Road network center line generation method and device based on adaptive search of road topology

Abstract

The method and device for generating road network center line based on adaptive search of road topology belong to the field of remote sensing image processing, and acquire a segmented binary image of a road remote sensing image; perform morphological processing on the segmented binary image to extract a road skeleton line; represent the road network center line as a graph structure in graph theory, that is, a graph G(V, E), wherein V represents a road node set, and E represents an edge set of connected nodes required for constructing a road network; then simplify the road network center line without changing the original road network topology, and perform optimization processing on the simplified road network to generate a road network center line based on adaptive search of road topology; through image morphological processing and graph theory processing of the road remote sensing image, a method for constructing a high-quality road network center line from a road surface generated by a deep learning algorithm is realized, automatic connection of broken road sections and simplification of redundant road networks can be achieved, and the generated road network has good visual effects and practicability.

Description

Technical Field

[0001] This invention belongs to the field of remote sensing image processing, and particularly relates to a method and apparatus for generating road network centerlines based on adaptive road topology search. Background Technology

[0002] Extracting crucial geographic information from remote sensing imagery has become an essential requirement for modern society. Roads, as key geographic elements, require efficient, real-time, and accurate information acquisition and updating for applications such as urban construction, emergency response, and transportation planning. Fully automated road information extraction from remote sensing imagery can significantly shorten acquisition time and greatly reduce labor costs. In recent years, deep learning has made significant breakthroughs in image processing, and its application in remote sensing interpretation has received widespread attention and in-depth research. However, limited by sample size and computer hardware, road elements automatically extracted from images using deep learning methods inevitably suffer from fragmentation or noise, making it impossible to obtain high-quality road network topology.

[0003] In current technologies, researchers have proposed using post-processing methods to further refine the road network results extracted by deep learning, aiming to improve road network connectivity and remove noise and redundant road segments. For example, some studies have used tensor voting algorithms to improve road network structure, while others have used neural network-based iterative methods to connect disconnected road segments. However, while existing technologies can improve the quality of automatically extracted road networks to some extent, the lack of a holistic consideration of road topology and the absence of targeted analysis for various types of incorrect disconnections and chaotic connections result in the quality of the processed road network topology still needing improvement. Summary of the Invention

[0004] The present invention aims to solve the above problems and provides a method and apparatus for generating road network centerlines based on adaptive road topology search.

[0005] In a first aspect, the present invention provides a method for generating road network centerlines based on adaptive road topology search, comprising: Obtain segmented binary images of road remote sensing images; Morphological processing is performed on the segmented binary image to extract the road skeleton lines; The road skeleton is represented by a graph structure in graph theory, namely graph G(V,E), where V represents the set of road nodes and E represents the set of edges connecting the nodes required to construct the road network. Then, without changing the original road network topology, the road network centerline is simplified, and the simplified road network is optimized to generate a road network centerline based on adaptive search of road topology. The optimization process includes: removing burrs from the road network; finding disconnected points in the road network and automatically connecting them using the probability values ​​of the aforementioned segmented binary graph; automatically completing the road network connections between nodes; regularizing the misaligned intersections of the road network; converting the loop topology in the road network into a star topology; regularizing nodes with a degree of 3 in the road network using geometric relationships; and finally removing any existing noisy loop topologies.

[0006] Furthermore, the road network centerline generation method based on adaptive road topology search of the present invention includes the following steps for simplifying the road network centerline: Step 201: For a curve formed by a set of connected nodes V and the edges E connecting the nodes in the road network centerline diagram structure, take the endpoints A and B of the curve and obtain the straight line AB. The straight line AB is called the chord of the curve. Step 202: Obtain the point C on the curve that is furthest from the straight line segment, and calculate its distance d from AB; Step 203: Compare the distance with a pre-defined threshold. If it is less than the threshold, the straight line segment is used as an approximation of the curve. Once the curve segment is processed, proceed to the next curve segment. Step 204: If the distance is greater than the threshold, divide the curve into two segments AC and BC using point C, and process each segment separately according to steps 201-203. Step 205: Once all curves have been processed, connect the dividing points sequentially to form a broken line, which can be used as an approximation of the curve for processing the next segment of the curve.

[0007] Furthermore, the road network centerline generation method based on adaptive road topology search of the present invention, wherein finding the disconnected points in the road network and automatically connecting them by combining the probability values ​​of the aforementioned segmented binary map includes the following steps: Step 301: Traverse all nodes in graph G(V,E), determine the degree of the current node, and record the degree of the current node as v if the degree of the current node is 1. At the same time, perform the following steps. Step 302: Create a window of size (T1,T2) centered on node v with current degree 1, where T is a preset threshold, and find all edges of graph G(V,E) within this window range, denoted as e; Step 303: For the current node v, traverse all edges e that meet the requirements, calculate the projection point p of node v on edge e, and find the optimal edge best_e through the constraints. If best_e is not empty, proceed to the following steps; otherwise, go to step 301 to continue traversing the next node; Step 304: Combine the probability values ​​of the road segmentation binary graph to find the shortest connection path and automatically connect the broken road network, specifically including: Step 3041: Starting from the starting point p, treat p as a square waiting to be checked and put it into the "open list". The "open list" is a list that stores squares waiting to be checked. Step 3042: Find the reachable squares around the starting point p, put them into the "Open List", and set their parent square to p; Step 3043: Delete the starting point p from the "Open List" and put p into the "Close List". The "Close List" contains the squares that do not need to be checked. Step 3044: Calculate the F value for each square, where the F value is calculated using the following formula: F = M + H + (1 - P) * k; where M represents the movement cost from the starting point to the specified square; H represents the estimated cost from the specified square to the target point v; P represents the road segmentation probability value of the square to be moved to; and k is a constant value. Step 3045: Select the cell 'a' with the lowest F value from the "Open List", delete it from the "Open List", and add it to the "Close List". Step 3046: Check all adjacent and reachable squares of cell a; squares in the "closed list" are not considered; if these squares are not yet in the "open list", add them to the "open list", calculate the F value of these squares, and set the parent square to a; if an adjacent square c is already in the "open list", calculate the new path from p to square c, i.e., the path through a, and determine whether an update is needed, i.e., whether the M value is lower; if the new M value is lower, change the parent square to square a, recalculate the F value, and the H value does not need to be changed because the expected cost of the target point is fixed; if the new M value is higher, the value remains unchanged. Step 3047: Continue to find the one with the smallest F value from the "Open List", delete it from the "Open List", add it to the "Closed List", and continue to find the surrounding reachable blocks, and repeat this process. Step 3048: When the target block v appears in the "Open List", it means that the optimal path has been found. When there is no data in the "Open List", it means that there is no suitable path. Step 305: Based on the shortest connection path, establish new nodes and edges in graph G(V,E), that is, connect the starting point p to the target point v; at the same time, go to step 301 to continue traversing the next node until all nodes in graph G(V,E) have been traversed.

[0008] Furthermore, the road network centerline generation method based on adaptive road topology search of the present invention includes the following constraints: the node of the edge is not near node v; the angle formed by the edge where node v is located and the projection point p to node v is less than a certain threshold; and the distance from the projection point p to node v is the shortest.

[0009] Furthermore, the road network centerline generation method based on adaptive road topology search of the present invention includes the following steps for automatically completing the road network connections between nodes: Step 501: Traverse all nodes in graph G(V,E), determine the degree of the current node, and denote the degree of the current node as v when the degree of the current node is 1. Step 502: Create a window of size (T3,T4) centered on the node v with current degree 1, and find all edges of graph G(V,E) within the window range, denoted as e; Step 503: For the current node v, traverse all nodes of all edges e that meet the requirements, and at the same time find the optimal node best_vertex through the node constraints. The node constraints include: the node of the edge is not near node v; the angle between the edge containing node v and the node from node v to the current edge is less than a certain threshold; calculate the intersection of the edge formed by the current edge and node v with its nearby edges to prevent the current edge node from being the same as node v; and minimize the distance between the current edge node and node v. Obtain the best_vertex of the node v with a current degree of 1; if the node best_vertex is not empty, perform the following steps; otherwise, go to step 501 to continue traversing the next node. Step 504: Connect the current node v with best_vertex.

[0010] Furthermore, the road network centerline generation method based on adaptive road topology search of the present invention includes the following steps for the misalignment and normalization of road network intersections: traversing each node with a degree of 3, taking the current node v with a degree of 3 as the starting node, and obtaining the connectivity length d from it to the nearest node with a degree of 3; if d is less than a set threshold T, proceed to the next step; otherwise, traverse the next node with a degree of 3 until the end; calculate the center point of the edge formed by two nodes with a degree of 3; and use the obtained center point as a new node to replace the two nodes with a degree of 3.

[0011] Furthermore, the method for generating the centerline of a road network based on adaptive road topology search according to the present invention includes the following steps in the process of converting the ring topology structure in the road network into a star topology structure: Step 701: Traverse each node with degree 3, create a window of size (T5,T6) centered on the current node v with degree 3, and find all edges in graph G(V,E) that are connected to node v within the window, and treat this local road network graph as a directed graph. Step 702: Let the number of nodes in the directed graph above be n, the number of edges be m, and its adjacency matrix be an n-order square matrix; use a linked list or array to record the visited nodes in order; starting from the node v1 with the current degree of 2, search for a node adjacent to node v1 and determine whether the node has been visited; if it has been visited, it indicates that a cycle has been formed; shrink all the nodes in the cycle to a single point, generate a new adjacency matrix, and continue searching and judging until there are no cycles; Step 703: Calculate the total edge length of the searched ring topology. If the total length is greater than a preset threshold, continue with step 701. If the total length is less than the preset threshold, obtain the center F of the ring topology through geometric relationships, connect each node of the ring topology, delete the original topological connection method, and transform it from a ring topology to a star topology. Then, continue with step 701 until all nodes with a degree of 3 in the road network have been traversed.

[0012] Furthermore, the road network centerline generation method based on adaptive road topology search of the present invention includes two methods for normalizing nodes of degree 3 in the road network through geometric relationships: Method 1 is to normalize nodes of degree 3 according to the length of the edge where the node of degree 3 is located; Method 2 is to normalize nodes of degree 3 according to the included angle of the edge where the node of degree 3 is located. Method 1, which normalizes nodes of degree 3 based on the length of the edge containing the node of degree 3, includes the following steps: Step 8011: Traverse each node with a degree of 3, and denote the current node with a degree of 3 as v; Step 8012: Calculate the length of the edge containing the node v with degree 3, and arrange them in ascending order as e1, e1 with nodes (v1, v), e2, e2 with nodes (v2, v), e3, and e3 with nodes (v3, v). Calculate the angle formed by the two shortest edges e1 and e2, and determine whether the angle is greater than a set threshold. If it is greater than the set threshold, continue to the next step; if it is less than the set threshold, continue traversing the next node with degree 3. Step 8013: Calculate the angle between the vector formed by nodes v3 and v and the vector formed by v1 and v2. If this angle is less than the threshold T7 or greater than the threshold T8, calculate the intersection point p of edge e3 and the line segment formed by v1 and v2, and determine whether the intersection point p exists. If it does not exist, calculate the projection point of node v on the line segment formed by v1 and v2, and make p equal to this projection point. If this angle is greater than the threshold T7 and less than the threshold T8, calculate the projection point of node v3 on the line segment formed by v1 and v2, and make p equal to this projection point. Step 8014: Make v equal to p, that is, replace the node v with degree 3 with p; then, continue with step 8011 until all nodes with degree 3 in the road network have been traversed. Method 2, which normalizes nodes with an angle of 3 on the edges containing nodes with an angle of 3, includes the following steps: Step 8021: Traverse each node with a degree of 3, and denote the current node with a degree of 3 as v; Step 8022: Calculate the angle between any two edges containing a node v with degree 3, and label the two edges with the largest angle as e1, with nodes (v1, v) and e2, and nodes (v2, v) for e2. Label the other edge as e3, with nodes (v3, v). Simultaneously, determine if the angle between e1 and e2 is greater than a set threshold. If it is, continue to the next step; otherwise, continue traversing the next node with degree 3. Step 8023: Calculate the node... The angle between the vector formed by points v3 and v and the vector formed by v1 and v2 is calculated. If this angle is less than the threshold T7 or greater than the threshold T8, the intersection point p of the line segment formed by edge e3 and v1 and v2 is calculated. At the same time, it is determined whether the intersection point p exists. If it does not exist, the projection point of node v on the line segment formed by v1 and v2 is calculated, and p is equal to this projection point. If this angle is greater than the threshold T7 and less than the threshold T8, the projection point of node v3 on the line segment formed by v1 and v2 is calculated, and p is equal to this projection point. Step 8024: Make v equal to p, that is, replace the node v with degree 3 with p; then, continue with step 8021 until all nodes with degree 3 in the road network have been traversed.

[0013] Furthermore, the road network centerline generation method based on adaptive road topology search of the present invention includes the following steps in the process of removing existing noise loop topology structures: Step 901: Traverse each node with degree 2, create a window of size (T5, T6) centered on the current node v with degree 2, and find all edges in graph G(V, E) that are connected to node v within the window range, and treat this local road network graph as a directed graph. Step 902: Let the number of nodes in the directed graph described in the previous step be n, the number of edges be m, and its adjacency matrix be an n-order square matrix; use a linked list or array to record the visited nodes sequentially; starting from the node v with a current degree of 2, search for a node adjacent to node v, and determine whether the node has been visited; if it has been visited, it indicates that a cycle has been formed; shrink all nodes in the cycle to a single point, generate a new adjacency matrix, and continue the search until there are no cycles; Step 903: Calculate the total edge length of the searched ring topology. If the total length is greater than the preset threshold, continue with step 901. If the total length is less than the preset threshold, perform ring-removal topology processing. Then, continue with step 901 until all nodes with a degree of 2 in the road network have been traversed.

[0014] Secondly, the present invention provides a road network centerline generation system based on adaptive road topology search, including an image acquisition module, a skeleton extraction module, a road network simplification module, and a road network optimization module; The image acquisition module is used to acquire segmented binary images of road remote sensing images; The skeleton extraction module is used to perform morphological processing on the segmented binary image to extract the road skeleton lines; The road network simplification module is used to represent the road network centerline using a graph structure in graph theory, i.e., graph G(V,E), where V represents the set of road nodes and E represents the set of edges connecting the nodes required to construct the road network; the road network centerline is simplified without changing the original road network topology. The road network optimization module is used to optimize the simplified road network and generate a road network centerline based on adaptive search of road topology. The optimization process includes: removing burrs from the road network; finding disconnected points in the road network and automatically connecting them using the probability values ​​of the aforementioned segmented binary image; automatically completing the road network connections between nodes; straightening out misaligned intersections in the road network; converting the loop topology in the road network into a star topology; straightening out nodes with a degree of 3 in the road network using geometric relationships; and finally removing existing noisy loop topologies.

[0015] Thirdly, the present invention provides a road network centerline generation device based on adaptive road topology search, comprising a processor and a memory electrically connected to each other; the memory is used to store a computer program; when the processor executes the aforementioned computer program, it can realize the road network centerline generation method based on adaptive road topology search described in the first aspect.

[0016] Fourthly, the present invention provides a computer-readable storage medium storing a computer program; when the computer program is executed, it can implement the road network centerline generation method based on road topology adaptive search described in the first aspect above.

[0017] The road network centerline generation method and apparatus based on adaptive road topology search of the present invention utilizes image morphology processing and graph theory to process road remote sensing images, and realizes the construction of high-quality road network centerlines from road surfaces generated by deep learning algorithms. The method and apparatus of the present invention can automatically connect broken road segments and simplify redundant road networks. The generated road network has good visual effects and practicality. Attached Figure Description

[0018] Figure 1 The remote sensing image described in Embodiment 1 of the present invention; Figure 2 This is a binary image of the road segmentation result from remote sensing imagery as described in Embodiment 1 of the present invention; Figure 3 The result of road segmentation from remote sensing imagery as described in Embodiment 1 of the present invention; Figure 4 This is a schematic diagram of the neighborhood of point p1 as described in Embodiment 1 of the present invention; Figure 5 This is a comparison image of the morphological processing before and after the process described in Embodiment 1 of the present invention; Figure 6 This is a schematic diagram of the road skeleton line result as described in Embodiment 1 of the present invention; Figure 7 This is a schematic diagram of the road network structure described in Embodiment 1 of the present invention; Figure 8 This is a simplified schematic diagram of the road network centerline as described in Embodiment 1 of the present invention; Figure 9 This is a comparison image before and after burr removal as described in Embodiment 1 of the present invention; Figure 10 This is a schematic diagram illustrating the node degree definition as described in Embodiment 1 of the present invention; Figure 11 This is a schematic diagram of the angle formed by the side where node v is located and the projection point p to node v, as described in Embodiment 1 of the present invention. Figure 12 This is a comparison diagram of the fractured connection before and after the fracture, as described in Embodiment 1 of the present invention; Figure 13 These are detailed comparison images of the connection before and after the breakage described in Embodiment 1 of the present invention; Figure 14 This is a schematic diagram of the angle formed by the edge containing node v and the current edge, as described in Embodiment 1 of the present invention. Figure 15 This is a comparison diagram before and after the connection as described in Embodiment 1 of the present invention; Figure 16 This is a comparison diagram of the details before and after the connection as described in Embodiment 1 of the present invention; Figure 17 This is a schematic diagram of the road network crossroads misalignment correction according to Embodiment 1 of the present invention; Figure 18 This is a comparison diagram of the road network intersection before and after the misalignment correction described in Embodiment 1 of the present invention; Figure 19 This is a schematic diagram illustrating the transformation of the road network ring topology into a star topology according to Embodiment 1 of the present invention; Figure 20 This is a comparison diagram of the road network before and after the ring topology structure described in Embodiment 1 of the present invention is transformed into a star topology structure; Figure 21 This is a schematic diagram of a node with a regularity of 3 as described in Embodiment 1 of the present invention; Figure 22 This is a before-and-after comparison of nodes with a regularity of 3 in Embodiment 1 of the present invention; Figure 23 This is a schematic diagram of a loop-free topology for a network according to an embodiment of the present invention. Figure 1 ; Figure 24 This is a schematic diagram of a loop-free topology for a network according to an embodiment of the present invention. Figure 2 ; Figure 25 This is a schematic diagram of a loop-free topology for a network according to an embodiment of the present invention. Figure 3 ; Figure 26 This is a comparison diagram of the topology before and after removing the noisy loop in Embodiment 1 of the present invention; Figure 27 This is a schematic diagram of the road network centerline generation system based on adaptive road topology search as described in Embodiment 2 of the present invention. Detailed Implementation

[0019] The method and apparatus for generating road network centerlines based on adaptive road topology search according to the present invention will be described in detail below with reference to the accompanying drawings and embodiments. Example 1

[0020] This disclosure presents a method for generating road network centerlines based on adaptive road topology search, comprising the following steps: Figure 1 , Figure 2 , Figure 3 As shown, the road segmentation result of the remote sensing image is obtained by deep learning. The result is a binary image, that is, the foreground pixel value is 255 and the background pixel value is 0.

[0021] For the segmented road binary image, morphological processing is first performed. In this embodiment, dilation is used to remove small holes in the road binary image. Then, based on the morphologically processed result, in this embodiment, a thinning algorithm is used to extract the road skeleton lines. The thinning algorithm is an iterative algorithm, and the entire iterative process is divided into two steps: The first step is to loop through all foreground pixels and delete pixels that meet the following four conditions, i.e., pixel value 255 is changed to pixel value 0: 11)2 <= N(p1) <= 6; 12) S(p1) = 1; 13) p2*p4*p6=0; 14) p4*p6*p8=0; Where N(p1) represents the number of foreground pixels among the 8 pixels adjacent to p1. Condition 2 <= N(p1) ensures that p1 is not an endpoint or an isolated point, and condition N(p1) <= 6 ensures that p1 is a boundary point, not an interior point. S(p1) represents the number of times these points change from 0 to 1 in the order of p2, p3, ..., p9. This condition ensures connectivity after deleting the current pixel. In the first step, conditions 13) and 14) only remove the southeast boundary point, without considering the northwest boundary point. The neighborhood diagram of p1 is shown below. Figure 4 As shown.

[0022] Step 2: After step 1, loop through all foreground pixels and delete pixels that meet the following four conditions, i.e., pixel value 255 is changed to pixel value 0: 21)2 <= N(p1) <= 6; 22)S(p1)=1; 23) p2*p4*p8=0; 24) p2*p6*p8=0; In the second step, conditions 23) and 24) simply involve moving the boundary point in the northwest.

[0023] Morphological processing results as follows Figure 5 As shown, where Figure 5 (a) is a schematic diagram before morphological processing; Figure 5 (b) is a schematic diagram after morphological processing. Road skeleton line results. Figure 6 As shown, where Figure 6 (a) is a schematic diagram of the road skeleton line results; Figure 6 (b) is a schematic diagram of a partial result of the road skeleton line.

[0024] The road skeleton lines obtained through the refinement algorithm are represented using the corresponding graph structure in graph theory, that is, the road network centerline is represented using a graph structure, namely G(V,E) (where V represents the set of road nodes and E represents the set of edges connecting the nodes required to construct the road network). This is the road network graph structure. Figure 7 As shown.

[0025] In this embodiment, the Douglas-Peucker algorithm is used to simplify the centerline of the road network without changing the original road network topology, thereby improving the efficiency of subsequent road network completion algorithms. In this embodiment, the Douglas-Peucker algorithm mainly consists of the following steps: Step 201: For a curve formed by a set of connected nodes V and the edges E connecting the nodes in the road network centerline diagram structure, take the endpoints A and B of the curve and obtain the straight line AB. The straight line AB is called the chord of the curve. Step 202: Obtain the point C on the curve that is furthest from the straight line segment, and calculate its distance d from AB; Step 203: Compare the distance with a pre-defined threshold. If it is less than the threshold, the straight line segment is used as an approximation of the curve. Once the curve segment is processed, proceed to the next curve segment. Step 204: If the distance is greater than the threshold, divide the curve into two segments AC and BC using point C, and process each segment separately according to steps 201-203. Step 205: When all curves have been processed, connect the various dividing points in sequence to form a broken line, which can be used as an approximation of the curve for the next segment of curve processing.

[0026] The simplified result of the road network centerline is as follows Figure 8 As shown.

[0027] Because the road segmentation results are affected by the quality of the training data and the training environment, the resulting road network contains some jagged edges. To address this issue, this embodiment removes jagged edges from the simplified road network, specifically deleting curves with a length less than a certain threshold. The diagrams before and after jagged edge removal are shown below. Figure 9 As shown, where Figure 9 (a) is a schematic diagram before deburring; Figure 9 (b) is a schematic diagram after deburring.

[0028] To identify potential disconnections, this public example uses the A* algorithm combined with the probability values ​​of the road segmentation binary image to automatically connect the points. The specific steps are as follows: Step 301: Traverse all nodes in graph G(V,E), and determine the degree of the current node, such as... Figure 10 The diagram shows the definition of node degree; if the degree of the current node is 1, it is denoted as v, and the following steps are performed simultaneously; Step 302: Create a window of size (T1,T2) centered on node v with current degree 1, where T is a preset threshold, and find all edges of graph G(V,E) within this window range, denoted as e; Step 303: For the current node v, traverse all edges e that meet the requirements, calculate the projection point p of node v on edge e, and find the optimal edge best_e through the corresponding constraints. The constraints include: 1) The nodes of the edge are not near node v (to avoid self-connection or forming small loop topology); 2) The angle formed by the edge containing node v and the projection point p to node v is less than a certain threshold. A schematic diagram of the angle is shown below. Figure 11 As shown in the figure, the included angle is a vector. sum vector The included angle. 3) The distance from the projection point p to the node v is the shortest; Based on the above three constraints, the optimal edge best_e for the current node v with degree 1 is obtained. If best_e is not empty, the following steps are performed; otherwise, proceed to step 301 to continue traversing the next node. Step 304: Combining the probability values ​​of the road segmentation binary map, use the A* algorithm to find the shortest connection path and automatically connect the broken road network. The specific steps are as follows: Step 3041: Starting from the starting point p, treat p as a square waiting to be checked and put it into the "open list". The "open list" is a list that stores squares waiting to be checked. Step 3042: Find the reachable squares around the starting point p (maximum 8, i.e., the 8-neighborhood of the starting point), put them into the "open list", and set their parent square to p; Step 3043: Delete the starting point p from the "Open List" and put p into the "Close List". The "Close List" contains the squares that do not need to be checked. Step 3044: Calculate the F value for each square, where the formula for calculating the F value is shown below: F = M + H + (1 - P) * k Where M represents the movement cost from the starting point to the specified square, assuming a horizontal movement of one square costs 10 and a diagonal movement of one square costs 14 (the specific value can be modified according to the specific application). H represents the estimated cost of moving from the specified square to the target point v. P represents the road segmentation probability value of the square to be moved to. k is a constant value; Step 3045: Select the cell 'a' with the lowest F value from the "Open List", delete it from the "Open List", and add it to the "Close List"; Step 3046: Check all adjacent and reachable squares of cell a. Squares in the "Closed List" are ignored. If these squares are not yet in the "Open List," add them to the "Open List," calculate their F value, and set the parent square to a. If an adjacent square c is already in the "Open List," calculate the new path from p to square c (i.e., the path through a), and determine if an update is needed, i.e., whether the value of M is lower. If the new M value is lower, change the parent square to square a, recalculate the F value, and the H value does not need to be changed because the expected cost of the target point is fixed. If the new M value is higher, the value remains unchanged (M value remains unchanged and is not updated). Step 3047: Continue to find the one with the smallest F value from the "Open List", delete it from the "Open List", add it to the "Closed List", and continue to find the surrounding reachable blocks, and repeat this process. Step 3048, End Judgment: When the target block v appears in the "Open List", it means that the optimal path has been found. When there is no data in the "Open List", it means that there is no suitable path. Step 305: Based on the shortest path algorithm of A*, establish new nodes and edges in graph G(V,E), that is, connect the starting point p to the target point v. At the same time, go back to step 301 to continue traversing the next node until all nodes in graph G(V,E) have been traversed.

[0029] Schematic diagram before and after connection breakage is shown below Figure 12 As shown, where Figure 12 (a) is a schematic diagram before connection; Figure 12 (b) is a schematic diagram after connection; Figure 13 These are detailed comparison images before and after the fracture. Figure 13 (a) is a schematic diagram before connection; Figure 13 (b) is a schematic diagram after connection.

[0030] The automatic completion of the road network connecting nodes follows these steps: Step 501: Traverse all nodes in graph G(V,E), determine the degree of the current node, and record the degree of the current node as v when the degree of the current node is 1. At the same time, perform the following steps. Step 502: Create a window of size (T3,T4) centered on node v with current degree 1, and find all edges of graph G(V,E) within the window range, denoted as e; Step 503: For the current node v, traverse all nodes that satisfy the requirements for edge e, and simultaneously find the optimal node best_vertex through the corresponding constraints. The constraints include: 1) The nodes of the edge are not near node v (to avoid self-connection or forming small loop topology); 2) The angle between the edge containing node v and the node from node v to the current edge is less than a certain threshold. A diagram illustrating this angle is shown below. Figure 14 As shown in the figure; it can be seen from the figure that the included angle is a vector. sum vector The included angle; 3) Calculate the intersection of the edge formed by the current edge node and node v with its nearby edges to prevent the current edge node from being the same as node v; 4) The distance between the current edge node and node v is the shortest; Based on the above four constraints, the optimal node best_vertex for the current node v with degree 1 is obtained. If node best_vertex is not empty, the following steps are performed; otherwise, proceed to step 501 to continue traversing the next node. Step 504: Connect the current node v with best_vertex.

[0031] The diagram before and after automatic completion of the road network connecting the nodes is shown below. Figure 15 As shown, where Figure 15 (a) is a schematic diagram before connection; Figure 15 (b) is a schematic diagram after connection; Figure 16 This is a comparison image showing the details before and after the connection. Figure 16 (a) is a schematic diagram before connection; Figure 16 (b) is a schematic diagram after connection.

[0032] Road network intersection misalignment correction: This involves standardizing the road network nodes at intersections, as illustrated in the diagram below. Figure 17 As shown. Figure 17 It is known that road segments AF and BE intersect, forming a crossroads. However, the road network obtained based on the previous steps resulted in two T-junctions, meaning that road segment BE was incorrectly segmented into BC, CD, and DE. Due to the limitations of deep learning segmentation accuracy and the complexity of real-world scenarios, this type of problem is quite common in automatic road network extraction. To address this issue, potential errors, such as... Figure 17 As shown on the left, nodes C and D represent the original misaligned nodes; as Figure 17 As shown on the right, node G is the intersection of the obtained road network crossroads. The specific implementation process is as follows: First, traverse each node with a degree of 3, taking the current node v with a degree of 3 as the starting node, and obtain the connected length d from it to the nearest node with a degree of 3. If d is less than the set threshold T, proceed to the next step; otherwise, traverse the next node with a degree of 3 until the end. Calculate the center point of the edge formed by two nodes with a degree of 3. Use the obtained center point as the new node, replacing the two nodes with a degree of 3.

[0033] The diagrams before and after the alignment of the road network intersections are shown below. Figure 18 As shown, where Figure 18 (a) is a schematic diagram before reorganization; Figure 18 (b) is a schematic diagram after adjustment.

[0034] Transforming a ring topology into a star topology: This involves converting a ring topology where all nodes have a degree of 3 into a star topology. A schematic diagram is shown below. Figure 19 As shown. By Figure 19 It can be seen that the original road network topology has a ring topology consisting of 3 nodes (CBD). Since the degree of each of the 3 nodes in the ring topology is 3, we obtain the center F of this ring topology and connect the nodes of the ring topology. We then delete the original topological connections, transforming it from a ring topology into a star topology, as shown below. Figure 19 As shown on the right, the dashed line represents the original connection method, and the solid line in the middle represents the topology connection method after the ring-to-star topology is transformed; the specific implementation steps are as follows: Step 701: First, in order to improve the efficiency of finding loops, traverse each node with a degree of 3, create a window of size (T5,T6) centered on the current node v with a degree of 3, and find all edges in graph G(V,E) that are connected to node v within the window range, and treat this local road network graph as a directed graph. Step 702: Let the directed graph have n nodes and m edges, and its adjacency matrix be an n-order square matrix. We use a linked list or array to sequentially record the visited nodes. Starting from node v1 with a current degree of 2, we use a depth-first search (DFS) algorithm to find a node adjacent to v1 and determine if that node has been visited. If it has been visited, it indicates a cycle has been formed. We shrink all nodes in the cycle to a single point, generate a new adjacency matrix, and continue the DFS depth search until no cycles remain. Step 703: Calculate the total edge length of the found ring topology. If the total length is greater than a preset threshold, continue with step 701. If the total length is less than the preset threshold, obtain the center F of the ring topology through geometric relationships, connect all nodes of the ring topology, delete the original topological connection method, and transform it from a ring topology to a star topology. Then, continue with step 701 until all nodes with a degree of 3 in the road network have been traversed.

[0035] The diagrams before and after the road network's ring topology was transformed into a star topology are shown below. Figure 20 As shown, where Figure 20 (a) is a schematic diagram before the conversion; Figure 20 (b) is a schematic diagram after the conversion.

[0036] For nodes with a regularity of 3, the intersection points of their bifurcations will have some deviation due to the conversion of road segmentation results into a road network map. Figure 21 As shown. To solve this problem, it was normalized using geometric relations, and the normalization result is shown below. Figure 21 As shown on the right side of the diagram; specifically, it includes two methods; Method 1: Regularize nodes with degree 3 based on the length of the edge containing the node with degree 3. The specific implementation steps are as follows: Step 8011: Traverse each node with degree 3 and denote the current node with degree 3 as v. Step 8012: Calculate the length of the edge containing the node v with degree 3, and arrange them in ascending order as e1 (the node of e1 is (v1,v)), e2 (the node of e2 is (v2,v)), and e3 (the node of e3 is (v3,v)). Calculate the angle formed by the two shortest edges e1 and e2, and determine whether the angle is greater than the set threshold. If it is greater than the set threshold, continue to the next step; if it is less than the set threshold, continue to traverse the next node with degree 3. Step 8013: Calculate the angle between the vector formed by nodes v3 and v and the vector formed by v1 and v2. If this angle is less than the threshold T7 or greater than the threshold T8, calculate the intersection point p of edge e3 and the line segment formed by v1 and v2. At the same time, determine whether the intersection point p exists. If it does not exist, calculate the projection point of node v on the line segment formed by v1 and v2, and make p equal to this projection point. If this angle is greater than the threshold T7 and less than the threshold T8, calculate the projection point of node v3 on the line segment formed by v1 and v2, and make p equal to this projection point. Step 8014: Make v equal to p, that is, replace the node v with degree 3 with p; then, continue with step 8011 until all nodes with degree 3 in the road network have been traversed.

[0037] Method 2: Based on the angle of the edge containing the node with degree 3, normalize the nodes with degree 3. The specific implementation steps are as follows: Step 8021: Traverse each node with a degree of 3, and denote the current node with a degree of 3 as v; Step 8022: Calculate the angle between any two edges containing a node v with degree 3, and denote the two edges with the largest angles as e1 (nodes of e1 are (v1, v)) and e2 (nodes of e2 are (v2, v)), and the other edge as e3 (nodes of e3 are (v3, v)). Simultaneously, determine if the angle between e1 and e2 is greater than a set threshold. If it is, continue to the next step; if it is less than the set threshold, continue traversing the next node with degree 3. Step 8023: Calculate the angle between the vector formed by nodes v3 and v and the vector formed by v1 and v2. If this angle is less than the threshold T7 or greater than the threshold T8, calculate the intersection point p of edge e3 and the line segment formed by v1 and v2. At the same time, determine whether the intersection point p exists. If it does not exist, calculate the projection point of node v on the line segment formed by v1 and v2, and make p equal to this projection point. If this angle is greater than the threshold T7 and less than the threshold T8, calculate the projection point of node v3 on the line segment formed by v1 and v2, and make p equal to this projection point. Step 8024: Make v equal to p, that is, replace the node v with degree 3 with p; then, continue with step 8021 until all nodes with degree 3 in the road network have been traversed. A diagram illustrating the nodes with degree 3 before and after is shown below. Figure 22 As shown, 22(a) is a schematic diagram before regularization; Figure 22 (b) is a schematic diagram after normalization.

[0038] Remove existing noisy loop topologies, such as Figure 23 The ring topology of the road network shown on the left has 3 vertices with a degree of 3; as shown... Figure 24 The ring topology of the road network shown on the left has two vertices with a degree of 3; as shown... Figure 25The loop topology of the road network shown on the left has one vertex with a degree of 3. When the total length of the loop's edges is less than a certain threshold, a loop-removal topology is performed, as shown below. Figure 23 , Figure 24 , Figure 25 As shown on the right. The specific implementation steps are as follows: Step 901: In order to improve the efficiency of finding loops, traverse each node with a degree of 2 (since nodes with a degree of 1 cannot form a loop topology), create a window of size (T5,T6) centered on the current node v with a degree of 2, and find all edges in graph G(V,E) that are connected to node v within the window range, and treat this local road network graph as a directed graph. Step 902: Let the directed graph have n nodes and m edges, and its adjacency matrix be an n-order square matrix. We also use a linked list or array to sequentially record the visited nodes. Starting from node v with a current degree of 2, in this embodiment, a depth-first search (DFS) algorithm is used to find a node adjacent to node v, and it is determined whether this node has been visited. If it has been visited, it indicates that a cycle has been formed. All nodes in this cycle are shrunk to a single point, a new adjacency matrix is ​​generated, and the DFS depth search continues until there are no more cycles. Step 903: Calculate the total edge length of the found loop topology. If the total length is greater than a certain threshold, continue with step 901; if the total length is less than a certain threshold, perform loop removal processing. Then, continue with step 901 until all nodes with a degree of 2 in the road network have been traversed. The diagram before and after removing noisy loop topologies is shown below. Figure 26 As shown, where, Figure 26 (a) is a schematic diagram before removal; Figure 26 (b) is a schematic diagram after removal. Example 2

[0039] This disclosure presents a road network centerline generation system based on adaptive road topology search, such as... Figure 27 As shown, the system includes an image acquisition module, a skeleton extraction module, a road network simplification module, and a road network optimization module. The image acquisition module is used to acquire segmented binary images of road remote sensing images. The skeleton extraction module is used to perform morphological processing on the segmented binary images to extract road skeleton lines. The road network simplification module is used to represent the road network centerline using a graph structure in graph theory, i.e., graph G(V,E), where V represents the set of road nodes and E represents the set of edges connecting the nodes required to construct the road network. The road network centerline is simplified without changing the original road network topology.

[0040] The road network optimization module is used to optimize the simplified road network and generate a road network centerline based on adaptive search of road topology. The optimization process includes: removing burrs from the road network; finding disconnected points in the road network and automatically connecting them using the probability values ​​of the aforementioned segmented binary image; automatically completing the road network connections between nodes; straightening out misaligned intersections in the road network; converting the loop topology in the road network into a star topology; straightening out nodes with a degree of 3 in the road network using geometric relationships; and finally removing existing noisy loop topologies. Example 3

[0041] This disclosure presents a road network centerline generation device based on adaptive road topology search, comprising a processor and a memory electrically connected to each other; the memory is used to store a computer program; when the processor executes the aforementioned computer program, it can implement the road network centerline generation method based on adaptive road topology search described in Embodiment 1 above, and the specific image generation steps are the same as those in Embodiment 1 above, and will not be repeated here. Example 4

[0042] This disclosure provides a computer-readable storage medium storing a computer program. When the computer program is executed, it can implement the road network centerline generation method based on road topology adaptive search as described in Embodiment 1 above. The specific image generation method steps are the same as those in Embodiment 1 above, and will not be repeated here.

[0043] The computer described in this application embodiment can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. The computer-readable storage medium can be any usable medium that a computer can read, or a data storage device such as a server or data center that integrates one or more usable media. The usable medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., digital versatile optical disc (DVD)), or a semiconductor medium (e.g., solid-state drive (SSD)). The software formed by the computer's stored code can reside in random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, or other storage media that are mature in the art.

[0044] In the various embodiments of this application, the functional modules can be integrated into one processing unit or module, or each module can exist physically separately, or two or more modules can be integrated into one unit or module. In the above embodiments, they can be implemented entirely or partially through software, hardware, firmware, or any combination thereof. When implemented using software, they can be implemented entirely or partially in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated.

[0045] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for generating road network centerlines based on adaptive road topology search, characterized in that... include: Obtain segmented binary images of road remote sensing images; Morphological processing is performed on the segmented binary image to extract the road skeleton lines; The road framework is constructed using graph structures in graph theory. Indicates the centerline of the road network, where, Represents the set of road nodes. This represents the set of edges connecting the nodes required to construct the road network; Then, without changing the original road network topology, the road network centerline is simplified, and the simplified road network is optimized to generate a road network centerline based on adaptive search of road topology. The optimization process includes: first, removing burrs from the road network; Secondly, locate the disconnected points in the road network and automatically connect them using the A* algorithm combined with the probability values ​​of the segmented binary image; Secondly, the automatic completion of road network connections between nodes in the road network includes the following steps: Step 501: Traverse the graph Given all nodes, determine the degree of the current node. If the degree of the current node is 1, denote it as... ; Step 502: Starting with the node with a current degree of 1 Create size for center The window, and look for the image. All edges within this window range are denoted as ; Step 503, for the current node Iterate through all edges that satisfy the requirements. Find the optimal node by considering all nodes and the node constraints. ; The node constraints include: nodes with edges and nodes not present. Nearby; node Edges and nodes If the angle between the current edge node and the node formed by the current edge node is less than a certain threshold, calculate the angle between the current edge node and the node formed by the current edge node. The intersection of the formed edge and its neighboring edges prevents nodes of the current edge from intersecting with nodes. Same; the node of the current edge and the node The shortest distance is ; Get the node with a current degree of 1 optimal node When node If the value is not empty, proceed with the following steps; otherwise, go to step 501 and continue traversing the next node. Step 504: Connect the current node and ; Next, the intersections in the road network are staggered and normalized. This involves traversing each node with a degree of 3, starting with the node currently at degree 3. Starting with a node, obtain the connected length from it to the nearest node with a degree of 3. ;if Less than the set threshold If the condition is met, proceed to the next step; otherwise, traverse the next node with a degree of 3 until the end; calculate the center point of the edge formed by two nodes with a degree of 3; use the obtained center point as the new node, replacing the two nodes with a degree of 3; transform the ring topology with nodes of degree 3 into a star topology; regularize the nodes with a degree of 3 in the road network using geometric relationships in two ways: Method 1 is to regularize the nodes with a degree of 3 according to the length of the edge containing the node with a degree of 3; Method 2 is to regularize the nodes with a degree of 3 according to the included angle of the edges containing the nodes with a degree of 3. Finally, the existing noisy loop topology is removed, completing all optimization processes.

2. The method for generating road network centerlines based on adaptive road topology search according to claim 1, characterized in that, The simplification of the road network centerline includes the following steps: Step 201: For a curve formed by a set of connected nodes V and the edges E connecting the nodes in the road network centerline diagram structure, take the endpoints A and B of the curve and obtain the straight line AB. The straight line AB is called the chord of the curve. Step 202: Obtain the point C on the curve that is furthest from the straight line segment, and calculate its distance d from AB; Step 203: Compare the distance with a pre-defined threshold. If it is less than the threshold, the straight line segment is used as an approximation of the curve. Once the curve segment is processed, proceed to the next curve segment. Step 204: If the distance is greater than the threshold, divide the curve into two segments AC and BC using point C, and process each segment separately according to steps 201-203. Step 205: Once all curves have been processed, connect the dividing points sequentially to form a broken line, which can be used as an approximation of the curve for processing the next segment of the curve.

3. The method for generating road network centerlines based on adaptive road topology search according to claim 1, characterized in that, The process of finding disconnected points in the road network and automatically connecting them using the A* algorithm combined with the probability values ​​of the segmented binary image includes the following steps: Step 301: Traverse the graph Given all nodes, determine the degree of the current node. If the degree of the current node is 1, denote it as... At the same time, perform the following steps; Step 302: Starting with the node with a current degree of 1 Create a center with a size of The window, where T is a preset threshold, is used to find the graph. All edges within this window range are denoted as ; Step 303, for the current node Iterate through all edges that satisfy the requirements. and calculate the nodes On the side projection point At the same time, the optimal edge is found through constraints. ; when If the value is not empty, proceed to the following steps; otherwise, go to step 301 to continue traversing the next node. Step 304: Combining the probability values ​​of the road segmentation binary map, find the shortest connection path and automatically connect the broken road network, specifically including: Step 3041, Starting from the beginning Begin, put As a square waiting to be checked, it is placed into the "Open List", which is a list that stores squares waiting to be checked; Step 3042: Find the starting point Add all reachable squares to the "Open List" and set their parent square to [parent square]. ; Step 3043: Delete the starting point from the "Start List" and will Add them to the "Close List", which contains the squares that do not need to be checked; Step 3044: Calculate the F value for each square, where, The formula for calculating the value is shown below: ;in, Indicates the movement cost from the starting point to the specified square; This indicates moving from the specified square to the target point. Expected consumption; This represents the probability value of the road segmentation of the cell to which the player moves. It is a constant value; Step 3045: Select from the "Open List" The square with the lowest value Remove it from the "On List" and add it to the "Off List"; Step 3046, Inspection All adjacent and reachable squares; squares in the "Closed List" are not considered; if these squares are not yet in the "Open List," add them to the "Open List" and calculate their value. Value, and set the parent cell to If a certain adjacent square The new path is already being calculated from the "Open List". Reaching the square That is, after The path is used to determine whether an update is needed. Is the value lower? If the new If the value is lower, then modify the parent square to a square. Recalculate value, The value does not need to be changed because the expected cost at the target point is fixed; if the new value is changed... If the value is higher, the value remains unchanged; Step 3047: Continue to find from the "Open List" The one with the smallest value is removed from the "open list" and added to the "closed list". Then, the system continues to find reachable blocks in the surrounding area and repeats this process. Step 3048: When the target block appears in the "Open List" When the number of records is 1, it means that the optimal path has been found; when there is no data in the "Open List", it means that there is no suitable path. Step 305: Based on the shortest connection path, construct a new node and edge graph. In the middle, that is, connecting from the starting point To the target point Meanwhile, proceed to step 301 to continue traversing the next node until the graph is complete. The traversal of the middle nodes is complete.

4. The method for generating road network centerlines based on adaptive road topology search according to claim 3, characterized in that, The constraints include: the node of the edge is not a node. Nearby; node The edge and the projection point To the node The angle formed is less than a certain threshold; projection point To the node The shortest distance is [the shortest distance].

5. The method for generating road network centerlines based on adaptive road topology search according to claim 1, characterized in that, The process of transforming a ring topology with all nodes having a degree of 3 into a star topology includes the following steps: Step 701: Traverse each node with a degree of 3, starting with the node with the current degree of 3. Create a center with a size of The window, and look for the image. Within this window range and nodes Connect all edges and treat this local road network as a directed graph; Step 702: Let the number of nodes in the above directed graph be... The number of sides is Its adjacency matrix is A square matrix of degree 2; using a linked list or array to record visited nodes sequentially; starting from the node with a current degree of 2. Start by searching for a node For adjacent nodes, determine whether the node has been visited; if it has been visited, it indicates that a cycle has been formed; shrink all nodes in the cycle to a single point, generate a new adjacency matrix, and continue searching and determining until there are no cycles. Step 703: Calculate the total length of the edges of the searched ring topology. If the total length is greater than the preset threshold, continue to step 701. When the total length is less than a preset threshold, the center F of the ring topology is obtained through geometric relationships. At the same time, each node of the ring topology is connected, the original topological connection method is deleted, and the ring topology is transformed into a star topology. Then, step 701 is continued until all nodes with a degree of 3 in the road network are traversed.

6. The method for generating road network centerlines based on adaptive road topology search according to claim 1, characterized in that, Method 1, which normalizes nodes of degree 3 based on the length of the edge containing the node of degree 3, includes the following steps: Step 8011: Traverse each node with a degree of 3, and denote the current node with a degree of 3 as... ; Step 8012: Calculate the node with degree 3. The lengths of the sides, arranged in ascending order, are: , The node is , , The node is , , The node is And calculate the two shortest edges. , The angle formed by the angle is determined, and it is also determined whether the angle is greater than the set threshold. If it is greater than the set threshold, the next operation is continued; if it is less than the set threshold, the next node with a degree of 3 is traversed. Step 8013, Calculate the node and The vector formed by and and The angle between the vectors formed, if this angle is less than a threshold Or greater than the threshold Calculate the edges and and The intersection of the line segments formed At the same time, determine the intersection point. Does it exist? If it does not exist, then calculate. Node at and The projection points of the line segment formed, and such that Equal to this projection point; if this included angle is greater than the threshold And less than the threshold ,calculate Node at and The projection points of the line segment formed, and such that Equal to this projection point; Step 8014, making equal That is, using Nodes with a substitution degree of 3 Then, continue with step 8011 until all nodes with a degree of 3 in the road network have been traversed. Method 2, which normalizes nodes with an angle of 3 on the edges containing nodes with an angle of 3, includes the following steps: Step 8021: Traverse each node with a degree of 3, and denote the current node with a degree of 3 as... ; Step 8022: Calculate the node with degree 3. The included angle between any two sides of a given side is denoted as , and the two sides with the largest included angle are denoted as . , The node is and , The node is The other edge is denoted as , The node is Simultaneously judge and If the included angle is greater than the set threshold, continue to the next step; if it is less than the set threshold, continue traversing the next node with a degree of 3. Step 8023, Calculate the node and The vector formed by and and The angle between the vectors formed, if this angle is less than a threshold Or greater than the threshold Calculate the edges and and The intersection of the line segments formed At the same time, determine the intersection point. Does it exist? If not, then calculate. Node at and The projection points of the line segment formed, and such that Equal to this projection point; if this included angle is greater than the threshold And less than the threshold ,calculate Node at and The projection points of the line segment formed, and such that Equal to this projection point; Step 8024, making equal That is, using Nodes with a substitution degree of 3 Then, continue with step 8021 until all nodes with a degree of 3 in the road network have been traversed.

7. The method for generating road network centerlines based on adaptive road topology search according to claim 1, characterized in that, The process of removing existing noise loop topology includes the following steps: Step 901: Traverse each node with a degree of 2, starting with the node with the current degree of 2. Create size for center The window, and look for the image. Within this window range and nodes Connect all edges and treat this local road network as a directed graph; Step 902: Let the number of nodes in the directed graph described in the previous step be... The number of sides is Its adjacency matrix is A square matrix of degree 2; using a linked list or array to record visited nodes sequentially; starting from the node with a current degree of 2. Starting from there, the search found a node. For adjacent nodes, determine whether the node has been visited; if it has been visited, it indicates that a cycle has been formed; shrink all nodes in the cycle to a single point, generate a new adjacency matrix, and continue the search until there are no more cycles. Step 903: Calculate the total edge length of the searched ring topology. If the total length is greater than the preset threshold, continue with step 901. If the total length is less than the preset threshold, perform ring-removal topology processing. Then, continue with step 901 until all nodes with a degree of 2 in the road network have been traversed.

8. A road network centerline generation system based on adaptive road topology search, characterized in that: It includes an image acquisition module, a skeleton extraction module, a road network simplification module, and a road network optimization module; The image acquisition module is used to acquire segmented binary images of road remote sensing images; The skeleton extraction module is used to perform morphological processing on the segmented binary image to extract the road skeleton lines; The road network simplification module is used to transform the road skeleton using graph structures in graph theory. Indicates the centerline of the road network, where, Represents the set of road nodes. This represents the set of edges connecting the nodes required to construct the road network; it simplifies the centerline of the road network without changing the original road network topology. The road network optimization module is used to optimize the simplified road network and generate a road network centerline based on adaptive search of road topology; the optimization process includes: first, removing burrs from the road network; Secondly, locate the disconnected points in the road network and automatically connect them using the A* algorithm combined with the probability values ​​of the segmented binary image; Secondly, the automatic completion of road network connections between nodes in the road network includes the following steps: Step 501: Traverse the graph Given all nodes, determine the degree of the current node. If the degree of the current node is 1, denote it as... ; Step 502: Starting with the node with a current degree of 1 Create size for center The window, and look for the image. All edges within this window range are denoted as ; Step 503, for the current node Iterate through all edges that satisfy the requirements. Find the optimal node by considering all nodes and the node constraints. ; The node constraints include: nodes with edges and nodes not present. Nearby; node Edges and nodes If the angle between the current edge node and the node formed by the current edge node is less than a certain threshold, calculate the angle between the current edge node and the node formed by the current edge node. The intersection of the formed edge and its neighboring edges prevents nodes of the current edge from intersecting with nodes. Same; the node of the current edge and the node The shortest distance is ; Get the node with a current degree of 1 optimal node When node If the value is not empty, proceed with the following steps; otherwise, go to step 501 and continue traversing the next node. Step 504: Connect the current node and ; Next, the intersections in the road network are staggered and normalized. This involves traversing each node with a degree of 3, starting with the node currently at degree 3. Starting with a node, obtain the connected length from it to the nearest node with a degree of 3. ;if Less than the set threshold If the condition is met, proceed to the next step; otherwise, traverse the next node with a degree of 3 until the end; calculate the center point of the edge formed by two nodes with a degree of 3; use the obtained center point as the new node, replacing the two nodes with a degree of 3; transform the ring topology with nodes of degree 3 into a star topology; normalize the nodes with a degree of 3 in the road network through geometric relationships in two ways: Method 1 is to normalize the nodes with a degree of 3 according to the length of the edge containing the node with a degree of 3; Method 2 is to normalize the nodes with a degree of 3 according to the included angle of the edges containing the nodes with a degree of 3; finally, remove the existing noisy loop topology to complete all optimization processing.

9. A road network centerline generation device based on adaptive road topology search, comprising a processor and a memory electrically connected in phase; the memory is used to store a computer program; characterized in that: When the processor executes the aforementioned computer program, it can implement the road network centerline generation method based on road topology adaptive search as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that: The storage medium stores a computer program; when the computer program is executed, it can implement the road network centerline generation method based on road topology adaptive search as described in any one of claims 1-7.

Images