A mountainous area bridge identification method, device, equipment and medium

Abstract

The application discloses a mountainous area bridge identification method and device, equipment and medium, and relates to the technical field of remote sensing. The method comprises the following steps: acquiring first image, second image and DEM data of a target area; performing semantic segmentation processing on the first image and the second image based on a pre-trained semantic segmentation network to obtain a first segmentation mask and a second segmentation mask respectively; obtaining a mountainous area road binary image by taking the intersection of the first segmentation mask and the second segmentation mask; and obtaining mountainous area road vector data through morphological processing; extracting a road center line based on the mountainous area road vector data, obtaining the elevation values of each sampling point on the road center line according to the DEM data; traversing each sampling point along the road center line, calculating the elevation difference between adjacent sampling points, determining the bridge starting point according to the elevation difference, combining the highway construction specification to determine the bridge ending point, and obtaining the bridge area according to the bridge starting point and the bridge ending point. The method realizes high adaptability and high accuracy identification of bridges in a special mountainous area environment.

Description

Technical Field

[0001] This invention relates to the field of remote sensing technology, and in particular to a method, device, equipment and medium for identifying bridges in mountainous areas. Background Technology

[0002] Bridges, as crucial transportation infrastructure connecting both sides of mountains, occupy an important position in mountainous transportation networks. Located in mountainous areas, bridges are susceptible to deformation and even collapse due to complex terrain, extreme weather, and geological disasters. Therefore, accurately identifying and extracting data on mountain bridges is of great significance for providing reliable data support for disaster early warning and infrastructure maintenance in mountainous regions.

[0003] Mountainous areas are frequently shrouded in fog and have undulating terrain, which limits optical imaging conditions. Existing bridge recognition technologies have significant shortcomings when applied to mountainous scenarios: on the one hand, optical satellite remote sensing images are obscured by fog, making it difficult to clearly present bridge structural information; on the other hand, traditional image recognition methods cannot effectively cope with complex weather conditions and are unable to effectively distinguish bridges from ordinary roads, resulting in problems such as low recognition accuracy and high false positive rate. Moreover, the accuracy of model recognition heavily depends on the quality of sample data, and the training samples cannot completely cover all bridge types, leading to poor adaptability of the recognition algorithm.

[0004] For example, Chinese invention patent CN118732987A (Invention title: Remote sensing image bridge detection method based on deep learning semantic segmentation) uses natural water areas for bridge detection, making it difficult to identify bridges in non-water environments such as mountain valleys, and thus unable to meet the diverse recognition needs of complex mountain terrain. Chinese invention patent CN118823481A (Invention title: A bridge detection image recognition method and system based on artificial intelligence) improves image recognition performance under hazy weather by evaluating and enhancing visibility of the original bridge image, but still cannot fully extract the bridge structure in the cloudy weather common in mountainous areas.

[0005] Therefore, there is an urgent need for a bridge identification method that is highly adaptable and accurate for the special environment of mountainous areas. Summary of the Invention

[0006] This invention provides a method, apparatus, equipment, and medium for identifying bridges in mountainous areas to solve the aforementioned technical problems.

[0007] This invention is achieved through the following technical solution:

[0008] A first aspect of the present invention provides a method for identifying bridges in mountainous areas, comprising:

[0009] Acquire the first image, second image, and DEM data of the target area;

[0010] The first image and the second image are semantically segmented based on a pre-trained semantic segmentation network to obtain the first segmentation mask and the second segmentation mask for mountain roads, respectively.

[0011] The intersection of the first segmentation mask and the second segmentation mask is used to obtain a binary image of the mountain road.

[0012] Morphological processing is performed on the binary image of the mountain road to obtain the vector data of the mountain road;

[0013] Based on the mountain road vector data, the road centerline is extracted, and the elevation values ​​of each sampling point on the road centerline are obtained according to the DEM data.

[0014] Traverse each sampling point along the centerline of the road, calculate the elevation difference between adjacent sampling points, determine the bridge starting point based on the elevation difference, and determine the bridge ending point in conjunction with highway construction specifications. Obtain the bridge area based on the bridge starting point and bridge ending point.

[0015] This invention overcomes the limitations of clouds, fog, and terrain by processing image data from different sources separately, providing a comprehensive and accurate data foundation for road and bridge identification. By performing semantic segmentation on mountain roads and taking the intersection of multi-source segmentation results, it provides an accurate basis for bridge identification. Compared with semantic segmentation schemes that can only target specific types of bridges, it also enhances the adaptability of bridge identification in mountainous areas. Finally, it filters bridge areas based on both elevation data and highway construction specifications, utilizing the characteristics of mountain roads and bridge construction to accurately distinguish between bridges and roads.

[0016] Furthermore, based on a pre-trained semantic segmentation network, semantic segmentation processing is performed on the first image and the second image to obtain a first segmentation mask and a second segmentation mask for mountain roads, respectively, including:

[0017] The first image is sliced ​​to obtain multiple first image blocks of a predetermined size;

[0018] The first image block is semantically segmented using a pre-trained first semantic segmentation network to obtain the first segmentation mask for mountain roads.

[0019] The second image is sliced ​​to obtain multiple second image blocks of a predetermined size;

[0020] The second image block is semantically segmented by a pre-trained second semantic segmentation network to obtain the second segmentation mask for mountain roads.

[0021] Furthermore, the first semantic segmentation network is trained using a first image dataset, which includes multiple first image slices of predetermined size and a real mask corresponding to each first image slice.

[0022] The second semantic segmentation network is trained using a second image dataset, which includes multiple second image slices of predetermined size and a real mask corresponding to each second image slice.

[0023] The real mask is obtained by annotating the mountain road scene in the first image slice and the second image slice.

[0024] Furthermore, the first image is an optical remote sensing image, and the second image is a SAR image.

[0025] Furthermore, determining the bridge starting point based on the elevation difference includes:

[0026] If the elevation difference of the current sampling point is greater than the adaptive elevation difference threshold for the first time, then the current sampling point will be taken as the starting point of the bridge; where the elevation difference of the current sampling point is the elevation value of the current sampling point minus the elevation value of the next sampling point;

[0027] The adaptive elevation difference threshold is calculated as follows:

[0028]

[0029]

[0030] in, This represents the adaptive elevation difference threshold. It is the standard deviation of the elevation difference of all sampling points within a sliding window [1, m] centered on the current sampling point. This is an empirical coefficient used to amplify the standard deviation to adapt to actual terrain detection needs;

[0031] This represents the mean elevation difference of all sampling points within the sliding window [1, m]. This represents the elevation difference of the i-th sampling point within the sliding window [1, m]. , , These are the elevation values ​​of the i-th sampling point and the (i+1)-th sampling point, respectively.

[0032] Furthermore, the bridge endpoint is determined in accordance with highway construction specifications, including:

[0033] S2-1, Taking the starting point of the bridge as the starting point of the candidate region, and taking the next sampling point along the traversal direction as the ending point of the candidate region, to obtain the current candidate region;

[0034] S2-2, calculate the slope of the current candidate region and determine whether it meets the bridge deck slope requirements of highway construction specifications; and,

[0035] Determine whether the elevation value of the endpoint of the current candidate region meets the bridge endpoint condition, i.e.:

[0036]

[0037] in, This represents the elevation value of the endpoint of the current candidate region. This is the elevation value at the starting point of the bridge;

[0038] , The length of the current candidate region along the road centerline. To meet the bridge deck slope specifications in accordance with highway construction standards;

[0039] S2-3: If the current candidate region satisfies the bridge deck slope condition and the bridge endpoint condition, then the endpoint of the current candidate region is determined as the bridge endpoint; otherwise, take the next sampling point of the endpoint of the current candidate region as the endpoint of the candidate region along the traversal direction, update the current candidate region, and repeat steps S2-2 to S2-3.

[0040] Further, by traversing each sampling point along the road centerline, the elevation difference between adjacent sampling points is calculated. The bridge starting point is determined based on this elevation difference, and the bridge ending point is determined in conjunction with highway construction specifications. The bridge area is then obtained based on the bridge starting point and the bridge ending point, including:

[0041] Traverse each sampling point along the centerline of the road in the forward direction, calculate the elevation difference between adjacent sampling points, determine the forward starting point of the bridge based on the elevation difference, and determine the forward ending point of the bridge in combination with the highway construction specifications. The first bridge area is obtained based on the forward starting point and the forward ending point of the bridge.

[0042] Traverse each sampling point in reverse along the centerline of the road, calculate the elevation difference between adjacent sampling points, determine the reverse starting point of the bridge based on the elevation difference, and determine the reverse ending point of the bridge based on the highway construction specifications. Obtain the second bridge area based on the reverse starting point and the reverse ending point of the bridge.

[0043] The intersection of the first bridge region and the second bridge region is taken to obtain the final bridge region identification result.

[0044] A second aspect of the present invention provides a bridge identification device for mountainous areas, comprising:

[0045] The data acquisition module is used to acquire the first image, the second image, and the DEM data of the target area;

[0046] The road segmentation module is used to perform semantic segmentation processing on the first image and the second image based on a pre-trained semantic segmentation network to obtain a first segmentation mask and a second segmentation mask for mountain roads, respectively.

[0047] The first image processing module is used to take the intersection of the first segmentation mask and the second segmentation mask to obtain a binary image of the mountain road.

[0048] The second image processing module is used to perform morphological processing on the binary image of the mountain road to obtain vector data of the mountain road.

[0049] The road extraction module is used to extract the road centerline based on the mountain road vector data and obtain the elevation values ​​of each sampling point on the road centerline according to the DEM data.

[0050] The bridge identification module is used to traverse each sampling point along the centerline of the road, calculate the elevation difference between adjacent sampling points, determine the bridge starting point based on the elevation difference, and determine the bridge ending point in conjunction with highway construction specifications, and obtain the bridge area based on the bridge starting point and bridge ending point.

[0051] A third aspect of the present invention provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the mountain bridge identification method according to any one of the first aspects of the present invention.

[0052] A third aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the mountain bridge identification method according to any one of the first aspects of the present invention.

[0053] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0054] By integrating multi-source data with innovative recognition algorithms, the system enhances its adaptability to severe weather conditions such as cloud and fog in mountainous areas, and solves the problem of difficulty in extracting bridge structural information due to cloud and fog obstruction.

[0055] By using elevation data to confirm the starting point of the bridge and considering the characteristics of bridge construction in mountainous areas and the relationship with the terrain, a secondary constraint is applied to the end point range of the bridge to obtain the final bridge identification result. This enhances the adaptability of the bridge identification algorithm and effectively reduces the false positive rate and false negative rate. Attached Figure Description

[0056] 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:

[0057] Figure 1This is a flowchart of a method for identifying bridges in mountainous areas according to an embodiment of the present invention;

[0058] Figure 2 This is a flowchart of a method for determining the end point of a bridge according to highway construction specifications, according to an embodiment of the present invention.

[0059] Figure 3 This is a flowchart of a method for bidirectional traversal identification of bridge regions according to an embodiment of the present invention;

[0060] Figure 4 This is a schematic diagram of a mountain bridge identification device according to an embodiment of the present invention. Detailed Implementation

[0061] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the embodiments and accompanying drawings. The illustrative embodiments and descriptions of this invention are only for explaining this invention and are not intended to limit this invention.

[0062] It should be noted that the terms "comprising" and "having" and any variations thereof in the specification, claims and accompanying drawings of this invention are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to or includes other steps or units inherent to the device.

[0063] The terminology used in the various embodiments of the invention is for the purpose of describing particular embodiments only and is not intended to limit the various embodiments of the invention. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Unless otherwise defined, all terms used herein (including technical and scientific terms) have the same meaning as commonly understood by one of ordinary skill in the art to which the various embodiments of the invention pertain. The terms (such as those defined in commonly used dictionaries) are to be interpreted as having the same meaning as in the context of the relevant technical field and are not to be interpreted as having an idealized or overly formal meaning, unless clearly defined in the various embodiments of the invention.

[0064] This invention aims to address the significant shortcomings of existing bridge recognition technologies when applied to mountainous scenarios due to the unique climate and geographical environment of mountainous areas. It proposes a bridge recognition method based on multi-source image data and highway construction standards. This method not only improves the accuracy of bridge recognition in mountainous areas but also adapts to the diverse bridge structures in mountainous regions, reducing the false positive and false negative rates.

[0065] The key to this task is to accurately identify bridges and extract their distribution under conditions of frequent cloud cover and complex geographical features in mountainous areas. To approximate the true bridge distribution, the task first requires enhancing adaptability to adverse weather conditions such as cloud cover and fog in mountainous areas, addressing the challenge of extracting bridge structural information due to cloud obstruction. Secondly, it necessitates overcoming the limitations of geographical factors to achieve the identification of various types of bridges in mountainous regions. Finally, it aims to improve the accuracy of bridge identification in mountainous areas, thereby providing precise and reliable bridge identification results for practical applications such as disaster early warning and infrastructure maintenance in mountainous regions.

[0066] To address severe weather conditions, this invention employs a multi-source data combination to enhance the representation of bridge structural information. To improve the algorithm's adaptability and accuracy, a deep learning network is used to extract information about mountain roads from the data source. Simultaneously, a bridge extraction method based on elevation data and highway construction standards is constructed. Using mountain roads as a foundation, elevation data is used to confirm the bridge's starting point, and the endpoint range of the bridge is subject to secondary constraints based on bridge construction standards, thereby obtaining the final bridge identification result. For the specific identification process, please refer to [link / reference]. Figure 1 As shown, the steps include the following.

[0067] S1, acquire the first image, second image and DEM data of the target area.

[0068] The first and second images are image data from different sources. The first image provides high-resolution structural details of the mountainous terrain, such as high-resolution optical imagery, while the second image provides comprehensive monitoring data to withstand clouds, fog, and severe weather, such as infrared and radar imagery. This multi-source data complementarity provides a comprehensive data foundation for bridge identification. Digital Elevation Model (DEM) data provides a reliable basis for subsequent accurate extraction of bridge structures, thereby improving data quality.

[0069] The first image, second image, and DEM data ensure coverage of the target mountainous area and potential bridge locations, and the acquired images are preprocessed, such as image correction, noise reduction, interpolation, and image alignment.

[0070] S2, based on the pre-trained semantic segmentation network, performs semantic segmentation processing on the first image and the second image to obtain the first segmentation mask and the second segmentation mask for the mountain road, respectively.

[0071] Unlike conventional semantic segmentation, the present invention identifies mountain roads in multi-source data. It only requires training the segmentation model to identify mountain roads, simplifying the processing of sample data and model training.

[0072] Semantic segmentation network models such as FCN, SegNet, U-Net, and SPIN are trained on a mountain road sample set before being applied. For different image data sources, corresponding image sample data are used for separate training to obtain semantic segmentation networks applied to the first image and those applied to the second image.

[0073] S3. Take the intersection of the first segmentation mask and the second segmentation mask to obtain a binary image of the mountain road.

[0074] This step fuses the segmentation results from different images to compensate for the deficiencies of each data source, resulting in more accurate road identification results in mountainous areas, and thus providing reliable basic data for bridge identification.

[0075] S4 performs morphological processing on the binary image of mountain roads to obtain vector data of mountain roads.

[0076] Morphological processing includes removing small noise areas, filling road voids, and connecting broken road segments to ultimately obtain complete and accurate vector data of mountain roads.

[0077] S5. Extract the road centerline based on mountain road vector data, and obtain the elevation values ​​of each sampling point on the road centerline based on DEM data.

[0078] This step combines mountain road vector data and DEM data to extract the elevation sequence of the road centerline. First, the road centerline is extracted based on the mountain road vector data. Then, the road centerline is sampled at fixed intervals (e.g., 30m) to obtain a series of sampling points. Based on the coordinates of these sampling points, the elevation values ​​of each point are extracted from the DEM data, forming the elevation sequence along the road centerline. , This represents the total number of sampling points along the road centerline.

[0079] S6. Traverse each sampling point along the centerline of the road, calculate the elevation difference between adjacent sampling points, determine the bridge starting point based on the elevation difference, and determine the bridge ending point in conjunction with the highway construction specifications. Obtain the bridge area based on the bridge starting point and the bridge ending point.

[0080] Based on the unique characteristics of abrupt elevation changes and gentle bridge surface at both ends of a mountain bridge, and the gradual elevation changes of ordinary mountain roads extending along the undulating terrain, this invention identifies the starting point of the bridge by quantifying the differences in elevation changes and determining candidate areas in conjunction with highway construction specifications, ultimately accurately pinpointing the bridge's location.

[0081] The elevation difference between adjacent sampling points is calculated as follows: , This represents the elevation difference of the i-th sampling point. , These represent the elevation values ​​of the i-th sampling point and the (i+1)-th sampling point, respectively. .

[0082] If the elevation difference of the current sampling point meets the threshold condition for a sudden change in altitude, then the sampling point is taken as the starting point of the bridge area. In combination with highway construction specifications, such as the bridge deck slope not exceeding 6 degrees and the bridge end elevation value meeting the set conditions, then the range from the starting point to the end point is identified as the bridge area. If the bridge deck or end point conditions are not met, then the starting point is an incorrect bridge starting point, and there is no bridge area in this section of the road. Continue to identify other roads.

[0083] In a further improved implementation, step S2 performs semantic segmentation processing on the first image and the second image based on a pre-trained semantic segmentation network to obtain a first segmentation mask and a second segmentation mask for mountain roads, respectively. Specifically, (1) the first image is sliced ​​to obtain multiple first image blocks of a predetermined size, and each first image block is semantically segmented using a pre-trained first semantic segmentation network to obtain a first segmentation mask for mountain roads; (2) the second image is sliced ​​to obtain multiple second image blocks of a predetermined size, and each second image block is semantically segmented using a pre-trained second semantic segmentation network to obtain a second segmentation mask for mountain roads.

[0084] Furthermore, the training method is as follows: a basic segmentation network model is trained using the first image dataset to obtain the first semantic segmentation network; the same basic segmentation network model is trained using the second image dataset to obtain the second semantic segmentation network.

[0085] The first image dataset is constructed using mountain road images from the same source as the first image dataset. The acquired mountain road images are sliced ​​into multiple first image slices of predetermined sizes, and these slices are labeled with mountain road information. The resulting ground truth mask data serves as the sample labels. Therefore, the first image dataset includes multiple first image slices of predetermined sizes and a ground truth mask corresponding to each first image slice. Similarly, the second image dataset is constructed using mountain road images from the same source as the second image dataset, employing the same method as the first image dataset.

[0086] Preferably, the first image is an optical remote sensing image, and the second image is a SAR image.

[0087] Optical remote sensing relies on the reflection of sunlight in the visible and near-infrared bands. Optical sensors capture the light reflected from the Earth's surface to generate high-resolution, color-rich images. These images, especially on clear days, clearly show the Earth's surface landscape and provide high-resolution details of road and bridge structures. Synthetic Aperture Radar (SAR) is an active Earth observation system that generates images by transmitting microwave signals and receiving their echoes. Because SAR is independent of sunlight, it can operate in any weather conditions, including nighttime and foggy weather. SAR images typically have high resolution and rich texture information, revealing the shape, structure, and surface roughness of surface objects, and providing comprehensive information on road and bridge structures.

[0088] Taking the combination of optical remote sensing imagery and SAR imagery as an example, a mountain road dataset based on optical remote sensing imagery and a mountain road dataset based on SAR imagery are constructed. The SPIN semantic segmentation network is used as the basic recognition model to train the first semantic segmentation network and the second semantic segmentation network respectively.

[0089] (1) Construction of mountain road dataset: Optical remote sensing images and SAR images with similar time phases are sliced ​​into 1024×1024 image blocks. For each image block, typical mountain road scene samples are labeled to obtain a real mask. Each sample contains a 1024×1024 high-resolution remote sensing image, a 1024×1024 SAR image, and a 1024×1024 real mask, and finally a multi-source mountain road dataset is obtained.

[0090] (2) Extracting mountain roads based on multi-source data: The multi-source mountain road dataset is divided to obtain an optical remote sensing image dataset (including multiple 1024×1024 optical remote sensing image slices and corresponding real masks) and a SAR image dataset (including multiple 1024×1024 SAR image slices and corresponding real masks). The SPIN semantic segmentation network is trained twice to obtain optical training weights (corresponding to the first semantic segmentation network) and SAR training weights (corresponding to the second semantic segmentation network).

[0091] SPIN uses a graph reasoning module to construct a fully connected graph, capturing the locational dependencies of different regions through a spatial similarity matrix, thus enhancing the connectivity of road segments. The similarity matrix calculation employs dot product operations, combined with 1×1 convolution and global pooling, to project features into the interaction space. SPIN distinguishes roads from other features (such as buildings and trees) using a learnable similarity matrix, achieving accurate semantic segmentation boundaries.

[0092] The first semantic segmentation network is used to infer the optical remote sensing image of the target area, and the second semantic segmentation network is used to infer the SAR image of the target area. The inference results include a first segmentation mask based on the optical remote sensing image and a second segmentation mask based on the SAR image.

[0093] In a further improved implementation, an adaptive elevation difference threshold is used to determine the bridge starting point.

[0094] Select a sliding window centered on the current sampling point and calculate the standard deviation of the elevation difference sequence within that window:

[0095]

[0096] in, It is the standard deviation of the elevation difference of all sampling points within a sliding window [1, m] centered on the current sampling point. This represents the mean elevation difference of all sampling points within the sliding window [1, m]. This represents the elevation difference of the i-th sampling point within the sliding window [1, m]. Let be the elevation difference of the i-th sampling point. , These are the elevation values ​​of the i-th sampling point and the (i+1)-th sampling point, respectively.

[0097] Then, based on the standard deviation, an adaptive elevation difference threshold is generated:

[0098]

[0099] in, This represents the adaptive elevation difference threshold. This is an empirical coefficient used to amplify the standard deviation to adapt to actual terrain detection needs.

[0100] Due to the complex terrain in mountainous areas, with some sections having steep slopes and others gentle slopes, using a fixed threshold to determine whether an elevation difference represents a sudden change can easily lead to misjudgments. On ordinary highways, elevation differences might exceed the fixed threshold and be misjudged as bridges; on gentle slopes, small elevation changes where bridges connect to mountains might be below the fixed threshold and thus missed. An adaptive threshold is obtained by amplifying the standard deviation using an empirical coefficient k. Based on the current road segment's fluctuations, elevation differences exceeding normal fluctuations are defined as sudden changes, avoiding the influence of the road segment's inherent slope.

[0101] In a further improved implementation, after determining the starting point of the bridge, the method for determining the ending point of the bridge in accordance with highway construction specifications is as follows: Figure 2 As shown.

[0102] S2-1: Take the starting point of the bridge as the starting point of the candidate region, and take the next sampling point along the traversal direction as the ending point of the candidate region to obtain the current candidate region;

[0103] S2-2, calculate the slope of the current candidate region and determine whether it meets the bridge deck slope conditions of the highway construction specifications; and determine whether the elevation value of the endpoint of the current candidate region meets the bridge endpoint conditions, i.e.:

[0104]

[0105]

[0106] in, This represents the elevation value of the endpoint of the current candidate region. The elevation value at the starting point of the bridge. The length of the current candidate region along the road centerline. To meet the bridge deck slope requirements of highway construction specifications, .

[0107] S2-3: If the current candidate region meets the bridge deck slope condition and the bridge endpoint condition, then the endpoint of the current candidate region is determined as the bridge endpoint (i.e., the bridge region identification result is output); otherwise, take the next sampling point of the endpoint of the current candidate region as the endpoint of the candidate region along the traversal direction, update the current candidate region, and repeat steps S2-2 to S2-3.

[0108] The method described in this invention combines elevation data with highway construction standards, identifies the starting point of the bridge based on the abrupt change in altitude, and constrains the end point range of the bridge by limiting the slope of the bridge deck. It selects candidate areas for bridges from a professional perspective of road elevation changes. Compared with existing technologies that rely solely on simple geometric features or a single data source for identification, this method fully considers the characteristics of bridge construction in mountainous areas and the relationship with the terrain, significantly improving the accuracy and reliability of bridge identification.

[0109] In a further improved implementation, the final bridge identification result is determined by taking the intersection of the results from two bidirectional traversals, such as... Figure 3 As shown, the steps include the following.

[0110] S3-1, traverse each sampling point along the centerline of the road in the forward direction, calculate the elevation difference between adjacent sampling points, determine the forward starting point of the bridge based on the elevation difference, and determine the forward ending point of the bridge in combination with the highway construction specifications, and obtain the first bridge area based on the forward starting point and the forward ending point of the bridge.

[0111] S3-2, traverse each sampling point in reverse along the centerline of the road, calculate the elevation difference between adjacent sampling points, determine the reverse starting point of the bridge based on the elevation difference, and determine the reverse ending point of the bridge based on the highway construction specifications, and obtain the second bridge area based on the reverse starting point and the reverse ending point of the bridge.

[0112] S3-3, take the intersection of the first bridge region and the second bridge region to obtain the final bridge region identification result.

[0113] By combining the results of two traversals in both the forward and reverse directions, calculation errors caused by unidirectional data trends can be eliminated, making the identification results more reliable. "Forward" and "reverse" have no specific direction; they simply refer to traversing the elevation data sequence in two directions.

[0114] Embodiments of the present invention also provide a mountain bridge identification device for performing the identification method in any of the above embodiments of the present invention. For example... Figure 4 As shown, the device includes:

[0115] The data acquisition module is used to acquire the first image, the second image, and the DEM data of the target area;

[0116] The road segmentation module is used to perform semantic segmentation processing on the first image and the second image based on a pre-trained semantic segmentation network to obtain the first segmentation mask and the second segmentation mask for mountain roads, respectively.

[0117] The first image processing module is used to take the intersection of the first segmentation mask and the second segmentation mask to obtain a binary image of the mountain road.

[0118] The second image processing module is used to perform morphological processing on the binary image of mountain roads to obtain vector data of mountain roads.

[0119] The road extraction module is used to extract the road centerline based on mountain road vector data and obtain the elevation values ​​of each sampling point on the road centerline according to the DEM data.

[0120] The bridge identification module is used to traverse each sampling point along the centerline of the road, calculate the elevation difference between adjacent sampling points, determine the bridge starting point based on the elevation difference, and determine the bridge ending point in conjunction with highway construction specifications. The bridge area is obtained based on the bridge starting point and the bridge ending point.

[0121] Furthermore, the road segmentation module includes a first segmentation module and a second segmentation module. The first segmentation module is used to perform semantic segmentation processing on the first image to obtain a first segmentation mask for mountain roads; the second segmentation module is used to perform semantic segmentation processing on the second image to obtain a second segmentation mask for mountain roads.

[0122] The road segmentation module also includes a slicing module, which is used to divide the first image and the second image into image blocks of a predetermined size, and then input them into the first segmentation module and the second segmentation module for processing.

[0123] Embodiments of the present invention also provide an electronic device, which includes a processor and a memory, wherein the number of processors may be one or more. The memory, as a computer-readable storage medium, can be used to store software programs, computer-executable programs, and modules. The processor executes various functional applications and data processing of the electronic device by running the software programs, instructions, and modules stored in the memory, thereby implementing the mountain bridge identification method of any of the above embodiments of the present invention.

[0124] The memory may primarily comprise a program storage area and a data storage area. The program storage area may store the operating system and at least one application program required for a given function; the data storage area may store data created based on terminal usage. Furthermore, the memory may include high-speed random access memory (RAM) and non-volatile memory, such as at least one disk storage device, flash memory, or other non-volatile solid-state storage device. In some instances, the memory may further include memory remotely located relative to the processor, which can be connected to the electronic device via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks (LANs), mobile communication networks, and combinations thereof.

[0125] Embodiments of the present invention also provide a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the mountain bridge identification method of any embodiment of the present invention.

[0126] The computer storage medium of this invention can be any combination of one or more computer-readable media. A computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. A computer-readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or 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 or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this invention, a 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, apparatus, or device.

[0127] Computer-readable signal media may include data signals propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. Computer-readable signal media may also be any computer-readable medium other than computer-readable storage media, capable of sending, propagating, or transmitting programs for use by or in connection with an instruction execution system, apparatus, or device.

[0128] Embodiments of the present invention also provide a computer program product that, when run on a computer, causes the computer to execute the mountain bridge identification method of any of the above embodiments of the present invention.

[0129] 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 identifying bridges in mountainous areas, characterized in that, include: Acquire the first image, second image, and DEM data of the target area; The first image and the second image are semantically segmented based on a pre-trained semantic segmentation network to obtain the first segmentation mask and the second segmentation mask for mountain roads, respectively. The intersection of the first segmentation mask and the second segmentation mask is used to obtain a binary image of the mountain road. Morphological processing is performed on the binary image of the mountain road to obtain the vector data of the mountain road; Based on the mountain road vector data, the road centerline is extracted, and the elevation values ​​of each sampling point on the road centerline are obtained according to the DEM data. Traverse each sampling point along the centerline of the road, calculate the elevation difference between adjacent sampling points, and if the elevation difference at the current sampling point meets the threshold condition for an elevation change point, then the current sampling point is taken as the starting point of the bridge. The bridge endpoint is determined in accordance with highway construction standards, including: S2-1, Taking the starting point of the bridge as the starting point of the candidate region, and taking the next sampling point along the traversal direction as the ending point of the candidate region, to obtain the current candidate region; S2-2, calculate the slope of the current candidate region and determine whether it meets the bridge deck slope requirements of highway construction specifications; and, Determine whether the elevation value of the endpoint of the current candidate region meets the bridge endpoint condition, i.e.: in, This represents the elevation value of the endpoint of the current candidate region. This is the elevation value at the starting point of the bridge; , The length of the current candidate region along the road centerline. To meet the bridge deck slope specifications in accordance with highway construction standards; S2-3: If the current candidate region satisfies the bridge deck slope condition and the bridge endpoint condition, then the endpoint of the current candidate region is determined as the bridge endpoint; otherwise, take the next sampling point of the endpoint of the current candidate region as the endpoint of the candidate region along the traversal direction, update the current candidate region, and repeat steps S2-2 to S2-3. The bridge area is obtained based on the bridge's starting point and ending point.

2. The method for identifying bridges in mountainous areas according to claim 1, characterized in that, Based on a pre-trained semantic segmentation network, semantic segmentation processing is performed on the first image and the second image to obtain a first segmentation mask and a second segmentation mask for mountain roads, respectively, including: The first image is sliced ​​to obtain multiple first image blocks of a predetermined size; The first image block is semantically segmented using a pre-trained first semantic segmentation network to obtain the first segmentation mask for mountain roads. The second image is sliced ​​to obtain multiple second image blocks of a predetermined size; The second image block is semantically segmented by a pre-trained second semantic segmentation network to obtain the second segmentation mask for mountain roads.

3. The method for identifying bridges in mountainous areas according to claim 2, characterized in that, The first semantic segmentation network is trained using a first image dataset, which includes multiple first image slices of predetermined size and a real mask corresponding to each first image slice. The second semantic segmentation network is trained using a second image dataset, which includes multiple second image slices of predetermined size and a real mask corresponding to each second image slice. The real mask is obtained by annotating the mountain road scene in the first image slice and the second image slice.

4. The method for identifying bridges in mountainous areas according to any one of claims 1-3, characterized in that, The first image is an optical remote sensing image, and the second image is a SAR image.

5. The method for identifying bridges in mountainous areas according to claim 1, characterized in that, Determining the bridge starting point based on the elevation difference includes: If the elevation difference of the current sampling point is greater than the adaptive elevation difference threshold for the first time, then the current sampling point will be taken as the starting point of the bridge; where the elevation difference of the current sampling point is the elevation value of the current sampling point minus the elevation value of the next sampling point; The adaptive elevation difference threshold is calculated as follows: in, This represents the adaptive elevation difference threshold. It is the standard deviation of the elevation difference of all sampling points within a sliding window [1, m] centered on the current sampling point. This is an empirical coefficient used to amplify the standard deviation to adapt to actual terrain detection needs; This represents the mean elevation difference of all sampling points within the sliding window [1, m]. This represents the elevation difference of the i-th sampling point within the sliding window [1, m]. , , These are the elevation values ​​of the i-th sampling point and the (i+1)-th sampling point, respectively.

6. The method for identifying bridges in mountainous areas according to claim 1 or 5, characterized in that, Traverse each sampling point along the centerline of the road, calculate the elevation difference between adjacent sampling points, determine the bridge starting point based on the elevation difference, and determine the bridge ending point in accordance with highway construction specifications. Based on the bridge starting point and bridge ending point, obtain the bridge area, including: Traverse each sampling point along the centerline of the road in the forward direction, calculate the elevation difference between adjacent sampling points, determine the forward starting point of the bridge based on the elevation difference, and determine the forward ending point of the bridge in combination with the highway construction specifications. The first bridge area is obtained based on the forward starting point and the forward ending point of the bridge. Traverse each sampling point in reverse along the centerline of the road, calculate the elevation difference between adjacent sampling points, determine the reverse starting point of the bridge based on the elevation difference, and determine the reverse ending point of the bridge based on the highway construction specifications. Obtain the second bridge area based on the reverse starting point and the reverse ending point of the bridge. The intersection of the first bridge region and the second bridge region is taken to obtain the final bridge region identification result.

7. A bridge identification device for mountainous areas, characterized in that, include: The data acquisition module is used to acquire the first image, the second image, and the DEM data of the target area; The road segmentation module is used to perform semantic segmentation processing on the first image and the second image based on a pre-trained semantic segmentation network to obtain a first segmentation mask and a second segmentation mask for mountain roads, respectively. The first image processing module is used to take the intersection of the first segmentation mask and the second segmentation mask to obtain a binary image of the mountain road. The second image processing module is used to perform morphological processing on the binary image of the mountain road to obtain vector data of the mountain road. The road extraction module is used to extract the road centerline based on the mountain road vector data and obtain the elevation values ​​of each sampling point on the road centerline according to the DEM data. The bridge identification module is used to traverse each sampling point along the center line of the road, calculate the elevation difference between adjacent sampling points, determine the bridge starting point based on the elevation difference, and determine the bridge ending point in conjunction with highway construction specifications, and obtain the bridge area based on the bridge starting point and bridge ending point. The method for determining the bridge starting point based on the elevation difference is as follows: if the elevation difference traversed to the current sampling point meets the threshold condition for the elevation change point, then the current sampling point is taken as the bridge starting point. The method for determining the bridge endpoint in conjunction with highway construction specifications is as follows: S2-1, Taking the starting point of the bridge as the starting point of the candidate region, and taking the next sampling point along the traversal direction as the ending point of the candidate region, to obtain the current candidate region; S2-2, calculate the slope of the current candidate region and determine whether it meets the bridge deck slope requirements of highway construction specifications; and, Determine whether the elevation value of the endpoint of the current candidate region meets the bridge endpoint condition, i.e.: in, This represents the elevation value of the endpoint of the current candidate region. This is the elevation value at the starting point of the bridge; , The length of the current candidate region along the road centerline. To meet the bridge deck slope specifications in accordance with highway construction standards; S2-3: If the current candidate region satisfies the bridge deck slope condition and the bridge endpoint condition, then the endpoint of the current candidate region is determined as the bridge endpoint; otherwise, take the next sampling point of the endpoint of the current candidate region as the endpoint of the candidate region along the traversal direction, update the current candidate region, and repeat steps S2-2 to S2-3.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the mountain bridge identification method according to any one of claims 1-6.

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

Images