A method, system, device, and medium for directional connectivity road detection

Abstract

The application discloses a kind of direction connectivity road detection method, system, equipment and medium, specifically related to road detection field, its gist is: the association between multiple road nodes is obtained, and with road mask vector as node, the association between road nodes is edge, and topological network is constructed;Obtain multiple target image blocks of target area remote sensing image, and extract the target road mask of road in each target image block and the road node of road center line;The slope of the road where breakpoint position is located is obtained at each breakpoint, based on the slope direction in each target image block to build the first search area corresponding to each breakpoint, to build the second search area corresponding to each breakpoint between each target image block;First associated point with breakpoint having connectivity in first search area is obtained using topological network, and second associated point with breakpoint having connectivity in second search area, and road detection result is generated using first associated point and second associated point.

Description

Technical Field

[0001] This invention relates to the field of road inspection, and specifically to a method, system, equipment, and medium for detecting directional connectivity roads. Background Technology

[0002] Road extraction, as a key technology in remote sensing image interpretation, is becoming increasingly important with the exponential growth in demand for geospatial information. In smart city construction, road network topology analysis provides crucial data support for traffic flow simulation and infrastructure layout optimization; in disaster emergency response scenarios, rapid extraction of damaged road information can shorten the golden rescue window; in the field of surveying and mapping in remote areas, automated road identification technology has increased the efficiency of traditional manual surveying many times over, significantly reducing operating costs. In recent years, breakthroughs in high-resolution Earth observation technology have enabled commercial satellite imagery to break through the sub-meter resolution level, and UAV oblique photography can capture centimeter-level ground details, laying a solid data foundation for refined road identification. Meanwhile, the iterative evolution of deep learning technology has given rise to new architectures such as U-Net++ and Swin-Transformer, which significantly enhance the model's ability to represent road continuity features and complex topological relationships through multi-level feature fusion and self-attention mechanisms. However, this technology still faces multiple challenges in practical applications: the three-dimensional occlusion effect caused by dense urban buildings and elevated roads leads to a high road fragmentation rate in vertical projection images; spectral confusion caused by vegetation cover in mountainous and forested scenes results in a high misjudgment rate of dirt road boundaries; and the degradation of optical image quality caused by cloudy and rainy weather directly leads to a decrease in the stability of feature extraction. In addition, differences in road construction standards across regions, seasonal changes in land cover, and limitations of low-light imaging at night continue to test the robustness and generalization ability of the algorithm.

[0003] Traditional road extraction methods primarily rely on manual feature engineering and rule-based modeling techniques. Template-matching-based geometric search algorithms, while capable of rapidly locating regular roads by matching pre-defined road geometric templates within a limited area, are insufficient for curved roads. Morphological processing schemes (such as multi-threshold segmentation and straight-segment detection) combine multi-band spectral analysis to generate road networks through binarization and skeleton extraction; however, their morphological filtering process is susceptible to interference from vegetation shadows. Hough transform isoline detection techniques were once widely used in low-resolution imagery, optimizing broken line segment connections using genetic algorithms; however, when processing sub-meter high-resolution imagery, the modeling accuracy of road topology decreases. While Markov random field (MRF) models can improve road network continuity through probabilistic graphical optimization, the time-consuming manual feature design accounts for a significant portion of the overall process, and the computational complexity increases exponentially.

[0004] Compared to traditional methods, deep learning technology significantly improves the efficiency of road extraction through a data-driven approach. Chinese patent CN117173557A proposes the GDU-Net architecture, which captures long-range dependencies of rural roads through a global contextual attention module. Chinese patent CN117351360A innovatively designs a cyclic cross-attention module, strengthening road feature expression through spatiotemporal feature interaction, and, combined with a channel-space dual-dimensional attention mechanism, improves the fragmentation characteristics of complex road networks. Chinese patent CN117636154A proposes an edge-semantic collaborative network, employing a dual-branch architecture to simultaneously perform semantic segmentation and edge detection, achieving refined segmentation guided by edge information through a feature fusion module. However, while current deep learning-based semantic segmentation road extraction schemes can efficiently process large-scale data, they neglect road connectivity. Summary of the Invention

[0005] The technical problem this invention aims to solve is that existing technologies neglect road connectivity in road detection, leading to reduced accuracy of detection results. The invention provides a method, system, device, and medium for directional connectivity road detection. By acquiring the relationships between multiple road nodes, a topological network representing road connectivity is constructed to enhance the representation of road connectivity when roads are occluded by shadows or other obstructions. Simultaneously, through hierarchical settings of local optimization in the first search region and global optimization in the second search region, and utilizing correlation point search within a single image block and correlation point search between multiple image blocks, the efficiency of correlation point search is improved, thus solving the problem of existing technologies neglecting road connectivity in road detection, resulting in reduced accuracy of detection results.

[0006] This invention is achieved through the following technical solution:

[0007] A method for detecting directional connectivity roads, the method comprising:

[0008] The road extraction model trained is used to infer sample image blocks and obtain road mask vectors; the road centerline is obtained using the road mask vectors, and multiple road nodes are obtained based on the road centerline to construct discontinuous roads; wherein, the road nodes include discontinuities and intersections;

[0009] By comparing disconnected roads with centerline masks, the relationships between multiple road nodes are obtained, and a topological network is constructed using road mask vectors as nodes and the relationships between road nodes as edges.

[0010] Acquire multiple target image patches from the remote sensing image of the target area, and extract the target road mask and road nodes of the road centerline in each target image patch;

[0011] At each breakpoint, the slope of the road where the breakpoint is located is obtained. Based on the slope direction, a first search region corresponding to each breakpoint is constructed within each target image block, and a second search region corresponding to each breakpoint is constructed between each target image block.

[0012] The first associated point connected to the breakpoint within the first search area and the second associated point connected to the breakpoint within the second search area are obtained using the topology network. The road detection results are generated using the first and second associated points.

[0013] Furthermore, before using the trained road extraction model to infer sample image patches and obtain road mask vectors, the method also includes:

[0014] Obtain remote sensing images of the sample area and road masks and centerline masks of roads in the sample area remote sensing images. Cut the sample remote sensing images of the area to obtain multiple sample image blocks and the corresponding road masks and centerline masks of the sample image blocks, and construct a training set.

[0015] The sample image patches in the training set are input into the semantic segmentation model to extract the road extraction mask. The road extraction mask is then compared with the road mask in the training set. Based on the comparison results, the model is iteratively trained to obtain the optimal weights.

[0016] Furthermore, the breakpoint refers to a point on the road centerline with a degree of 1; the intersection point refers to a point on the road centerline with a degree greater than 2.

[0017] Furthermore, based on the slope direction, a first search region corresponding to each breakpoint is constructed within each target image block, and a second search region corresponding to each breakpoint is constructed between each target image block, specifically as follows:

[0018] Extend a distance of 0.1 times the side length of the target image block along the slope direction of each breakpoint as the first radius, and take the target image block area within ±90° of the first radius as the center of each breakpoint as the first search area;

[0019] Extend a distance of twice the side length of the target image block along the slope direction of each breakpoint as the second radius, and obtain multiple target image block regions within ±90° of the second radius with each breakpoint as the center as the second search area.

[0020] Further, obtain the remote sensing image of the sample area and the road mask and centerline mask of the roads in the remote sensing image of the sample area, specifically as follows:

[0021] The remote sensing images of the sample area and the corresponding road SHP annotation files are obtained. The road centerline is extracted from the road SHP annotation files using vector data processing tools, and the road is converted into a binary mask to obtain the road mask and the centerline mask.

[0022] Furthermore, the topological network is used to obtain the first associated point in the first search region that is connected to the breakpoint, and the second associated point in the second search region that is connected to the breakpoint, specifically as follows:

[0023] Obtain all first road nodes within the first search area of ​​each breakpoint, and use the topology network to determine the connectivity between each obtained first road node and the corresponding breakpoint in order to filter out the first associated points with correlation.

[0024] Obtain all second road nodes within the second search area of ​​each breakpoint, and use the topology network to determine the connectivity between each obtained second road node and the corresponding breakpoint, so as to filter out the second associated points with connectivity.

[0025] The present invention also provides a directional connectivity road detection system, which is used in any of the above-described directional connectivity road detection methods, the system comprising:

[0026] The road node acquisition module is used to infer sample image blocks using a trained road extraction model to obtain road mask vectors; to obtain road centerlines using road mask vectors, and to obtain multiple road nodes based on road centerlines to construct discontinuous roads; wherein, the road nodes include discontinuities and intersections; it is also used to acquire multiple target image blocks of remote sensing images of the target area, and to extract the target road mask and road nodes of the road centerline in each target image block;

[0027] The topology network construction module is used to obtain the relationship between multiple road nodes by comparing disconnected roads and centerline masks, and to construct a topology network with road mask vectors as nodes and the relationship between road nodes as edges.

[0028] The search region construction module is used to obtain the slope of the road where the breakpoint is located at each breakpoint, construct the first search region corresponding to each breakpoint within each target image block based on the slope direction, and construct the second search region corresponding to each breakpoint between each target image block.

[0029] The target road detection module is used to obtain a first associated point in a first search area that is connected to the breakpoint, and a second associated point in a second search area that is connected to the breakpoint, using the first and second associated points to generate road detection results.

[0030] The present invention also provides a computer device, including a system memory and a processor, wherein the system memory stores a computer program, and the processor executes the computer program to implement the steps of any of the methods described above.

[0031] The present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of any of the methods described above.

[0032] The present invention also provides a computer program product containing instructions that, when executed by a cluster of computer devices, cause the cluster of computer devices to perform the method described in any of the preceding claims.

[0033] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0034] In this invention, by acquiring the association relationships between multiple road nodes, a topological network representing road connectivity is constructed to enhance the representation of road connectivity when roads are occluded by shadows or other factors. At the same time, through the hierarchical setting of local optimization in the first search region and global optimization in the second search region, the efficiency of association point search is increased by utilizing association point search within a single image block and association point search between multiple image blocks. This solves the problem that existing technologies neglect road connectivity in road detection, leading to reduced accuracy of detection results. Attached Figure Description

[0035] To more clearly illustrate the technical solutions of the exemplary embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly described below. It should be understood that the following drawings only show some embodiments of the present invention and should not be considered as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort. In the drawings:

[0036] Figure 1 This is a schematic diagram of the method flow for a directional connectivity road detection method in this embodiment;

[0037] Figure 2(a) is a schematic diagram of the remote sensing image of the sample image block in this embodiment;

[0038] Figure 2(b) is a schematic diagram of the road SHP annotation of the sample image block in this embodiment;

[0039] Figure 2(c) is the centerline mask image of the sample image block in this embodiment;

[0040] Figure 2(d) is the mask image of the sample image block after randomly erasing part of the center line in this embodiment;

[0041] Figure 3(a) is a schematic diagram of the remote sensing image of the target image block in this embodiment;

[0042] Figure 3(b) is a schematic diagram of the road SHP annotation of the target image block in this embodiment;

[0043] Figure 3(c) is the centerline mask image of the target image block in this embodiment;

[0044] Figure 4(a) is a schematic diagram of the mask before local road filling in this embodiment;

[0045] Figure 4(b) is a schematic diagram of the effect after partial road filling in this embodiment;

[0046] Figure 5 This is a schematic diagram of the module connections of a directional connectivity road detection system in this embodiment;

[0047] Figure 6 This is a schematic diagram of the structure of a computer device in this embodiment. Detailed Implementation

[0048] The exemplary embodiments of this disclosure are described below with reference to the accompanying drawings, including various details of the embodiments to aid understanding, and should be considered merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope of this disclosure. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description.

[0049] In this disclosure, unless otherwise stated, the use of terms such as "first," "second," etc., to describe various elements is not intended to limit the positional, temporal, or importance relationships of these elements; such terms are merely used to distinguish one element from another. In some examples, the first element and the second element may refer to the same instance of that element, while in other cases, based on the context, they may refer to different instances.

[0050] The terminology used in the description of the various examples in this disclosure is for the purpose of describing particular examples only and is not intended to be limiting. Unless the context explicitly indicates otherwise, an element may be one or more unless the number of elements is specifically limited. Furthermore, the term "and / or" as used in this disclosure covers any one of the listed items and all possible combinations thereof.

[0051] Example 1

[0052] See Figure 1 , Figure 1 A flowchart illustrating a method for detecting directional connectivity roads is shown, wherein the method includes:

[0053] S1: Use the trained road extraction model to infer sample image blocks and obtain road mask vectors; use the road mask vectors to obtain road centerlines, and obtain multiple road nodes based on the road centerlines to construct disconnected roads; wherein, the road nodes include breakpoints and intersections;

[0054] It should be noted that in this embodiment, before S1, the method further includes: acquiring high-resolution remote sensing images of the sample area (0.3-1m spatial resolution) and the corresponding road SHP annotation files; extracting sub-pixel level road centerlines from the road SHP annotation files using vector data processing tools; randomly truncating the road centerlines by 15%-25% using a random linear masking algorithm to simulate real-world breakage scenarios such as tree occlusion and shadow coverage; and performing binary masking on the road to obtain a road mask and a centerline mask; and processing the original remote sensing images of the sample area... The road mask and centerline mask were spatially aligned, and the regional sample remote sensing image was segmented using a sliding window to obtain multiple 512×512 pixel sample image blocks and their corresponding road masks and centerline masks. These were then divided into training and test sets at a ratio of 9:1. Figure 2(a) shows a schematic diagram of the remote sensing image of the sample image block; Figure 2(b) shows a schematic diagram of the road SHP annotation of the sample image block; Figure 2(c) shows the centerline mask image of the sample image block; Figure 2(d) shows the mask image after randomly erasing part of the centerline.

[0055] The sample image patches in the training set are input into the semantic segmentation model to extract the road extraction mask. The road extraction mask is then compared with the road mask in the training set. Based on the comparison results, the model is iteratively trained to obtain the optimal weights. The semantic segmentation model is composed of the Transformer-based SPIN road extraction network.

[0056] It should be noted that in this embodiment, the model performance is evaluated using a test set, and the evaluation method adopts conventional technical solutions in the field, which will not be elaborated on here.

[0057] Specifically, in this embodiment, the road mask vector is obtained by using the model's optimal weight inference sample image blocks; the road centerline is obtained by using the road mask vector; and all road nodes with a degree of 1 or greater than 2 in the road centerline are obtained according to the road centerline to construct all disconnected roads.

[0058] S2: By comparing disconnected roads and centerline masks, the association relationships between multiple road nodes are obtained, and a topological network is constructed using road mask vectors as nodes and the association relationships between road nodes as edges.

[0059] Specifically, in this invention, all disconnected roads are compared with the centerline mask to obtain the association relationship between multiple road nodes, and a topological network is constructed using the road mask vector as nodes and the association relationship between road nodes as edges; wherein, the association relationship refers to whether two road nodes are connected or not.

[0060] S3: Acquire multiple target image blocks of the remote sensing image of the target area, and extract the target road mask and road nodes of the road centerline in each target image block;

[0061] Specifically, in this embodiment, remote sensing images of the target area are acquired, and the remote sensing images of the target area are segmented to obtain multiple target image blocks. The target road mask and road nodes of the road centerline are extracted from each target image block. Figure 3(a) shows a schematic diagram of the remote sensing image of the target image block. Figure 3(b) shows a schematic diagram of the road SHP annotation of the target image block. Figure 3(c) shows the centerline mask image of the target image block.

[0062] S4: At each breakpoint, obtain the slope of the road where the breakpoint is located, construct the first search region corresponding to each breakpoint within each target image block based on the slope direction, and construct the second search region corresponding to each breakpoint between each target image block;

[0063] Specifically, in this embodiment, the slope of the road at each breakpoint is first obtained. Then, a distance of 0.1 times the side length of the target image block is extended along the slope direction of each breakpoint as a first radius. The target image block region within ±90° of the first radius is obtained with each breakpoint as the center as a first search region. A distance of 2 times the side length of the target image block is extended along the slope direction of each breakpoint as a second radius. Multiple target image block regions within ±90° of the second radius are obtained with each breakpoint as the center as a second search region.

[0064] S5: Use the topology network to obtain the first associated point in the first search area that is connected to the breakpoint, and the second associated point in the second search area that is connected to the breakpoint. Use the first associated point and the second associated point to generate the road detection result.

[0065] Specifically, in this embodiment, all first road nodes within the first search area of ​​each breakpoint are obtained, and the connectivity between each obtained first road node and the corresponding breakpoint is determined using a topology network to filter out first associated points with correlation; all second road nodes within the second search area of ​​each breakpoint are obtained, and the connectivity between each obtained second road node and the corresponding breakpoint is determined using a topology network to filter out second associated points with connectivity.

[0066] A complete road detection map is formed using the first and second association points. (See [link]) Figures 4(a)-4(b)As shown in Figure 4(a), a schematic diagram of the mask before local road filling is presented, and Figure 4(b) shows a schematic diagram of the effect after local road filling is presented. In this embodiment, all roads are also cut into independent line segments using a skeletal map, mainly by cutting the line segments with a degree greater than 2. Then, the width of each skeletal point is obtained according to the road mask. The average width of each line segment is calculated using the line segment length and the width of the point as the width of the road. The road is connected by a breakpoint, and the width of the breakpoint is used as the width of the filled road.

[0067] Specifically, in this invention, by acquiring the association relationships between multiple road nodes, a topological network representing road connectivity is constructed to enhance the representation of road connectivity when roads are occluded by shadows or other factors. At the same time, through the hierarchical setting of local optimization in the first search region and global optimization in the second search region, the efficiency of association point search is increased by utilizing association point search within a single image block and association point search between multiple image blocks. This solves the problem that existing technologies neglect road connectivity in road detection, leading to reduced accuracy of detection results.

[0068] Example 2

[0069] See Figure 5 The present invention also provides a directional connectivity road detection system, which is used in any of the above-described directional connectivity road detection methods, the system comprising:

[0070] The road node acquisition module 100 is used to infer sample image blocks using a trained road extraction model to obtain road mask vectors; to obtain road centerlines using road mask vectors, and to acquire multiple road nodes based on road centerlines to construct discontinuous roads; wherein, the road nodes include discontinuities and intersections; it is also used to acquire multiple target image blocks of remote sensing images of the target area, and to extract the target road mask and road nodes of the road centerline in each target image block;

[0071] The topology network construction module 200 is used to obtain the relationship between multiple road nodes by comparing disconnected roads and centerline masks, and to construct a topology network with road mask vectors as nodes and the relationship between road nodes as edges.

[0072] The search region construction module 300 is used to obtain the slope of the road where the breakpoint is located at each breakpoint, construct a first search region corresponding to each breakpoint within each target image block based on the slope direction, and construct a second search region corresponding to each breakpoint between each target image block.

[0073] The target road detection module 400 is used to obtain a first associated point in a first search area that is connected to the breakpoint, and a second associated point in a second search area that is connected to the breakpoint, using the topology network, and to generate road detection results using the first associated point and the second associated point.

[0074] It should be noted that the modules in the system of Embodiment 2 correspond to the steps in the method of Embodiment 1. The steps in the method of Embodiment 1 have been described in detail in Embodiment 1, and the module content in the system will not be described in detail in this Embodiment 2.

[0075] Example 3

[0076] See Figure 6 This embodiment also provides a computer device, including a system memory 1005 and a processor 1001. The system memory 1005 stores a computer program, and the processor 1001 executes the computer program to implement the steps of any of the methods described above.

[0077] It should be noted that the processor 1001 is used to execute the steps in the above method embodiments according to the instructions in the program code. Alternatively, when the processor 1001 executes the computer program, it implements the functions of each module / unit in the above system / device embodiments.

[0078] Specifically, in this embodiment, the computer program can be divided into one or more modules / units. One or more modules / units are stored in the system memory 1005 and executed by the processor 1001 to complete this application. One or more modules / units can be a series of computer program instruction segments capable of performing specific functions, which describe the execution process of the computer program in the terminal device.

[0079] The terminal device can be a desktop computer, laptop, handheld computer, or cloud server, etc. The terminal device may include, but is not limited to, a processor 1001 and a system memory 1005. Those skilled in the art will understand that this does not constitute a limitation on the terminal device, which may include more or fewer components than shown, or a combination of certain components, or different components. For example, the terminal device may also include an input / output device 1003, a network access device 1002, a bus 1006, etc.

[0080] The processor 1001 can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor.

[0081] System memory 1005 can be an internal storage unit of the terminal device, such as a hard drive or RAM. System memory 1005 can also be a storage device 1004 of the terminal device, such as an external hard drive, SmartMedia Card (SMC), Secure Digital (SD) card, or FlashCard. Furthermore, system memory 1005 can include both internal storage units and storage device 1004. System memory 1005 is used to store computer programs and other programs and data required by the terminal device. System memory 1005 can also be used to temporarily store data that has been output or will be output.

[0082] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0083] Example 4

[0084] This embodiment provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of any of the methods described above.

[0085] The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or any combination thereof. More specific examples of computer-readable storage media (a non-exhaustive list) include: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM), registers, hard disks, optical fibers, compact disc read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof, or any other form of computer-readable storage medium in the art.

[0086] An exemplary storage medium is coupled to a processor, enabling the processor to read information from and write information to the storage medium. Of course, the storage medium can also be a component of the processor. The processor and storage medium can reside within an application-specific integrated circuit (ASIC). In embodiments of the invention, the computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, system, or device.

[0087] Example 5

[0088] This embodiment also provides a computer program product containing instructions that, when executed by a cluster of computer devices, cause the cluster of computer devices to perform the method described in Embodiment 1.

[0089] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for detecting directional connectivity roads, characterized in that the method... include: The trained road extraction model is used to infer sample image blocks and obtain road mask vectors. The road centerline is obtained using a road mask vector, and multiple road nodes are obtained based on the road centerline to construct a discontinuous road; wherein, the road nodes include discontinuities and intersections; By comparing disconnected roads with centerline masks, the relationships between multiple road nodes are obtained, and a topological network is constructed using road mask vectors as nodes and the relationships between road nodes as edges. Acquire multiple target image patches from the remote sensing image of the target area, and extract the target road mask and road nodes of the road centerline in each target image patch; At each breakpoint, the slope of the road where the breakpoint is located is obtained. Based on the slope direction, a first search region corresponding to each breakpoint is constructed within each target image block, and a second search region corresponding to each breakpoint is constructed between each target image block. The first associated point connected to the breakpoint within the first search area and the second associated point connected to the breakpoint within the second search area are obtained using the topology network. The road detection results are generated using the first and second associated points.

2. The method for detecting directional connectivity roads according to claim 1, characterized in that, Before using the trained road extraction model to infer sample image patches and obtain road mask vectors, the method also includes: Obtain remote sensing images of the sample area and road masks and centerline masks of roads in the sample area remote sensing images. Cut the sample remote sensing images of the area to obtain multiple sample image blocks and the corresponding road masks and centerline masks of the sample image blocks, and construct a training set. The sample image patches in the training set are input into the semantic segmentation model to extract the road extraction mask. The road extraction mask is then compared with the road mask in the training set. Based on the comparison results, the model is iteratively trained to obtain the optimal weights.

3. The method for detecting directional connectivity roads according to claim 1, characterized in that, The breakpoint refers to a point on the road centerline with a degree of 1; the intersection refers to a point on the road centerline with a degree greater than 2.

4. The method for detecting directional connectivity roads according to claim 1, characterized in that, Based on the slope direction, a first search region corresponding to each breakpoint is constructed within each target image block, and a second search region corresponding to each breakpoint is constructed between each target image block. Specifically: Extend a distance of 0.1 times the side length of the target image block along the slope direction of each breakpoint as the first radius, and take the target image block area within ±90° of the first radius as the center of each breakpoint as the first search area; Extend a distance of twice the side length of the target image block along the slope direction of each breakpoint as the second radius, and obtain multiple target image block regions within ±90° of the second radius with each breakpoint as the center as the second search area.

5. The method for detecting directional connectivity roads according to claim 2, characterized in that, Obtain the remote sensing image of the sample area and the road mask and centerline mask of the roads in the remote sensing image of the sample area, specifically as follows: The remote sensing images of the sample area and the corresponding road SHP annotation files are obtained. The road centerline is extracted from the road SHP annotation files using vector data processing tools, and the road is converted into a binary mask to obtain the road mask and the centerline mask.

6. The method for detecting directional connectivity roads according to claim 2, characterized in that, The topology network is used to obtain the first associated point in the first search region that is connected to the breakpoint, and the second associated point in the second search region that is connected to the breakpoint, specifically as follows: Obtain all first road nodes within the first search area of ​​each breakpoint, and use the topology network to determine the connectivity between each obtained first road node and the corresponding breakpoint in order to filter out the first associated points with correlation. Obtain all second road nodes within the second search area of ​​each breakpoint, and use the topology network to determine the connectivity between each obtained second road node and the corresponding breakpoint, so as to filter out the second associated points with connectivity.

7. A directional connectivity road detection system, characterized in that, This system is used in a directional connectivity road detection method according to any one of claims 1-6, the system comprising: The road node acquisition module is used to infer sample image blocks using a trained road extraction model to obtain road mask vectors; to obtain road centerlines using road mask vectors, and to obtain multiple road nodes based on road centerlines to construct discontinuous roads; wherein, the road nodes include discontinuities and intersections; it is also used to acquire multiple target image blocks of remote sensing images of the target area, and to extract the target road mask and road nodes of the road centerline in each target image block; The topology network construction module is used to obtain the relationship between multiple road nodes by comparing disconnected roads and centerline masks, and to construct a topology network with road mask vectors as nodes and the relationship between road nodes as edges. The search region construction module is used to obtain the slope of the road where the breakpoint is located at each breakpoint, construct the first search region corresponding to each breakpoint within each target image block based on the slope direction, and construct the second search region corresponding to each breakpoint between each target image block. The target road detection module is used to obtain a first associated point in a first search area that is connected to the breakpoint, and a second associated point in a second search area that is connected to the breakpoint, using the first and second associated points to generate road detection results.

8. A computer device comprising a system memory and a processor, wherein the system memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method described in any one of claims 1 to 6.

10. A computer program product containing instructions, characterized in that, When the instructions are executed by a cluster of computer devices, the cluster of computer devices causes the cluster of computer devices to perform the method as described in any one of claims 1 to 6.

Images