, where k=1~t, q=1~m, j=1~w;

[0022] S66 sequentially calculates the minimum bounding rectangle Qj of the third region Vq corresponding to each advanced building Bk, finds the rectangle edge point Qjd that is closest to the third edge point Pj, and obtains the rectangle edge point set {Q1d,Q2d,..,Qjd,...,Qwd}; where d = 1 to 4; and uses each rectangle edge point Qjd in the rectangle edge point set {Q1d,Q2d,..,Qjd,...,Qwd} as the optimized position of the corresponding third edge point Pj in the third edge point set {P1,P2,...,Pj..,Pw}.

[0023] S67 sequentially connects adjacent rectangular edge points in the rectangular edge point set {Q1d,Q2d,..,Qjd,...,Qwd}, and connects the rectangular edge points with the adjacent first edge points that do not belong to the third edge point set, forming multiple intermediate tessellation lines. These multiple intermediate tessellation lines constitute an intermediate tessellation line network.

[0024] Specifically, the preset threshold mentioned in step S63 or S65 is 50m.

[0025] Specifically, the road vector data in step S7 includes multiple road vector lines.

[0026] Specifically, step S7 further includes:

[0027] Based on road vector data, S71 extracts intermediate mosaic lines that intersect with a road vector line and have an angle of less than 45° from the intermediate mosaic line network as intermediate mosaic lines to be optimized, thus obtaining a set of intermediate mosaic lines to be optimized; at the same time, the two end points of each intermediate mosaic line to be optimized in the set of intermediate mosaic lines to be optimized are recorded as nodes to be optimized.

[0028] S72 constructs perpendicular lines to the road vector lines that intersect each of the two nodes to be optimized for each intermediate mosaic line. The two intersection points formed by the two perpendicular lines and the road vector lines are taken as the two optimized nodes.

[0029] The two optimized intersections of S73 are connected to form an optimized intermediate tessellation line;

[0030] S74 Repeat steps S72-S73 to obtain the optimized intermediate tessellation lines corresponding to all intermediate tessellation lines to be optimized.

[0031] S75 replaces all intermediate tessellation lines to be optimized with the corresponding optimized intermediate tessellation lines to obtain the advanced tessellation line network.

[0032] Specifically, in step S1, the eight-neighbor boundary tracing algorithm is used to extract the effective regions of each orthophoto in the image set in turn; and the Douglas-Peucker algorithm is used to simplify and thin all effective regions to obtain the effective convex polygons of each orthophoto.

[0033] Specifically, step S2 further includes:

[0034] S21 uses a polygon clipping algorithm to calculate the first intersection region between the effective convex polygons of a pair of adjacent orthophotos;

[0035] S22 performs a search from left to right and from bottom to top in the first intersecting region to find the first valid pixel, which is taken as the first intersection point;

[0036] S23, based on the first intersection point, sets the search azimuth angle of this intersection point to 0° and defines the initial search direction as the upper left; along the upper left of the first intersection region, it sequentially searches for valid pixels; when the search azimuth angle is < 90°, the found valid pixels are used as upper boundary points to form an upper boundary point set; when the search azimuth angle is ≥ 90° and < 180°, the found valid pixels are used as another intersection point; when the search azimuth angle is ≥ 180°, the found valid pixels are used as lower boundary points to form a lower boundary point set.

[0037] S24 The upper boundary point set is formed by two adjacent upper boundary points, and multiple upper boundary lines form an upper boundary line set; the lower boundary set is formed by two adjacent lower boundary points, and multiple lower boundary lines form a lower boundary line set; one intersection point forms two special boundary lines with the nearest upper boundary point and the lower boundary line respectively, and another focus forms two special boundary lines with the nearest upper boundary point and the lower boundary line respectively.

[0038] S25 constructs a constrained triangulation within the first intersecting region based on the upper boundary point set, lower boundary point set, intersection point set, upper boundary line set, lower boundary line set, and special boundary line set of the first intersecting region.

[0039] Specifically, the method for constructing the constrained triangulation in step S25 is as follows: the first intersection region between the effective convex polygons of a pair of adjacent orthophotos is taken as the primary triangulation; a set of constraint edges is added to the primary triangulation to obtain the constrained triangulation; the set of constraint edges includes multiple pairs of constraint edges, each pair of constraint edges includes two constraint edges and the two constraint edges intersect at an upper boundary point; each pair of constraint edges is formed by connecting an upper boundary point and two adjacent lower boundary points respectively; the lower boundary line formed by connecting each pair of constraint edges and two adjacent lower boundary points constitutes a triangular facet.

[0040] The beneficial effects of this invention are as follows:

[0041] (1) The method for generating orthophoto mosaic lines of the present invention transforms the problem of image topology search into a polygon selection problem of triangles by constructing constrained triangular meshes in the intersecting regions between the effective regions of adjacent orthophotos in sequence, thus avoiding direct local topology solving of the images. The generated primary mosaic line network is the global optimal solution. This method of dividing the intersecting regions by constrained triangular meshes is unique, non-redundant, and seamless. It is suitable for mosaicking large-scale multiple orthophotos, combining the flexibility and efficiency of mosaicking, avoiding the accumulation of errors, and the mosaicking process is independent of the order of the orthophotos.

[0042] (2) This invention optimizes the primary mosaicking network using building vector data. The optimization algorithm involved in this invention only involves the calculation of points, lines, and polygons, which has the advantages of low computational cost, higher accuracy, and higher efficiency. It can quickly obtain a mosaicking network that bypasses the building area, thus solving the problems of texture misalignment and low selection efficiency caused by the mosaicking network's inability to automatically bypass the building imaging area.

[0043] (3) The present invention further optimizes the intermediate mosaic network by using road vector data, so that the mosaic lines are more closely aligned with the road vector lines. The roads in the mosaicked orthophoto will appear more coherent and natural, without any obvious breaks or misalignments, thus improving the overall quality of the orthophoto mosaic. Attached Figure Description

[0044] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0045] Figure 1 is a technical flowchart of the method for generating mosaic lines in remote sensing orthophotos in this invention.

[0046] Figure 2 is a technical framework diagram of the primary mosaic line generation method for a pair of adjacent orthophotos in this invention;

[0047] Figure 3 is a schematic diagram of generating a primary tessellation network in this invention;

[0048] Figure 4 is a schematic diagram of the principle of calculating the first region of a building in the corresponding orthophoto in this invention;

[0049] Figure 5 is a schematic diagram of the intersection area of ​​multiple orthophotos of the second building in this invention. Detailed Implementation

[0050] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention are within the scope of protection of the present invention.

[0051] The invention will now be further described with reference to the accompanying drawings:

[0052] Please refer to Figure 1, which is a technical flowchart of a method for generating mosaic lines in remote sensing orthophotos. The method specifically includes the following steps:

[0053] S1 uses a tracking algorithm to sequentially obtain the effective regions of each orthophoto in the image set, and simplifies the effective regions to obtain the effective convex polygons of each orthophoto; the image set includes multiple orthophotos, multiple original images and corresponding DEM data; wherein multiple orthophotos constitute multiple pairs of adjacent orthophotos.

[0054] In an embodiment of the present invention, step S1 uses an eight-neighbor boundary tracing algorithm to sequentially extract the effective regions of each orthophoto in the image set; and uses the Douglas-Peucker algorithm to simplify and thin all effective regions to obtain the effective convex polygons of each orthophoto.

[0055] In embodiments of the present invention, an eight-neighbor boundary tracing algorithm is used to extract the effective region from the eight neighborhoods of pixels in each orthophoto. For any orthophoto, the search method is to search in eight directions for a pixel: left, upper left, upper, upper right, right, lower right, lower, and lower left, with an angle of 45 degrees between adjacent search directions. For each orthophoto, the search is performed in the order from left to right and from bottom to top to find the first effective pixel, which is used as the initial upper boundary point, i.e., the upper left boundary point. Based on the initial upper boundary point, the initial search direction is defined as the upper left. Valid pixels are searched sequentially along the upper left. When the search azimuth angle is <90°, the found valid pixels are used as the initial upper boundary point, forming the initial upper boundary point set. When the search azimuth angle is ≥90° and <180°, the two found valid pixels are used as two intersection points, forming the initial intersection point. When the search azimuth angle is ≥180°, the found valid pixels are used as the initial lower boundary point, forming the initial lower boundary point set. This process continues until the initial upper boundary point is returned. Connecting all the initial upper boundary point sets, intersection points, and initial lower boundary point sets obtained from these searches forms the effective region of the orthophoto. Boundary point search is a common technique in this field and will not be elaborated upon further.

[0056] An excessive number of boundary points hinders subsequent polygon intersection operations. To reduce computational complexity, this invention employs the Douglas-Peuker algorithm to thin the initial upper and lower boundary point sets. Since the effective region boundary of the orthophoto is a closed convex polygon, a starting point and the farthest boundary point can be selected from the point set. The convex polygon is then split into two curve segments for thinning, yielding the effective convex polygon of the orthophoto.

[0057] S2 identifies a first intersection region between effective convex polygons of a pair of adjacent orthophotos, the first intersection region including different types of boundary point sets and different types of boundary line sets; and constructs a constrained triangular mesh based on the different types of boundary point sets and boundary line sets in the first intersection region; the constrained triangular mesh includes multiple triangular patches; wherein different types of boundary point sets constitute different types of boundary line sets.

[0058] In an embodiment of the present invention, the different types of boundary point sets in step S2 include upper boundary point sets, lower boundary line point sets, and intersection point sets; different types of boundary lines include upper boundary line sets, lower boundary line sets, and special boundary line sets; two adjacent upper boundary points in the upper boundary point set constitute an upper boundary line; two adjacent lower boundary points in the lower boundary point set constitute a lower boundary line; multiple upper boundary lines constitute an upper boundary line set; multiple lower boundary lines constitute a lower boundary line set; and boundary lines containing intersection points are designated as special boundary lines.

[0059] Please refer to Figures 2 and 3. Figure 2 is a technical framework diagram of the method for generating primary mosaic lines from a pair of adjacent orthophotos in this invention; Figure 3 is a schematic diagram of generating a primary mosaic line network in this invention.

[0060] In an embodiment of the present invention, step S2 further includes:

[0061] S21 uses a polygon clipping algorithm to calculate the first intersection region EFGHIJKMNOPQR between the effective convex polygons of a pair of adjacent orthophotos.

[0062] S22 performs a sequential search from left to right and from bottom to top on the first intersecting region EFGHIJKMNOPQR to find the first valid pixel, which is taken as the first intersection point E;

[0063] S23, based on the first intersection point E, sets the search azimuth angle of this intersection point to 0° and defines the initial search direction as the upper left. Along the upper left of the first intersection region, valid pixel points are searched sequentially. When the search azimuth angle is < 90°, the found valid pixel points are used as upper boundary points, forming an upper boundary point set, as shown in Figure 3. The upper boundary point set includes upper boundary points F, G, H, I, and J. When the search azimuth angle is ≥ 90° and < 180°, the found valid pixel point is another intersection point K. When the search azimuth angle is ≥ 180°, the found valid pixel points are used as lower boundary points, forming a lower boundary point set, as shown in Figure 3. The lower boundary point set includes lower boundary points M, N, O, P, Q, and R.

[0064] S24 The upper boundary point set is formed by two adjacent upper boundary points, and multiple upper boundary lines form an upper boundary line set; the lower boundary set is formed by two adjacent lower boundary points, and multiple lower boundary lines form a lower boundary line set; one intersection point forms two special boundary lines with the nearest upper boundary point and the lower boundary line respectively, and another focus forms two special boundary lines with the nearest upper boundary point and the lower boundary line respectively.

[0065] As shown in Figure 3, adjacent upper boundary points F and G form an upper boundary line FG, and adjacent lower boundary points M and N form a lower boundary line MN. The set of upper boundary lines includes FG, GH, HI, and IJ; the set of lower boundary lines includes lower MN, NO, OP, PQ, and QR; and the special boundary lines include EF, ER, KM, and KJ.

[0066] S25 constructs a constrained triangulation within the first intersecting region based on the upper boundary point set, lower boundary point set, intersection point set, upper boundary line set, lower boundary line set, and special boundary line set of the first intersecting region.

[0067] In an embodiment of the present invention, the specific method for constructing a constrained triangulation in step S25 is as follows: the first intersection region EFGHIJKMNOPQR between effective convex polygons of a pair of adjacent orthophotos is used as a primary triangulation; a constrained edge set is added to the primary triangulation to obtain a constrained triangulation; as shown in Figure 3, the constrained edge set includes FR, FQ, GQ, GP, HP, HO, IO, IN, JN, and JM; the constrained edge set includes multiple pairs of constrained edge pairs, each pair of constrained edge pairs includes two constrained edges and the two constrained edges intersect at an upper boundary point; each pair of constrained edges is formed by connecting an upper boundary point and two adjacent lower boundary points respectively; the lower boundary line formed by connecting each pair of constrained edges and two adjacent lower boundary points constitutes a triangular facet, as shown in Figure 3, a pair of constrained edge pairs includes two constrained edges, FR and FQ, which intersect at an upper boundary point F, and FR and FQ are formed by connecting the upper boundary point F and two adjacent lower boundary points Q and R respectively. Two adjacent lower boundary points Q and R are connected to form a lower boundary line QR and a pair of constraint edges FR and FQ, forming a triangular patch FQR.

[0068] S3 determines the boundary point set category to which each vertex in each triangular facet belongs in the constrained triangular network, removes triangular facets where any vertex belongs to the intersection point set and where all vertices belong to the same boundary point set category, and extracts the target triangular facet set.

[0069] In an embodiment of the present invention, the specific method for extracting the target triangular facet set in step S3 is as follows:

[0070] In the constrained triangular mesh, the boundary point set category to which each vertex belongs is determined. Triangular faces with any vertex belonging to the intersection category or all vertices belonging to the same boundary point set category are removed, thus extracting the target triangular face set. As shown in Figure 3, the target triangular face set includes triangular faces FQR, GPQ, HOP, INO, and JMN.

[0071] S4 takes the midpoints of two non-boundary lines in each target triangular facet to obtain a midpoint set; the midpoint set includes multiple adjacent midpoints and two terminal midpoints; then, the adjacent midpoints in the midpoint set are connected pairwise, and the two terminal midpoints are connected to the nearest intersection point in the boundary point set of the first intersection region, respectively, to obtain a pair of primary mosaic lines for adjacent orthophotos. 。

[0072] S5 Repeat steps S2 to S4 to connect all the initial mosaic lines of adjacent orthophotos to form a primary mosaic line network.

[0073] S6 optimizes the primary tessellation network based on building vector data to obtain an intermediate tessellation network; the intermediate tessellation network includes multiple intermediate tessellation lines.

[0074] In an embodiment of the present invention, step S6 further includes:

[0075] S61 obtains the original building positioning model based on the original building vector data corresponding to each original building in multiple orthophotos, and uses the model to calculate the first region of each original building in the corresponding orthophoto in turn; the first region includes multiple edge points, denoted as the first edge point set.

[0076] Please refer to Figure 4, which is a schematic diagram of the principle of calculating the first region of the original building in the orthophoto in this invention; in the embodiments of this invention, the method for obtaining the positioning model of the original building specifically includes:

[0077] (1) The original building vector data is combined with the DEM to automatically generate the original building positioning model. Specifically, in the photogrammetric stereo mapping stage, the original building vector lines are obtained by constructing a stereo model through stereo image pairs and then performing stereo mapping. These vector lines not only have two-dimensional XY coordinates but also elevation coordinates Z, which are three-dimensional descriptions of the original building. If the original building vector polygon is projected vertically downwards onto the DEM of the corresponding original image, a simplified three-dimensional model of the original building can be obtained. For example, let ABCD be the original building vector polygon. Projecting points A, B, C, and D onto the DEM respectively yields points a, b, c, and d. Then, ABCD-abcd is the positioning model of the original building. There are many types of storage models for three-dimensional models, such as parametric models, CAD models, CSG models, and polyhedral structure models. Among them, the polyhedral structure model uses the topological relationships between one or more connected facets to construct the geometric structure of the entire original building, and then combines it with vertical walls to obtain a complete three-dimensional model of the original building. These facets that constitute the original building are usually represented by a triangular mesh in the form of triangular facets. Therefore, in this invention, the original building positioning model is stored in the form of a polyhedral structure model.

[0078] (2) Calculate the first region of each original building on the original image using the original building positioning model. The specific method is as follows:

[0079] ① Project all the triangular faces of the original building positioning model onto the image to obtain all the triangular faces on the original image;

[0080] ② Find the union of all triangular faces on the original image to obtain the first region of the original building in the corresponding original image;

[0081] Find the union of all triangular faces on the original image to obtain the first region of the original building on the original image. Point O in the figure is the photography center. The cuboid ABCD-abcd is assumed to be the positioning model of the original building. All its triangular faces are projected onto the original image according to formula (1). Then, the union is found to obtain the first region A'B'C'cba of the original building on the original image (each endpoint in the first region is represented by image coordinates).

[0082] In equation (1), (X,Y,Z) are the object coordinates of the point, (x,y) are the image coordinates of the point in the image plane coordinate system, and (X,Y,Z) are the object coordinates of the point. S ,Y S Z S ) represents the object coordinates of the photography center, f is the camera focal length, and a1, a2, a3, b1, b2, b3, c1, c2, c3 are the nine rotation parameters of the photogrammetric rotation matrix.

[0083] ③ Based on the DEM iterative approximation solution, the first region of the original building on the corresponding orthophoto is obtained from the first region on the original image; each corner point of the polygon A'B'C'cba on the original image is obtained according to Equation (2) and based on the DEM iterative approximation solution to obtain the corresponding region of the polygon on the orthophoto, that is, the first region of the building on the corresponding orthophoto A”B”C”D”c”b”a.

[0084] In equation (2), (X,Y,Z) are the object coordinates of the point, (x,y) are the image coordinates of the point in the image plane coordinate system, and (X,Y,Z) are the object coordinates of the point. S ,Y S Z S ) represents the object coordinates of the photography center, f is the camera focal length, and a1, a2, a3, b1, b2, b3, c1, c2, c3 are the nine rotation parameters of the photogrammetric rotation matrix.

[0085] S62 will select the original buildings that exist in the first region in multiple orthophotos to form a primary building set. Then, the common intersection of each target building in the primary building set with the multiple first regions in multiple orthophotos will be obtained to obtain the target intersection region of each target building in multiple orthophotos. The multiple edge points in the target intersection region will be denoted as the second edge point set.

[0086] In embodiments of the present invention, due to different photography centers, the first region of an original building in each orthophoto is also different. Please refer to Figure 5, which is a schematic diagram of the intersection region of the original building in multiple orthophotos in the present invention. O1 and O2 are the photography centers of two adjacent images. Assuming that the cuboid ABCD-abcd is the original building, the first region of the original building in orthophoto O1 is polygon A'B'C'cba, and the first region in orthophoto O2 is polygon B”C”D”dab. Therefore, the intersection region of the original building in multiple orthophotos is polygon A'B'EB”C”D”da. Each second node can be shared by a maximum of 4 valid convex polygons of orthophotos, that is, the maximum number of orthophotos with shared second nodes in the first region is 4.

[0087] S63 sequentially filters all original buildings within a preset threshold range, centered on each second edge point, to form a set of intermediate buildings {R1, R2, ..., Rn}. For each intermediate building in the set, its first region on different orthophotos is calculated, forming a set of first regions for each intermediate building. The common intersection of all the first regions in the set of first regions of each intermediate building is taken to obtain the second region of each intermediate building in multiple orthophotos. The second regions of all intermediate buildings in multiple orthophotos constitute a set of second regions {S1, S2, ..., Sn}. The correspondence between Ri and Si is recorded, where i = 1 to n.

[0088] S64 performs an intersection test on the second region set {S1,S2,...,Sn} for all intermediate buildings' second regions, merges the second intersecting regions that intersect each other to form a third region V1, and takes the non-intersecting second regions as other third regions to form the third region set {V1,V2,...,Vm (m≤n)}.

[0089] S65 extracts all second edge points concentrated in the third region set {V1,V2,...,Vq,...,Vm} to form the third edge point set {P1,P2,...,Pj,...,Pw}; sequentially, taking each third edge point Pj in the third edge point set {P1,P2,...,Pj,...,Pw} as the center, it searches for high-level buildings {B1,B2,...,Bk,...,Bt(t≤n)} within a preset threshold range, and at the same time finds the third region Vq corresponding to each high-level building Bk, where k=1~t, q=1~m, j=1~w;

[0090] S66 sequentially calculates the minimum bounding rectangle Qj of the third region Vq corresponding to each advanced building B, finds the rectangle edge point Qjd that is closest to the third edge point Pj, and obtains the rectangle edge point set {Q1d,Q2d,..,Qjd,...,Qwd}; where d = 1 to 4; and uses each rectangle edge point Qjd in the rectangle edge point set {Q1d,Q2d,..,Qjd,...,Qwd} as the optimized position of the corresponding third edge point Pj in the third edge point set {P1,P2,...,Pj..,Pw}.

[0091] S67 sequentially connects adjacent rectangular edge points in the rectangular edge point set {Q1d,Q2d,..,Qjd,...,Qwd}, and connects the rectangular edge points with the adjacent first edge points that do not belong to the third edge point set, forming multiple intermediate tessellation lines. These multiple intermediate tessellation lines constitute an intermediate tessellation line network.

[0092] In the embodiments of the present invention, the intermediate mosaicking network obtained after the above steps has edge points that are not within the first region of the original building in the corresponding original image. In step 6, all the first edge points in the primary mosaicking network that fall within the first region of the original building have been moved out of the first region of the original building, ensuring that the endpoints of each mosaicking line are outside the first region of the original building. However, each mosaicking line is shared by the effective convex polygons of two adjacent images. Therefore, for the optimization of the mosaicking line automatically bypassing the building, it is necessary to consider both the case of the original building shared by the two images and the selection of the optimal path.

[0093] In an embodiment of the present invention, the preset threshold in step S63 or S65 is 50m.

[0094] S7 optimizes the intermediate mosaic line network based on road vector data to obtain the advanced mosaic line network.

[0095] In an embodiment of the present invention, the road vector data in step S7 includes multiple road vector lines.

[0096] In an embodiment of the present invention, step S7 further includes:

[0097] Based on road vector data, S71 extracts intermediate mosaic lines that intersect with a road vector line and have an angle of less than 45° from the intermediate mosaic line network as intermediate mosaic lines to be optimized, thus obtaining a set of intermediate mosaic lines to be optimized; at the same time, the two end points of each intermediate mosaic line to be optimized in the set of intermediate mosaic lines to be optimized are recorded as nodes to be optimized.

[0098] S72 constructs perpendicular lines to the road vector lines that intersect each of the two nodes to be optimized for each intermediate mosaic line. The two intersection points formed by the two perpendicular lines and the road vector lines are taken as the two optimized nodes.

[0099] The two optimized intersections of S73 are connected to form an optimized intermediate tessellation line;

[0100] S74 Repeat steps S72-S73 to obtain the optimized intermediate tessellation lines corresponding to all intermediate tessellation lines to be optimized.

[0101] S75 replaces all intermediate tessellation lines to be optimized with the corresponding optimized intermediate tessellation lines to obtain the advanced tessellation line network.

[0102] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for generating mosaic lines in a remotely sensed orthophoto image, characterized in that, Includes the following steps: S1 uses a tracking algorithm to sequentially obtain the effective regions of each orthophoto in the image set, and simplifies the effective regions to obtain the effective convex polygons of each orthophoto; the image set includes multiple orthophotos, multiple original images and corresponding DEM data; Multiple orthophotos form multiple pairs of adjacent orthophotos; S2 identifies a first intersection region between effective convex polygons of a pair of adjacent orthophotos, the first intersection region including different types of boundary point sets and different types of boundary line sets; and constructs a constrained triangular mesh based on the different types of boundary point sets and boundary line sets in the first intersection region; the constrained triangular mesh includes multiple triangular patches; wherein different types of boundary point sets constitute different types of boundary line sets; S3 extracts the target triangular facet set by determining the boundary point set to which each vertex in each triangular facet of the constrained triangular mesh belongs; the target triangular facet set includes multiple target triangular facets; each target triangular facet consists of one boundary line and two non-boundary lines; S4 takes the midpoint of two non-boundary lines in each target triangular facet to obtain a midpoint set; the midpoint set includes multiple adjacent midpoints and two end midpoints; and sequentially connects the adjacent midpoints in the midpoint set to each other, and connects the two end midpoints to the nearest intersection point in the boundary point set in the first intersection region to obtain a pair of primary mosaic lines of adjacent orthophotos. S5 Repeat steps S2 to S4 to connect all the initial mosaic lines of adjacent orthophotos to form a primary mosaic line network. S6 optimizes the primary tessellation network based on building vector data to obtain an intermediate tessellation network; the intermediate tessellation network includes multiple intermediate tessellation lines; S7 optimizes the intermediate mosaic line network based on road vector data to obtain the advanced mosaic line network.

2. The method for generating mosaic lines in a remote sensing orthophoto image according to claim 1, characterized in that, The different types of boundary point sets mentioned in step S2 include upper boundary point sets, lower boundary point sets, and intersection point sets; the different types of boundary lines include upper boundary line sets, lower boundary line sets, and special boundary line sets; two adjacent upper boundary points in the upper boundary point set constitute an upper boundary line; two adjacent lower boundary points in the lower boundary point set constitute a lower boundary line; multiple upper boundary lines constitute an upper boundary line set; multiple lower boundary lines constitute a lower boundary line set; and boundary lines containing intersection points are considered special boundary lines.

3. The method for generating mosaic lines in a remote sensing orthophoto image according to claim 2, characterized in that, The specific method for extracting the target triangular facet set in step S3 is as follows: Determine the boundary point set category to which each vertex in each triangular facet belongs in the constrained triangular network, remove triangular faces where any vertex belongs to the intersection point set and where all vertices belong to the same boundary point set category, and extract the target triangular facet set.

4. The method for generating mosaic lines in a remote sensing orthophoto image according to claim 1, characterized in that, Step S6 further includes: S61 obtains the original building positioning model based on the original building vector data corresponding to each original building in multiple orthophotos, and uses the model to calculate the first region of each original building in the corresponding orthophoto in turn; the first region includes multiple edge points, denoted as the first edge point set; S62 will select the original buildings that exist in the first region in multiple orthophotos to form a primary building set. Then, the common intersection of each target building in the primary building set with the multiple first regions in multiple orthophotos will be obtained to obtain the target intersection region of each target building in multiple orthophotos. The multiple edge points in the target intersection region will be denoted as the second edge point set. S63 sequentially filters all original buildings within a preset threshold range, centered on each second edge point, to form a set of intermediate buildings {R1, R2, ..., Rn}. For each intermediate building in the set, its first region on different orthophotos is calculated, forming a set of first regions for each intermediate building. The common intersection of all the first regions in the set of first regions of each intermediate building is taken to obtain the second region of each intermediate building in multiple orthophotos. The second regions of all intermediate buildings in multiple orthophotos constitute a set of second regions {S1, S2, ..., Sn}. The correspondence between Ri and Si is recorded, where i = 1 to n. S64 performs an intersection test on the second region set {S1,S2,...,Sn} for all intermediate buildings' second regions, merges the second intersecting regions that intersect each other to form a third region V1, and takes the non-intersecting second regions as other third regions to form the third region set {V1,V2,...,Vm (m≤n)}. S65 extracts all second edge points concentrated in the third region set {V1,V2,...,Vq,...,Vm} to form the third edge point set {P1,P2,...,Pj,...,Pw}; sequentially, taking each third edge point Pj in the third edge point set {P1,P2,...,Pj,...,Pw} as the center, it searches for high-level buildings {B1,B2,...,Bk,...,Bt(t≤n)} within a preset threshold range, and at the same time finds the third region Vq corresponding to each high-level building Bk, where k=1~t, q=1~m, j=1~w; When S66 takes the third edge point Pj as the center, it sequentially calculates the minimum bounding rectangle Qj of the third region Vq corresponding to each advanced building Bk within the preset threshold range, finds the rectangle edge point Qjd that is the smallest distance from the third edge point Pj, and obtains the rectangle edge point set {Q1d,Q2d,..,Qjd,...,Qwd}; where d = 1 to 4; and takes each rectangle edge point Qjd in the rectangle edge point set {Q1d,Q2d,..,Qjd,...,Qwd} as the optimized position of the corresponding third edge point Pj in the third edge point set {P1,P2,...,Pj..,Pw}. S67 sequentially connects adjacent rectangular edge points in the rectangular edge point set {Q1d,Q2d,..,Qjd,...,Qwd}, and connects the rectangular edge points with the adjacent first edge points that do not belong to the third edge point set, forming multiple intermediate tessellation lines. These multiple intermediate tessellation lines constitute an intermediate tessellation line network.

5. The method for generating mosaic lines in a remote sensing orthophoto image according to claim 4, characterized in that, The preset threshold mentioned in step S63 or S65 is 50m.

6. The method for generating mosaic lines in a remote sensing orthophoto image according to claim 1, characterized in that, The road vector data in step S7 includes multiple road vector lines.

7. The method for generating mosaic lines in a remote sensing orthophoto image according to claim 6, characterized in that, Step S7 further includes: Based on road vector data, S71 extracts intermediate mosaic lines that intersect with a road vector line and have an angle of less than 45° from the intermediate mosaic line network as intermediate mosaic lines to be optimized, thus obtaining a set of intermediate mosaic lines to be optimized; at the same time, the two end points of each intermediate mosaic line to be optimized in the set of intermediate mosaic lines to be optimized are recorded as nodes to be optimized. S72 constructs perpendicular lines to the road vector lines that intersect each of the two nodes to be optimized for each intermediate mosaic line. The two intersection points formed by the two perpendicular lines and the road vector lines are taken as the two optimized nodes. The two optimized intersections of S73 are connected to form an optimized intermediate tessellation line; S74 Repeat steps S72-S73 to obtain the optimized intermediate tessellation lines corresponding to all intermediate tessellation lines to be optimized. S75 replaces all intermediate tessellation lines to be optimized with the corresponding optimized intermediate tessellation lines to obtain the advanced tessellation line network.

8. The method for generating mosaic lines in a remote sensing orthophoto image according to claim 1, characterized in that, In step S1, the eight-neighbor boundary tracing algorithm is used to extract the effective regions of each orthophoto in the image set in turn; and the Douglas-Peucker algorithm is used to simplify and thin all effective regions to obtain the effective convex polygons of each orthophoto.

9. The method for generating mosaic lines in a remote sensing orthophoto image according to claim 2, characterized in that, Step S2 further includes: S21 uses a polygon clipping algorithm to calculate the first intersection region between the effective convex polygons of a pair of adjacent orthophotos; S22 performs a search from left to right and from bottom to top in the first intersecting region to find the first valid pixel, which is taken as the first intersection point; S23, based on the first intersection point, sets the search azimuth angle of this intersection point to 0° and defines the initial search direction as the upper left; along the upper left of the first intersection region, it sequentially searches for valid pixels; when the search azimuth angle is < 90°, the found valid pixels are used as upper boundary points to form an upper boundary point set; when the search azimuth angle is ≥ 90° and < 180°, the found valid pixels are used as another intersection point; when the search azimuth angle is ≥ 180°, the found valid pixels are used as lower boundary points to form a lower boundary point set. S24 The upper boundary point set is formed by two adjacent upper boundary points forming an upper boundary line, and multiple upper boundary lines form an upper boundary line set; the lower boundary point set is formed by two adjacent lower boundary points forming a lower boundary line, and multiple lower boundary lines form a lower boundary line set; one intersection point forms two special boundary lines with the nearest upper boundary point and the next lower boundary point respectively, and another intersection point forms two special boundary lines with the nearest upper boundary point and the next lower boundary point respectively; S25 constructs a constrained triangulation within the first intersecting region based on the upper boundary point set, lower boundary point set, intersection point set, upper boundary line set, lower boundary line set, and special boundary line set of the first intersecting region.

10. The method for generating mosaic lines in a remote sensing orthophoto image according to claim 9, characterized in that, The specific method for constructing the constrained triangulation in step S25 is as follows: the first intersection region between the effective convex polygons of a pair of adjacent orthophotos is taken as the primary triangulation; a set of constraint edges is added to the primary triangulation to obtain the constrained triangulation; the set of constraint edges includes multiple pairs of constraint edges, each pair of constraint edges includes two constraint edges and the two constraint edges intersect at an upper boundary point; each pair of constraint edges is formed by connecting an upper boundary point and two adjacent lower boundary points respectively; the lower boundary line formed by connecting each pair of constraint edges and two adjacent lower boundary points constitutes a triangular facet.

Images